File and Print Services Technical Overview

8/10/2019 File and Print Services Technical Overview

1/39


http://msdn.microsoft.com/en-us/library/cc750353.aspx 1/39

File and Print Services Technical Overview5 out of 7 rated this helpful

Archived content. No warranty is made as to technical accuracy. Content may contain URLs that were valid when originally published, but now link to sites or pages that no longer exist.

Abstract

File and print servers are the workhorses of every network. That does not mean, however, that they can be taken for grantedon the contrary, successful network operating systems must

incorporate the latest technological advances into these fundamental services. To retain the competitive position established by Microsoft Windows NT Server 4.0, Windows 2000

Server has updated its file system, indexing and search capabilities, storage services, and printer functions using stateofthe art technology. In addition, file and print operations are now

integrated both with an organization's internal intranet and with the external Internet. You gain new functionality when you add Windows 2000 file and print servers to an existing network

environment, and then gain more features when you upgradeat your own paceto a Windows 2000 network.

All of the changes to file and print services have been made with both network administrators and application developers in mind. Windows 2000 makes managing file and print services

more efficient, and the open architecture of Windows 2000 Server is designed to facilitate thirdparty developers' ef forts to provide additional functionality in response to evolving business

needs.

On This Page

Introduction

File Services

Indexing Service

Storage Management Services

Printing Services

Existing NT File & Print Servers

Summary

Introduction

File and printer sharing, information retrieval, and data storage are among the most f requently used network services. They are therefore crucial factors to consider when choosing a

network operating system.

Microsoft built the Windows 2000 Server operating system from the ground up as an integrated, multipurpose operating system. The operating system design responds to customer

demands for sophisticated but easytomanage file and print services, for integration of Web and media content with file and print information sharing, and for meeting exponential

growth in storage requirements while lowering storage cost. In addition, its open architecture lets thirdparty developers provide additional functionality in response to everchanging

business requirements.

Microsoft developed specific file and print features to meet widespread customer needs:

Reduced cost.Remote Storage migrates infrequently used files to lowercost secondary storage, yet keeps that data available if needed. Removable Storage helps reduce costs by

letting multiple client applications share local libraries and tape or disk drives while ensuring that client applications do not corrupt each other's data.

Better manageability.The improved NTFS file system, distributed file system Dfs, and Indexing Service make it easier to find and access files across expanding networks. New

interfaces make operating system services easier to manage; for example, the new printer interface makes it simpler for both administrators and endusers to configure and manage

their printing needs.


2/39



Increased availability and reliability.Dfs replication and File Replication service FRS synchronization help keep data available to users, even if a server or disk drive fails or a

shared folder or file becomes corrupted. Dynamic volumes formatted with NTFS 5 allow fewer reboots when adding disks and creating, extending, or mirroring a volume.

Scalability.The Windows 2000 NTFS version 5 f ile system and the Windows 2000 storage subsystems let users efficiently store and retrieve everlarger quantities of data.

Organizations that install Windows 2000 file and print servers in their existing network can take advantage of several new features. When they upgrade to a Windows 2000 network,

additional file and print capabilities become available.

This overview focuses primarily on the Windows 2000 Server implementation of the standard file and print services components. However, it includes mention of several Webrelatedfeatures where they are inextricably bound up with file and print services.

Top Of Page

File Services

The majority of network servers provide file service; that is, they offer centralized file storage that lets users easily share files. File servers often store private files as well as shared files, and

provide a single point of backup for both. File servers let users access their files even when they move to different workstations.

Windows 2000 Server introduces new and improved file services, including changes to the management of network shares and users and to the NTFS file system. In addition, Windows

2000 Server supports several other ondisk file system formats. Windows 2000 Dfs extends the capabilities present in Windows NT 4.0 Dfs. Not technically a file system driver, Dfs provides

what appears to users as a unified hierarchical file system, although the data actually resides on different servers across the network.

Network administrators typically install systems whose primary role is providing file service as member servers rather than as domain controllers. All file service features described in thispaper, except for domainbased Dfs and the FRS, are also available on standalone servers.

Windows 2000 Server filesystem related features include:

Managing shares, connected users, and open files

Distributed File System Dfs

NTFS and related enhancements

Other supported file systems

Each of these topics is covered in the following sections.

Managing Shares, Connected Users, and Open Files

The Shared Folders snapin1is a filesystemrelated tool in the Windows 2000 Server collection of System Tools. An updated version of capabilities found in Windows NT Server 4.0 in

Server Manager in the Control Panel, Shared Folders enables the creation of shares, manages the connections on local or remote computers, and displays open files. To open it, click Start,

click Programs, click Administrative Tools, and then click Computer Management. Figure 1 shows Shared Folders among the System Tools available in the Computer Management

console.


3/39



Figure 1: Shared Folders tool manages shares, connected users, and open files

For Windows 2000 Server, members of the Administrators, Server Operators, or Power Users groups can use the Shared Folders snapin. For Windows 2000 Professional, only members of

the Administrators or Power Users group can use Shared Folders.

The Shared Folders snapin lets you perform the following tasks:

Shares.Create, view, and set permissions for network shares, including shares running Windows NT 4.0.

Sessions.View and disconnect users connected to the computer over the network.

Files.View and close files opened by remote users.

Mac Shares.Configure Services for Macintosh so that Windows and Macintosh users can share volumes, files, and printers through a Windows 2000 Server. Mac clients can access

Macintosh volumes and printers using an Apple networking protocol. From the file server, you administer Mac shares from the Services and Applications node of the Computer

Management console tree2.

For Windows 2000 Server, the Shared Folders snapin also enables publishing a share as a Volume Object in the Active Directory directory service. Publishing an object in Active Directory

lets users query available resources and shares.

Distributed File System

The Microsoft distributed file system Dfs for Windows 2000 Server is a secondgeneration technology built on and improving the earlier Windows NT Server 4.0 Dfs. Dfs presents to users

a logical view of distributed physical storage, making both managing and finding network data easier. Dfs is not a new file system but software that gives users a view of what looks likea

unified hierarchical file system, even though the data is in reality distributed in different locations. For example, you can use Dfs to make marketing files scattered across multiple servers in

a domain appear as if all these files reside on a single server. This eliminates the need for users to go to multiple locations on the network to find the information they need. Dfs can

connect hundreds or thousands of published shares in a single logical system.

You use the Dfs Administrator snapin to administer Dfs volumes. Implementing Dfs is not mandatory in Windows 2000 Server, but network administrators should consider doing so if:

The users accessing shared resources are distributed across multiple sites.

Most users require access to multiple file servers.
http://msdn.microsoft.com/en-us/library/Cc750353.file01_big(l=en-us).gif


4/39



Network load balancing can be improved by redistributing shared resources.

Users require uninterrupted access to file servers.

The organization uses either internal or external Web sites see the section "IIS Can Use Dfs" later in this paper for how Dfs can help simplify Web management

.

Dfs is protocol independent, which means that both Windows and nonWindows servers can be included in the Dfs namespace. The Dfs client and server use the Common Internet File

System CIFS

to determine which file server will be accessed by the client. When the client then accesses the target file server, it uses the native protocol to access the file server. If the

server uses a NetWarebased operating system, the client uses NetWare Core Protocol NCP; if the server is Unix, the client uses network file system NFS. That is, clients use NCP or NFS

to connect to the file share afterDfs has directed them to it.

The purpose of Dfs is to let users and applications access files. Dfs is notdesigned to perform operations such as indexing, virus scanning, or backup, because accessing very large numbers

of files in a highly sequential/repetitive manner using Dfs would substantially increase network traffic. In addition, when using Dfs replicas you do not know which particular file server in a

replica set is being accessed, which means that Dfs is not suitable for backup and restore operations.

These Dfs topics are covered in the following subsections:

Dfs topology

Dfs replication

Dfs intelligent client caching

Dfs uses standard security

IIS can use Dfs

Windows 2000 improvements to Dfs compared to Dfs 4.x

Dfs Topology

Dfs root.The share at the top of the Dfs topology, which is the starting point for the Dfs links and shared folders that make up the Dfs namespace.

Dfs link.Located under the Dfs root, a Dfs link forms a connection to one or more domainbased volumes or shared folders, or to another Dfs root. In Dfs 4.xin Windows NT terminology,

a Dfs link is called a child node or junction point.

Dfs shared folder.Files or folders in the Dfs namespace that can be accessed by users who have proper permissions.

Replica set.Two or more Dfs roots or Dfs shared folders that participate in replication.

Replica.A folder within a replica set.

The Dfs topologyis the physical layout depicted in the Dfs Administrator console that consists of a root, links, shared folders, and replicas. The Dfs topology is not the same as the Dfs

namespace, which provides the view of shared resources on the network as seen by users.

Dfs topology begins with the root of the Dfs tree. An administrator maps a Dfs root, which is the top of the logical hierarchy, to a physical share. Currently, you can assign several thousandDfs links to a Dfs root. A Dfs linkthe point where a physical machine boundary is crossedmaps a Domain Name System DNS name to the standard universal naming convention UNC


5/39



name of the target shared folder or target Dfs root. When a Dfs client accesses a Dfs shared folder, the Dfs root server uses this mapping of a DNS name to a UNC name to return a

referral to the client so that it can locate the shared folder.

Mapping the DNS name to the UNC name makes the physical location of data transparent to users, who no longer have to remember on which server a folder is stored. If you move a file

or folder to another physical location, the user's view of it remains unchanged.

When a client machine requests a referral to a Dfs share, the Dfs root server uses the Partition Knowledge Table PKT to direct the client to the physical share. The PKT is stored in Active

Directory for domainbased Dfs and stored in the registry for standalone Dfs more about domainbased Dfs and standalone Dfs later. In a network environment, the PKT maintains all

information about Dfs topology, including its mappings to the underlying physical shares. After the Dfs root server refers the client to a list of replica shares that correspond to the

requested Dfs link, the client then uses Active Directory site topology3to contact a replica within the same site or, if one is not available, a replica outside the site.

Figure 2 shows an example of how an administrator can set up Dfs composed of links from multiple servers, and it shows how an enduser would see the result:

Figure 2: Dfs topology from the point of view of the administrator and the enduser

Only Windows 2000based machines can host Dfs roots and Dfs links Dfs shared folders. NonWindows 2000based machines can be the target of a Dfs link but cannot contain additional

Dfs links although of course they can host filesystem subfolders. Dfs shared folders on downlevel volumes volumes on computers running an earlier operating system than Windows

2000

include those published on Windows NT 4.0 Workstation and Server, Windows 95, Windows 98, Windows for Workgroups, and all nonMicrosoft shared folders for which client

redirectors are available. If duplicates of a shared folder exist, each copy is a replica in a replica set see the section "Dfs Replication" for more about this.

The Dfs root can be one of two types:

Domainbased Dfs root.A domainbased Dfs root must be hosted on a domain member server or domain controller. Such servers, called root replicas, provide referrals to the Dfs

namespace for client machines. Domainbased Dfs stores its configuration information in Active Directory. If you have more than one domain controller to keep the Dfs

configuration information available

and if you establish more than one Dfs server in the domain, domainbased Dfs can provide high availability for any Dfs file or folder in the

domain.

Standalone Dfs root.A standalone Dfs root can be hosted on three types of Windows 2000 servers: on a standalone server, a member server, or on a domain controller. A stand

alone Dfs server does not use Active Directory, cannot have rootlevel replicas, and can have only a single leve l of Dfs links. Standalone Dfs stores its configuration in the local

registry. Its purpose is to provide backward compatibility with earlier versions of Dfs. During an upgrade of Windows NT 4.0 to Windows 2000, any Dfs 4.xroots are converted

automatically to Windows 2000 standalone Dfs roots. You can manage Dfs 4.ximplementations with the Windows 2000 Dfs Administrator. You can migrate standalone Dfs roots at

your own pace. Some roots can remain as standalone; others can be migrated to domainbased. Both can coexist in a Windows 2000 domain.

Each server that hosts a domainbased Dfs root obtains the PKT from the Active Directory4. Thus, each of these servers must stay in sync with the Active Directory. This synchronization of

the Dfs root and Active Directory not the same as synchronization among Dfs replicas, described below is triggered in three ways:


6/39



Startup.When a server hosting a domainbased Dfs root boots, the booting server obtains the PKT from an active domain controller.

Changes.When changes are made to the PKT changes are made in Active Directory itself, all participating Dfs root servers are notified that changes have occurred that the servers

must retrieve from the directory.

Set interval.Domainbased Dfs root servers poll the directory service for current PKT information every 24 hours.

Dfs Replication and FRS Synchronization

Users access certain files frequently over the network. When only one copy of a file exists on a single file server, if that server goes down, no one can access the file. A different problem

arises if many users access a single file simultaneouslythe server experiences a heavy processing load, resulting in slower access to files on that server. Using Dfs replication helps solve

these problems of availability and load. The following subsections describe how Dfs uses replication to accomplish these tasks:

Replica sets provide high availability

FRS keeps replicas that change synchronized

Replica sets help reduce file server load

Replica sets make file server maintenance easier

Replica Sets Provide High Availability

High availabilityrefers to keeping important data available, even if a server or disk drive fails or a shared folder or file becomes corrupted. To ensure that the Dfs root server itself remains

available, you can use Dfs Administrator to create two kinds of replica sets. A replica set is two or more copies of a Dfs root or Dfs shared folder that participate in replication. Here are the

two types of replica sets:

Dfs root replicas replicate Dfs "knowledge." In domainbased Dfs, you can create Dfs root replicas, that is, two or more copies of the Dfs root, each on a different server. As stated

above, a Dfs root replica uses the PKT which contains the Dfs topology, including mappings from the Dfs logical namespace to the underlying physical shares

to provide referrals to

clients. This is why replication of Dfs roots is also referred to as replication of Dfs knowledge.

Dfs link replicas replicate Dfs "content."When you create a Dfs link, you can create a replica set by specifying multiple shared folders for that link. Each copy must be located on a

separate computer, but all share the same logical Dfs name. The fact that more than one copy which can be readwrite copies

exists is transparent to users. You can specify

hundreds of shared folders in a replica set. Just as Dfs root replicas support high availability for Dfs roots, Dfs link replicas support high availability for a portion of the Dfs

namespace.

In Figure 3, below, the box labeled "Replicating Dfs Root" shows two Dfs root servers. The two shares hosting the root are a replica set. Any new content local to those roots is kept

consistent by FRS; see the next subsection, "FRS Keeps Replicas That Change Synchronized." The box labeled "Replicating Dfs Links" shows two shares published as a single Dfs link.

Content on these two shares is also held consistent by a separate FRS replica set.


7/39



Figure 3: File replication of Dfs roots and Dfs links.

As already explained, you create Dfs root replicas to provide high availability for a Dfs root server. However, because the Dfs root uses the PKT, providing high availability for Dfs root

replicas is a more complicated issue than providing high availability for Dfs links. For Dfs links, simply establishing replica set replication provides high availability. For a domainbased Dfs

root, the PKT is stored in and must be fetched from Active Directory. Because each participating Dfs root server sends its PKT changes to Active Directory, all other participating Dfs root

servers must periodically obtain the updated PKT from the directory. Therefore, although you can establish Dfs replication for your network without also setting up Active Directory

replication5, without Active Directory replication it is not possible for Dfs replication to optimally ensure high availability of Dfs data. Active Directory replication is required to ensure the

availability of the PKTif one domain controller fails, Dfs can obtain the PKT from another domain controller.

Establishing replica sets and keeping them synchronized are re lated but separate issues. For information on when and how to implement synchronization, see the next subsection.

FRS Keeps Replicas That Change Synchronized

Windows 2000 Server introduces the File Replication Service FRS to provide full twoway file replication for NTFS 5 volumes. This replication is implemented with NTFS Change Journal

update sequence numbers USNs see also the section "NTFS Change Journal" later in this paper. FRS is a multimaster filebasedreplication service6and uses remote procedure call RPC

based connections. FRS lets you adjust the schedule for file replication and specify what gets replicated and which replica sets need to be kept synchronized.

FRS is usedfor different purposesby both Dfs and by Active Directory:

FRS & Dfs roots.Domainbased Dfs root replicas are always dynamic because at the least they need to keep updating the PKT recall that the PKT, which contains the Dfs

topology, is stored in Active Directory. Therefore, by default, FRS is always activated for domainbased Dfs root replicas.

FRS & Dfs links.By default, FRS is not enabled for Dfs replica sets because data contained in them could be static. In most cases, however, you do want to ensure that the

underlying shared folders for Dfs links are kept synchronized to present the same data to users regardless of the folder they access. Microsoft strongly recommends using FRS for

automatic replication of Dfs shared folders.

An example of when it is not necessary to activate FRS for Dfs link replicas is if you keep readonly shares of sitelicensed software on servers for employees to install when needed.

These applications can be replicated to Dfs share points in multiple sites so that users can install uptodate versions from nearby servers, but it is not necessary to activate FRS

because the applications are readonly.

Whether or not to implement FRS for shared folder replica sets also depends on whether the network has sufficient bandwidth to handle the traffic generated by automatic file

replication.


8/39



FRS & Active Directory.In an Active Directory environment that is, on a Windows 2000 domain controller, certain system files are not stored in the Windows 2000 directory

information tree DIT database and are therefore not replicated by Active Directory object replication. These system files are stored in a shared folder called the system volume

SYSVOL. SYSVOL is installed as a part of the domain controller promotion process Dcpromo.exe. FRS replicates the SYSVOL folder among participating domain controllers and

keeps the replicas synchronized. FRS replication of SYSVOL is an activity that is entirely independent of Active Directory object replication. Note also that, although Active Directory

uses FRS to replicate SYSVOL, Active Directory replicationdoes not use FRS because FRS has its own replication engine.

Figure 4 summarizes features of Dfs and FRS:

Figure 4: Dfs and FRS features

Replica Sets Help Reduce File Server Load

Load sharingautomatically distributes file access across multiple disk drives or servers, thus improving server response time to client requests during peak usage periods. Dfs provides a

degree of load sharing by taking advantage of the share redundancy provided by replica sets just described in the preceding section.

Replica Sets Make File Server Maintenance Easier

Using Dfs replicas, you can often perform file server maintenance, software upgrades, and other tasks without disrupting user access. When two or more root replicas exist on two or more

servers, you can take one server offline for maintenance without disrupting users because the Dfs share remains available on the replica

s

.

Dfs Intelligent Client Caching

When you add a Dfs link, in addition to specifying a name and path for the link, you also specify the duration for which a reference to this Dfs link will be cached locally on a Dfs client. The

first time a client machine requests a referral for a Dfs link, the client contacts the Dfs root server, which uses the PKT stored in Active Directory to direct the client to the share. However,

subsequent requests for the same referral are handled from the client's cache. Therefore, the client does not need to access the Dfs server after the first referral unless the cache entry has

expired. Caching thus allows quick access to frequently used network volumes, improving performance by reducing the processing load on the Dfs server.

Dfs Uses Standard Security

Windows 2000 Dfs saves time for administrators because, other than creating the necessary administrator permissions, the Windows 2000 Dfs service does not require the implementation

of any additional security measures.

Only administrators can make changes to the Dfs topology enterprise and domain administrators for domainbased Dfs; local administrators for standalone Dfs. All other security is
http://msdn.microsoft.com/en-us/library/Cc750353.file04a_big(l=en-us).gif


9/39



enforced by the file system underlying the network share referenced by Dfs. That is, when a user tries to access a Dfs shared folder or the files inside it, whether or not that access succeeds

depends on the underlying file system. A volume formatted with FAT is protected by sharelevel security; a volume formatted with NTFS has both sharelevel and filelevel NTFS

permissions.

IIS Can Use Dfs

Internet Information Services 5.0 IIS, which installs as a networking service of Windows 2000 Server, can take advantage of Dfs to make management of a Web site easier. If all the data for

a website is stored on a single machine, Dfs has no role to play. However, if the HTML files backing a Web server are stored on multiple machines, using Dfs can simplify Web site

management by grafting the multiple network shares into a unified namespace.

A Webmaster can now build a logical Dfs directory that includes the default Web pages of each department's Web server as a subdirectory of the main Internet or intranet Web. If a Web

page is physically moved from one server to another, the HTML links to other pages stored in Dfs do not have to be updated, provided an administrator reconfigures Dfs accordingly using

the Dfs Administrator tool

. This means that if the server hosting a user's Web page is removed and the page is republished on a different server, the links pointing to that page do not

have to be reconfigured.

Windows 2000 Improvements to Dfs When Compared to Dfs 4.x

Dfs server enhancements7include the following:

Windows 2000 installs the Dfs service automatically during installation of or upgrade to the Windows 2000 operating system.

You can pause or stop the Dfs service, but you cannot remove it from the administrative console.

The Dfs topology is stored in Active Directory for domainbased Dfs.

Dfs is integrated into the Active Directory namespace for domainbased Dfs.

Replicated Dfs roots eliminate the root as a single point of failure.

Support for FRS permits automatic replication of file changes between Dfs replicas.

The Dfs administrative tool is now graphical.

Status flags indicate the availability of replicas.

Dfs links can connect to other links on other Windows 2000based servers without a fresh referral.

The expiration of referrals that are cached by Dfs clients is configurable on an individual link basis.

Dfs now supports dynamic configuration of the Dfs topology.

Dfs now supports Cluster service.

NTFS and Related Enhancements

Microsoft first introduced the NTFS file system NTFS, a 64bit advanced file system for storing data on hard disk, with Windows NT 3.1 and now adds significant enhancements to NTFS

in Windows 2000 Server. Windows NT 3.51 and Windows NT 4.0 both use NTFS version 4. Windows 2000 uses NTFS version 5 during setup, NTFS 4 partitions are automatically converted

to NTFS 5. The minimum size for an NTFS volume is 10 megabytes MB; the recommended practical maximum is 2 terabytes. The maximum theoretical file size is 16 exabytes.


10/39



Windows 2000 Server continues to support the advanced features NTFS provided to earlier versions of the operating system, including:

Filelevel access control.NTFS governs which users and groups can access individual files and directories, and it can provide varying levels of access for different users. This is then

enforced by the core operating system. The NTFS file permissions are No Access, List, Read, Add, Add and Read, Change, Full Control, Special Directory Access, and Special File

Access which provides an even greater degree of granularity. Filelevel access control does not include file encryption; for Windows 2000 file encryption, see the section

"Encrypting File System."

Compression.NTFS compression allows for the compressed storage of files and directories so that less physical space is required. Compression is configurable on a volume,

directory, or file basis. With NTFS, if anything goes wrong physically with a portion of data in a compressed file, only that file is affected. This differs from earlier FAT compression on

Windows 95 and Windows 98, which could lose an entire volume of data if even one sector became corrupted.

Recovery log.NTFS logs all changes to the file system so that every file or directory update can be redone or undone to correct discrepancies caused by system failure or power

loss. NTFS cluster remappingcalled sector/cluster hot fixingrepairs hard disk failures on the fly without returning error messages to the calling application. If the data is corrupt,

NTFS flags that part of the hard disk as defective, and then rewrites the data to another location. Recovery log operations are fast and transparent to users.

POSIX support available only when running the POSIX subsystem.NTFS file names support the Portable Operating System Interface for UNIX POSIX standard for network

naming conventions, such as case sensitivity, lastaccess timestamping, and hard links.

New features in Windows 2000 Serverits directory service Active Directory, the IntelliMirror set of management features, and many new storage improvementsrequire an update to

the ondisk format for NTFS. Windows 2000 Server requires the use of the updated NTFS format on all domain controllers. Among the storagerelated improvements that use the new

NTFS format are volume mount points, remote storage, file system encryption, sparse files, disk quotas, and Microsoft Indexing Service all described later.

For existing NTFS volumes, the upgrade to the NTFS 5 ondisk format occurs automatically at the volume's f irst mount time; for volumes that setup uses, this conversion occurs during

installation. In addition, setup asks if you want to convert your FAT and FAT32 volumes, but doing so is optional. You can manually convert FAT and FAT32 volumes to NTFS 5 at any time.

Servers that dual boot Windows 2000 Server and Windows NT Server 4.0 must install Windows NT 4.0 Service Pack 4 or higher in order to access local NTFS 5 volumes when running

Windows NT 4.0. All network clients can remotely access NTFS volumes on Windows 2000 file and print servers. Whether a volume is formatted with NTFS 4 or NTFS 5 is transparent to the

network client.

Applications that assume knowledge of a volume's ondisk format must be updated. To learn more about how to do this, see the section "For More Information."

Enhancements that Windows 2000 Server adds to NTFS include:

NTFS reparse points and file system filter drivers

Encrypting File System EFS

NTFS volume mount points

NTFS sparse file support

Native property sets

Security ID SID searching and bulk access control list ACL checking

NTFS Change Journal

Distributed Link Tracking


11/39



The following subsections describe each of these topics.

NTFS Reparse Points and File System Filter Drivers

Windows 2000 Server introduces reparse points and installable file system filter drivers into its storage subsystem to provide additional features and to give independent software vendors

ISVs

a consistent mechanism for extending storage functionality.

Reparse pointsare NTFS file system objects that have a special file attribute called a reparse tag. One reparse point is allowed per file or directory. Input/output I/O calls use these tags to

differentiate types of reparse points. ISVs must apply to Microsoft for a reparse point tag value.

Installable file system filtersare kernellevel dev ice drivers that let ISVs intercept file system calls. When used in conjunction with ISVprovided file system filter drivers, reparse points offer a

mechanism for providing file and directory enhancements to existing storage applications and enable new types of storage management applications, such as Remote Storage.

The Windows 2000 Server implementation of reparse points and installable file system filter drivers frees ISVs from the need to write proprietary system functionality. They can instead

concentrate on responding directly to customers' business needs. Besides developing reparse points and installable file system filter drivers for thirdparty developers to use, Microsoft

introduces several Windows 2000 Server services of its own based on these features. These include:

Encrypting File System EFS

NTFS volume mount points

Remote Storage

This paper covers the first two of these features in the immediately following two subsections and the last one in the "Storage Management Services" section.

Encrypting File System

With the Encrypting File System EFS, Windows 2000 Server provides protection for sensitive user data stored on the NTFS 5 f ile system. EFS is not a file system, but rather a service that

provides filelevel encryption for locally stored data on an NTFS 5 volume. EFS uses cryptography to complement the existing access control security model on NTFS. This provides a new

level of protection for data stored on disk. The encryption technology uses a combination of public key technologies for key management and symmetric cryptographic algorithms for data

encryption.

In the past, someone with physical access to a computer could bypass earlier NTFS access control security. Therefore, to provide complete file system security it was necessary to ensurethat the machine was not physically accessible to unauthorized people. In Windows 2000 Server, EFS addresses this problem. EFS strengthens NTFS's ex isting security by providing localfile

and directory encryption on NTFS volumes8. EFS is intended to protect only data that you access locally. Data sent across the network is in clear text. If it were possible to access an

encrypted file across the network, that would defeat encryption because a network sniffer could be used to read the data packets.

Users can use Windows Explorer, administrative interfaces, or commandline tools to encrypt and decrypt files and directories they are authorized to use. Users do not need to decrypt a file

before using itthe file is stored encrypted on the disk, and all reads and writes to it are decrypted and reencrypted transparently. To see whether the file is encrypted, users can check

the file properties to see if the encrypted attribute bit is on. Unauthorized users trying to open an encrypted file get an access denied error because they don't possess a key to decrypt the

file.

To read more about EFS, see the section "For More Information."

NTFS Volume Mount Points


12/39



As an alternative to assigning a drive letter to a mounted drive, Windows 2000 can assign a drive path. This means you are no longer limited to 26 drive letters for mounting and accessing

volumes. For example, if you have a CDROM drive currently assigned to the D: drive and you have an NTFSformatted C: drive, you can mount the CDROM drive at an empty folder with

the following path: C:\CDROM. You can then remove the drive letter D: or use it for something else and can access the CDROM through the mounted drive path.

This is made possible by NTFS volume mount points, which are new file system objects. Placing a mount point which is implemented as an NTFS reparse point on a directory maps one

disk volume under a directory of another volume. Because volume mount points are based on NTFS 5 reparse points, they work only on NTFS 5 volumes. Multiple volume mount points can

target any volume. Windows automatically prevents resolution problems due to changes in the internal device name of the target volume for example, changes due to hardware device

reconfiguration

.

Say a laptop user has two logical volumes:he or she uses one to store operating system and personal files and the second to store workrelated data. Most personal productivity tools areset to open/save work at a common directory such as C:\My Documents. It would be convenient not to have to change drives depending on whether personal or workrelated data is

being used. The user can use the Disk Management utility to place an NTFS volume mount point in the C:\My Documents\Work directory so that it and its subdirectories will use physical

disk space on drive 2. Changing directories to C:\My Documents\Personal, however, would access drive 1. Figure 5 depicts this situation.

Figure 5: Example of volume mount points

Although this simple example demonstrates the use of an NTFS volume mount point, the real power of NTFS volume mount points is in enterprise server environments where volumes are

frequently added in response to data growth.

Notes:

You cannot mount a network drive with an NTFS volume mount point; instead, you can use Dfs to do this.

The system does not check for loops in the namespace. Although it is possible to mount a drive underneath itself to create a loop, doing so is not recommended because

applications that do recursive opens against directories using FindFirst/FindNext will go into an infinite loop.

NTFS Sparse File Support

A very large file w ithout a lot of data is said to contain a sparse data set sparse datahas large consecutive areas of 0 bits. Applications that use sparse data sets include image processors

and highspeed databases. In versions of NTFS prior to version 5, the portions of the file that did not contain useful data occupied valuable disk space.

The file compression introduced in NTFS 3.51 was a partial solution to the problem. The portions of the file that do not contain useful data were set to zero, creating large consecutive


13/39



areas of 0 bits, and file compression compacted the nondata portions. However, file compression has its own drawbacks: access time may increase due to data compression and

decompression.

NTFS 5 introduces another solution to the problem of sparse data, which both conserves disk space and improves disk performance. An administrator or application can use a new user

controlled file system attribute to mark files containing large consecutive areas of 0 bits as sparse, and NTFS will then allocate physical space for only the meaningful data that is, to only

those portions of a sparse file that are actually written to. NTFS stores only range information that indicates where the sparse data would be if it were allocated. On file access, the file

system returns allocated data as actual and deallocated data as zeros. APIs let application developers bypass file expansion and read the allocated ranges directly. This enables applications

to avoid processing large streams of zeros yielded by the file system and to copy or move potentially huge files with sparse data streams in an efficient manner.

For example, if data is written to the first 64 KB and last 64 KB of a 42 GB file that is marked as a sparse file, NTFS uses only 128 KB of disk space, although in other respects the filefunctions as if it were 42 GB.

Native Property Sets

Native property sets let any file or folder on an NTFS 5 volume have descriptive metadata associated with it. NTFS 5 supports native property sets on any file or folder. For example,

Component Object Model COM9documents, such as Microsoft Word or Microsoft Excel files, have associated properties, including Author, Title, Subject, and Comment. With NTFS 5, any

file can have associated properties, even singlestream text files.

The new Indexing Service feature in Windows 2000 can index all properties on a file, including both NTFS native properties and COM document properties. The result is that users can

quickly search the index not only for files containing specified words in the file contents but also for propertiesof the filesfor example, they can search on properties such as document

author. To see the properties for a file or directory, rightclick on it and select Properties.

SID Searching and Bulk ACL Checking

In Windows 2000 and Windows NT

a security ID, or SID, is a unique number that identifies each user, group, or computer account to the Windows security systems. Windows issues a SID

to every account on the network when the account is first created. Internal processes in Windows 2000 refer to an account's SID rather than to its user, group, or computer name.

NTFS 5 can perform a volumewide scan for f iles using the owner's SID. This feature lets administrators perform such tasks as finding all files that a given user owns and cleaning up a user's

files.

An access control list ACL allows or denies permissions also called access rights on a file or folder to specific users or groups. File permissions include Full Control, Modify, Read and

Execute, Read, and Write. Folder permissions include Full Control, Modify, Read and Execute, List Folder Contents, Read, and Write.

Windows 2000 offers substantially improved storage for ACLs. Because Windows 2000 uses an ACL index, unique ACLs are stored only once. Any files that use the same ACL share an index

entry, unlike Windows NT 4.0, which stored ACLs on each file. That is, NTFS stores unique ACLs only once even if ten objects share the ACL; in Windows NT 4.0, the ACL would be stored tentimes.

Indexing Service uses the Bulk ACL Checking feature to efficiently check security on all files returned by a search. Using Bulk ACL Checking ensures that a reference to a file the user cannot

read is not returned. In addition, NTFS 5 uses Bulk ACL Checking to test for authorization against multiple files at once. This lets you perform tasks such as determining what a given user

can do with given files or checking multiple ACLs simultaneously for file access.

NTFS Change Journal

Windows 2000 Server introduces the volumewide Change Journal10to track modifications to NTFS 5 files over time and across system reboots. The Change Journal itself is a sparse

stream, which means that only a small active range of the file uses any disk allocation see the earlier section "NTFS Sparse File Support". As files, directories, and other NTFS objects are

added, deleted, and modified, NTFS enters records into the Change Journal in streams, one record for each volume on the computer. Each record indicates the type of change read, write,

move, and so on and the object that was changed. The offset from the beginning of the stream for a particular record is called the update sequence number USN. New records areappended to the end of the stream.


14/39



The Change Journal logs the fact of the change to a file and the reason for it. However, it does not record enough information to reverse the change.

ISV developers of systemlevel applications such as file system indexing engines, content replication engines, and storage archiving and migration can make use of the Change Journal to

provide enhanced functionality: Applications that periodically scan the file system for changes can now, instead, use the Change Journal without resorting to namespace traversal. For large

volumes, this can reduce the time for scan operations from hours to seconds. For example, a backup application can consult the Change Journal to build its list of files before performing an

incremental backup.

Distributed Link Tracking

Distributed Link Tracking lets client applications track link sources that have been moved. NTFS 5 implements a volumewide indexed unique object identifier OID for each file. This OIDlets the new Distributed Link Tracking feature preserve shortcuts and object linking and embedding OLE

linkssuch as a Microsoft Excel worksheet embedded in a Microsoft Word

documentto NTFS files that have undergone a name and/or path change, including a move to a different volume or computer. For example, a client application can continue to access a

linked database, even if the database location has changed.

Other Supported File Systems

Windows 2000 Server supports multiple file systems. Some provide backward compatibility and others offer access to the latest storage media.

FAT16 and FAT32

To provide backward compatibility, Windows 2000 Server continues to support the FAT16 file system used originally in MSDOS, Windows 3.1, early Windows 95, and OS/2. Today,

almost all computers and operating systems can read FAT16 volumes. FAT16 volumes range from floppy disk size to a maximum of 4 GB. Maximum file size is 2 GB. However, for best

performance, volumes that are 2 GB in size or larger should be formatted using FAT32.

Windows 2000 Server introduces support for the FAT32 file system used until now only by Windows 95 OEM11Service Release 2 OSR2 and Windows 98. FAT32 supports volumes from

512 MB up to 2 terabytes TB in size. Maximum file size is 4 GB. For performance reasons, Windows 2000 will not let you create FAT32 volumes larger than 32 GB to create volumes larger

than 32 GB, use NTFS

. However, an existing FAT32 volume of any size created by Windows 95 OSR2 or Windows 98

can be mounted. FAT32's smaller cluster size gives a 20 to 30 percent

increase in disk space efficiency over FAT16.

To dual boot Windows 2000 and another non Windows NTbased operating system, the system partition must be formatted with either FAT16 or FAT32, whichever is appropriate:

FAT16. Format the partition using FAT16 if the installation partition is smaller than 2 GB, or if you are dual booting Windows 2000 and MSDOS or Windows 3.1.

FAT32. Format the partition using FAT32 if the installation is 2 GB or larger and you are dual booting with Windows 95 OSR2 or Windows 98.

Compact Disk File System

Windows 2000 Server continues to support the Compact Disk File System CDFS, which lets data be read from CDROM devices. The Microsoft implementation of CDFS meets the ISO12

9660 specification.

Universal Disk Format UDF

Windows 2000 Server introduces support for the Universal Disk Format UDF file system defined by the Optical Storage Technology Association OSTA13. UDF is compliant with ISO

13346; Windows 2000 supports version 1.5. UDF is the successor to CDFS ISO 9660, and is also used for data interchange between operating systems and for digital versatile disk DVD14.

Currently, Windows 2000 Server supports readonly operations for UDF, but will support write operations in the future.


15/39



Top Of Page

Indexing Service

Under Windows NT Server 4.0, Content Indexing Server shipped as part of the Microsoft Internet Information Server IIS, giving customers full text indexing and searching of documents

located on their Web sites. However, Windows 2000 Server ships Microsoft Indexing Service as part of the base operating system. This development extends the indexing and searching

services to locate information on file servers as well as on Web sites.

Windows 2000 Server and Professional Indexing Service indexes file system objects and intranet and Internet Web sites across volumes and machines so that they can be searched by

network, intranet, and Internet users alike. Making these search activities look similar to the user saves an organization time and money in training and supporting employees. All Indexing

Service operations are automatic, including index creation, index updating, and crash recovery in the event of a power failure.

To access Indexing Service in the Computer Management console, click Start, point to Settings, and then click Control Panel. Doubleclick Administrative Tools, and then doubleclick

Computer Management. Figure 6 shows Indexing Service under the Services and Applications node of the Computer Management console.

Figure 6: Indexing Service snapin.

These Indexing Service topics are covered in the following sections:

Indexing Service structure

Catalogs

Both data and property search

Search and retrieval

Indexing control and speed

Detecting changes using NTFS Change Journal

Index storage using sparse streams

Integrating searches into applications

Remote storage and retrieval integration


16/39



Mount/dismount tracking

Indexing Service Structure

The Indexing Service feature itself consists of two types of data: the actual index server and the catalog or index data structures:

The index servercan exist on any Windows 2000 computer and consists of binary files and registry entries similar to many other Windows services. The server's components are

backed up with all other components on the volume.

The catalog filescan be on any volume local to the index server service. You can also back up and restore catalog files using the Windows 2000 backup utility. For more aboutcatalogs, see the next subsection.

Catalogs

Indexing Service builds and maintains catalogs of the contents of local and remote disk drives. A catalog consists of index information and stored properties for a particular group of

directories.

At installation, Indexing Service creates the following two catalogs by default:

System catalog.When Indexing Service is installed with Windows 2000 Server, it automatically builds the System catalog, which, by default, lists all directories on all permanently

attached disk drives. It contains an index for all file system documents except certain system and temporary files.

Web catalog.If Internet Information Services IIS is installed, Indexing Service also creates a Web catalog, which contains an index of the content of IIS, the default virtual server of

the World Wide Web for Web pages stored on a Windows 2000 Server server.

Using the Indexing Service snapin, you can configure existing catalogs and can add and remove additional catalogs at any time. After you add a catalog, you must add the directories to

be included within the catalog's scope. This is the set of directories that the catalog covers, both those to be indexed and those specifically excluded from being indexed. For each included

or excluded directory, all of its subdirectories are also included or excluded.

Both Data and Property Search

Indexing Service is tightly integrated with NTFS 5 and makes use of the new NTFS feature Native Property Sets described earlier in the "NTFS and Related Enhancements" section

. The

index includes all words in the content of each document and all properties. Document properties are items of information about the document, such as file name, date created, date

modified, author name, number of characters, and, for Windows 2000 NTFS documents, all the Microsoft Office summary information properties as well. Many property values are set

automatically by the application that creates the document. In addition, users can create custom properties for any file on an NTFS 5 volume.

Users can search for documents that contain specific words or phrases or for properties. For example, you can search for all documents containing the word "product" or for all Microsoft

Office documents written by Kyle. Indexing Service can index the following types of files:

Text

HTML

Microsoft Office 95 and later

Internet mail and news

Any other file for which a document filter is available


17/39



Search and Retrieval

Users can perform searches using the Windows Explorer search pane, the Start menu's Search function, the Indexing Service MMC snapin query form, or a Web browser using a Web page

created by the Webmaster or administrator

. If the indexed results are stored on a local NTFS volume, access privileges determine whether a document is returned in the result set. In other

words, a user who does not have at least Read access will not know that the document exists because NTFS will not return even a reference to the document.

When results are returned on a Web page, Indexing Service can rank hits according to how well they fit the query criteria, and it can sort results on multiple levels according to the value of

any document property. When results are returned in the regular Windows user interface, the user can sort the documents by rank. The administrator can limit the maximum number of hits

returned to the user.

Indexing Control and Speed

You can control what content is indexed and can set indexing on a perfile and/or perdirectory basis. No background indexing occurs when you are performing a server management task.

Better tracking of user activity and input/output make faster indexing and query results possible. You can also set the priority of the indexing task by stopping the Indexing Service,

selecting All Tasks, and then selecting Tune Performance.

Detecting Changes Using NTFS Change Journal

Indexing Service uses the NTFS volume Change Journal to detect file additions, deletions, and modifications, even when the service is not running. This eliminates costly file rescans and

improves performance. The NTFS Change Journal is described earlier, in the section "NTFS and Related Enhancements."

Index Storage Using Sparse Streams

Indexing Service uses NTFS sparse streams for index storage, which lets index optimization occur using the existing allocated disk space. NTFS sparse files support is described earlier, in

the section "NTFS and Related Enhancements."

Integrating Searches into Applications

In Windows 2000 Server, programmers and script writers can take advantage of the native indexing and search features using Microsoft ActiveX Data Objects ADO and OLEDB

interfaces. Both interfaces provide a consistent way to access data regardless of how the data is structured

. This lets ISVs integrate full text and property searching into their applications.

Remote Storage and Retrieval Integration

Indexing Service includes Remote Storage content in its searchable index. Remote Storage, which is the Windows 2000 version of Hierarchical Storage Management HSM, stores content

remotely, retrieving it only when needed. Indexing Service lets users search for files in both the local volume and in remote storage. Archived data can be indexed and searched without

users knowing that it is no longer stored on the local volume, thus freeing up storage without having to buy additional hard disks. See the subsection "Remote Storage" in Storage

Management Services.

Mount/Dismount TrackingThis feature lets indexing interact invisibly with the CHKDSK and format utilities. More important, it lets you index and store catalogs on removable disks, such as Jaz and Zip cartridges.

Top Of Page

Storage Management Services

The quantity of data stored on distributed systems has increased exponentially over the last decade. As the number of client/server systems increases in an organization, so does the

number of storage subsystems. Up to 25 percent of a typical computing budget is spent on storage. The kind of data stored on client/server systems is changing as wellgrowth in

Internet/intranet usage and in 32bit/64bit architectures are major contributors to the changes in types of data found in the distributed network. These deve lopments are accelerating the

creation of large volumes of data and resulting in increased storage requirements at proportionately increased cost.

Windows 2000 Server offers several new or improved features that improve storage and reduce its cost. These include:

New disk architecture


18/39



Defragmentation utility

Remote Storage service

Removable Storage service

Disk quotas

The Single Instance Store SIS

System File Protection

Storage support for hardware innovations

Each of these features is described in the following subsections.

In addition, Windows 2000 Server includes many enhancements that provide ISVs the infrastructure they need to write enterpriseclass storage applications and features. Also new in

Windows 2000 Server is the Microsoft Management Console MMC, which lets administrators and ISVs create the administrative tools called snapins that manage Windows 2000

components and services. ISVs should plan to write an MMC snapin as part of the administrative user interface for their storage solution.

New Disk Architecture

Windows 2000 Server supports two types of disk storage:

Basic disks,divided into partitions, are the type of disk used by Windows before the introduction of the Windows 2000 operating system. The Fault Tolerant Disk manager FT Disk,

present in Windows NT 4.0, manages Windows 2000 basic disks and legacy FT Disk volume sets from Windows NT 4.0. A basic disk is characterized by the placement of a few

kilobytes of data a signature

at the beginning of the disk. Basic disks can by partitioned with and recognized by Microsoft MSDOS, Windows 95, Windows 98, and Windows NT.

Dynamic disks,divided into volumes, are unique to Windows 2000. The new Logical Disk Manager LDM manages dynamic disks and all new volume sets. Dynamic disks do still

have partitions at the disk level, but these partitions are not shown in the user interface. LDM places a 4 MB soft volume database for storing metadata at the end of any physical

disk containing an LDM volume in addition to the few kilobytes of data at the beginning of the disk that are characteristic of basic disks. The 4 MB database is replicated among all

dynamic disks in the system. Dynamic disks can contain simple volumes, spanned volumes, stripe volumes, mirrored volumes, and redundant array of independent disks level 5

RAID5 volumes.

Figure 7 shows the new Windows 2000 disk architecture.


19/39



Figure 7: Basic and dynamic disks in Windows 2000 Server

New disks are installed as dynamic disks. By default, existing system disks are installed as basic disks, which you can convert to dynamic disks. Both basic and dynamic disks can be

formatted with FAT, FAT32, or NTFS. If you don't do so at installation, you can upgrade from basic storage to dynamic storage at any time. When you do so, existing partitions are

converted into volumes, as shown in the following table:

A basic disk partition, logical drive, or set becomes a dynamic disk volume

A partition including system, boot, primary, and extended

partitions becomes

a simple volume can be extended if NTFS 5; has a single dynamic disk partition, but cannot contain NT4 and

earlier partitions or logical drives

A logical drive becomes a simple volume

A volume set becomes a spanned volume

A mirror set becomes a mirrored volume

A stripe set becomes a striped volume

A stripe set with parity becomes a RAID5 volume

Disk and Volume Management Features

The Disk Management MMC snapin which replaces the Windows NT Server 4.0 Disk Administrator utility displays disks and volumes in either a graphical view or a list view. It offers the

following features:

New user interface.The Disk Management feature's shortcut menus show which tasks you can perform on the selected object, and wizards guide you through initializing or

upgrading disks and creating partitions or volumes.

Fewer reboots.With dynamic volumes formatted with NTFS 5, you can often add disks and create, extend, or mirror a volume without rebooting, letting users work withoutinterruption and saving administrators' time.


20/39



Selfdescribing disks.The operating system keeps disk configuration metadata on each disk and replicates it. This is a change from Windows NT Server 4.0 and earlier, which stored

configuration metadata in the registry. Selfidentification of managed disks ensures that disk controller and other disk reconfigurations or cluster disk ownership transfers are error

free.

Volume mount points.You can use Disk Management to connect to, or mount, a local drive at any empty folder on a local NTFSformatted volume see the subsection "NTFS

Volume Mount Points"

.

Volume GUIDs support Plug and Play.Unique disk signatures identify both basic partitions and dynamic volumes. Disk signatures associate partitions or volumes with their

assigned logical names. The disk signature lets FT Disk or the LDM accurately identify a partition's or volume's presence or absence in environments that implement Plug and Play

storage devices.

Disk and Volume Management Tasks

The following table lists the specific operations you can perform with the Disk Management snapin:

DISK & VOLUME MANAGEMENT TASKS

Same for basic and dynamic disks Same for basic disks and dynamic volumes

Manage disks on a remote computer.

Add a new disk.Move disks to another computer.

View disk properties.

Update disk information.

Reactivate a missing or offline disk.

Upgrade a basic disk to a dynamic disk.

Change a dynamic disk back to a basic disk. Note: This destroys all data

on the disk.

Restore disk configuration information from previous versions of

Windows NT.

View properties.

Assign, change, or remove a drive letter you cannot change drive letter of system or boot volumeCreate a mounted drive partition or volume is assigned a drive path, not a drive letter

Format a partition or volume.

Delete a partition or volume you cannot delete system volume, boot volume, or volume that contains the

active paging file

.

SIMILAR FOR BOTH DYNAMIC VOLUMES & BASIC DISKS

Dynamic Volumes Basic Disks

Manage spanned volumes correspond to volume sets in Windows NT 4.0

.

Manage mirrored volumes correspond to mirror sets in Windows NT 4.0; provide fault

tolerance.

Manage striped volumes correspond to stripe sets in Windows NT 4.0

.

Manage RAID 5 volumes correspond to stripe sets with parity in Windows NT 4.0;

provide fault tolerance.

The management tasks listed in the left column are the same on basic disks, except that:

You cannot create newspanned, mirrored, or striped sets on basic disks, and you cannot

extend a spanned volume.

UNIQUE TO EITHER DYNAMIC VOLUMES OR BASIC DISKS

Dynamic Volumes Only Basic Disks Only


21/39



Create, extend, and delete simple volumes. Simple volumes are the dynamic storage equivalent of Windows NT 4.0 and earlier

primary

partitions. When you have only one dynamic disk, the only kind of volume you can create is a simple volume.

Create and manage primary partitions,

extended partitions, and logical drives.

Mark a partition as active on an Intel

based computer.

Defragmentation Utility

The disk defragmentation utility is another Windows 2000 Server enhancement that improves storage management. The utility tells the file system to move the data from one sector to

another. The fact that the file systemrather than the defragmentation utilitymoves the data makes the system far more robust.

Administrators or users can defragment volumes formatted for FAT, FAT32, or NTFS, which improves performance if the fragmentation is extensive. This utility can operate while the systemis up and running and disks are in active use. You can defragment only local volumes and only one volume at a time.

Remote Storage

The Remote Storage service, which is the Windows 2000 version of Hierarchical Storage Management HSM

, helps manage the cost associated with large quantities of data that must be

kept accessible. The Remote Storage hierarchy consists of two layers:

Local storagerefers to the NTFS volumes local to the Windows 2000 file server15hosting the Remote Storage software.

Remote storagerefers to data moved from the local hard disk to a remote storage device such as tape that can be recalled whenever needed.

If a file has not been used in the past thirty days, there is a high probability that it will not be accessed again. These infrequently used files consume the majority of disk space, and it isthese files that Remote Storage typically migrates to secondary storage. Remote Storage automatically moves data back and forth between highcost, faster disk drives and lowcost, high

capacity storage media tape library. Remote Storage monitors the amount of space available on local NTFS volumes, and when the amount of free space dips below the needed level,

eligible files are transferred from the hard disk to secondary storage. Yet, the user still sees and can still access these archived files. This frees up storage on the file server without requiring

the purchase and installation of additional hard disks.

Remote Storage storage media are not a substitute for primary backup media. Remote Storage is typically used to migrate infrequently used data, so frequently used data, which is more

likely to be urgently needed, is less likely to be stored on Remote Storage media. The purpose of Remote Storage is to ensure free space on file server volumes, not to protect enterprise

data.

Applications running on NTFS volumes that regularly open many files can cause a great deal of data to be recalled, and thus reduce the efficacy of Remote Storage. For best results, use

Remote Storageaware applications.

Managing Remote Storage

The Remote Storage service is not installed by default when you run Windows 2000 Server setup, but you can choose to install it during setup or afterwards. Before setting up Remote

Storage, you must ascertain that remote storage media are available for it to use. The local disk volumes on the file server that are under Remote Storage control, called managed volumes,

must be nonremovable. Administrators who intend to use data compression on managed volumes should compress the volumes before installing Remote Storage. Administrators who

intend to use content indexing must also first set up Microsoft Indexing Service.

The Remote Storage snapin is comprised of four components, all of which run on both Windows 2000 Server and Windows 2000 Professional the Remote Storage service runs only on

Windows 2000 Server computers, not on Windows 2000 Professional machines, but the Remote Storage user interfaceruns on both:

Recall Notification user interface.This interface lets an administrator cancel a recall of remote data, if the recall is invoked before the data transfer has started.

Windows Explorer component.Remote Storage adds a new page to the file and directory property sheets accessible through Windows Explorer to represent storage management


22/39



properties. These pages provide information on migration status, premigrated file data location in remote storage, premigrated date and time, and so on. Users have readonly

access to the new Remote Storage property pages; however, they can force immediate premigration of individual files or entire directories by setting a Premigrate Nowoption on

the Remote Storage page. For an explanation of premigration, see next subsection.

Disk Management component.This component includes a property sheet showing total used space, free space, premigrated file space, truncated files placeholders, untruncated

file disk usage, premigration space savings, truncated file compression ratio, percent of files that are placeholders, and other volume report information. For an explanation of

premigration, truncation, and placeholders, see the next subsection.

Remote Storage snapin.You can use the Remote Storage snapin one of the Administrative Tools, once it is installed to establish the following two administratordefined

guidelines for a managed volume:

Desired Free Space. This specifies the amount of free space to be kept on the managed volume and triggers an automatic file truncation when the free space is too low.

File Selection Criteria. This specifies which files can be migrated to secondary storage. These settings include minimum file size and time elapsed since the file was last accessed.

You can also specify file inclusion and exclusion rules. For example, you might specify that only files not accessed for at least three days should be migrated, and that executable files

should be excluded from migration. Remote Storage cannot copy hidden, system, extended attribute, encrypted, or sparse files to remote storage. The selection criteria decision

engine is extensible using a Component Object Model COM interface.

Premigration, Truncation, and Placeholders

Remote Storage scans a managed volume periodically for eligible data,premigrates copies eligible files to secondary storage, and attaches an NTFS reparse point to each copied file, but

it does nototherwise modify the file at this point. More frequently than the scan that results in the copying of eligible files, Remote Storage also checks whether free space on the volume isat or above the specified threshold. If it drops below the threshold, Remote Storage then creates free space by truncating files that have already been premigrated to secondary storage.

Before each file is truncated, Remote Storage uses the NTFS Change Journal to determine whether the data in secondary storage still represents the data in primary storage primary

storage refers to the NTFS volume local to the server running Remote Storage or mapped locally to that server by Dfs

. If the primary data haschanged, the file is not truncatedit is no

longer considered premigrated and is returned to normal status.

After a file is truncated, what remains on the primary storage deviceso that migrated files can be viewed in directories and recalled as neededis an NTFS file, called aplaceholder. The

placeholder points to the complete copy now in remote storage. A placeholder has the systemdefined $REPARSE POINT attribute set with information that can identify and retrieve the

data from remote storage. The placeholder is marked FILE_ATTRIBUTE_OFFLINE. Although Remote Storage has changed the physical size of the file on local storage, the file's logical size

and the date/time create, last modified, last accessed remain unchanged.

When a user or application reads, writes, or makes a memory map request for a truncated file, Remote Storage recalls the removed data from secondary storage and reconstructs the

primary data stream. This operation is transparent to users and to applications for example, it does not affect the Windows 2000 disk quotas feature, which monitors and limits disk space,except that I/O operations are blocked for several minutes until the requested data has been restored from tape.

Remote Storage initiates Automatic File Truncation whenever a managed volume's free space level goes below the specified Desired Free Space setting. You can also force the truncation of

premigrated files to placeholders using Schedule File Truncation. In this case, premigrated files that have not been modified, and that meet the File Selection Criteria, are truncated

regardless of the volume's free space level. You might do this in advance of a volumeintensive event, such as the installation of a large application.

In addition, you can use Validate Managed Files to ensure that all files on your managed volumes point to valid data in remote storage. Validation also detects files that have been moved

from one local volume to another, or that have been modified. Validation is automatically performed two hours after a backup program is used to restore a remote storage file. You should

also perform validation on a regular basis to correct inconsistencies, especially after restoring files on a local volume or after disk errors have occurred on a volume.

You can rename premigrated files and placeholders only on the same volume. Renaming does not cause the data to be recalled. If you copy or move placeholders between volumes, the

data is recalled and the entire file including the migrated data is copied. At the completion of a move operation, the original placeholder file is deleted. You can also move a placeholder

to another volume on the same system by using Windows 2000 Backup to backup the placeholder and then restore it to another volume. In this case, the moved placeholder correctly

points to remote storage and can initiate a recall.


23/39



Remote Storage Interoperates with Related Tools

Remote Storage has been designed to work with the following software:

Job Scheduler.Remote Storage uses the Windows 2000 Job Scheduler to schedule Remote Storage jobs. Job status can be monitored from within the Remote Storage snapin.

Using the Job Scheduler, you can specify a job window that limits the amount of time spent in a single scan, which is useful if the system has a large amount of data to manage and

the scan might take too much time. A bookmark is kept where the scan stopped, so the scan can continue from that point the next time it is run.

Removable Storage.Remote Storage uses Removable Storage to access the library tapes used for remote storage see next section for more about Removable Storage. All tapesused by Remote Storage exist in a single application media pool that is automatically created during Remote Storage Setup. You use Removable Storage to verify that sufficient

media have been moved to a free media pool, so that Remote Storage can use tapes from that pool if needed. You cannot move tapes from the Remote Storage application media

pool to another application media pool. Remote Storage can support only a single tape type for use as remote storage. You specify which type is supported when you run the

Remote Storage Setup wizard. You cannot change this type later.

Backup.Remote Storage interoperates with the updated Windows 2000 Backup utility to correctly handle data recovery. To prevent legacy backup applications from backing up

Remote Storage files, set the registry value HKEY_LOCAL_MACHINE \SYSTEM \CurrentControlSet \Services \RSFilter \Parameters \SkipFilesForLegacyBackupto 1.

Thirdparty applications.In addition to the cost savings in storage media that Remote Storage provides within a company, thirdparty developers also benefit from Remote

Storage integration with the operating system. Instead of developing proprietary solutions, developers of applications such as backup programs, content indexing, and antivirus

agents can now easily create solutions by calling common APIs. However, because file recall involves latency the delay experienced during data retrieval, when I/O operations are

blocked until the requested data has been restored, developers of storage applications must be aware of possible interactions with Remote Storage.

Security

Remote Storage supports NTFS security features and thus recalls a file only if the user has valid access to the placeholder pointing to the data in remote storage.

Removable Storage

The new Windows 2000 Removable Storage service manages removable storage media tapes and optical disks and robotic storage libraries attached to a computer running Windows

2000 Server or Professional. Removable Storage moves media around within and between libraries and controls access to that media. Removable Storage consists of a user interface

implemented as an MMC snapin, a Windows 2000 service with API, and a database.

Removable Storage lets administrators perform the following tasks:

Create media pools groups of media and set media pool properties.

Insert and eject media in a robotic library.

Mount and dismount media.

View the operations state of media and libraries.

Perform library inventories.

Set security parameters for users.

12/8/2014 Fil d P i S i T h i l O i


24/39



Figure 8 illustrates the components of Removable Storage, managed devices, and Windows 2000 Server. Windows 2000 Removable Storage components are shown in white.

Figure 8: Removable Storage components

To ascertain which of the many drives and robotic libraries available are supported for Removable Storage, see the "Microsoft Windows 2000 Hardware Compatibility List HCL" entry in

the "For More Information" section at the end of this document. Also consult the HCL for the proper configuration settings for all supported drives and robotic libraries.

You can create command scripts using the Removable Storage commandline program Rsm.exe. Use this program to perform routine or automated activities, such as ejecting tapes,

creating media pools, and so on.

Removable Storage and Client Applications

Removable Storage helps reduce costs by letting multiple client applications share local libraries and tape or disk drives while ensuring that client applications do not corrupt each other's

data. Removable Storage gives applications access to storage devices through an API that lets applications catalog all removable media except floppy disks, whether housed online in

robotic libraries or offline on shelves. This API and the services it provides hide the details of the various drives and libraries. Because an application does not need to know what specific

hardware is being used, media support provided by Removable Storage lets ISVs that create storage applications concentrate on customer features rather than on hardware issues.

Managing media contentis handled by Removable Storage client applications called datamover applications, such as backup applications. For example, Removable Storage mounts tapeswhen needed by a backup application, but the backup application itself tracks the backup sets stored on that media.

For information about writing Removable Storageaware applications, see the section "For More Information."

Removable Storage Database

Removable Storage uses its database to track mediarelated system components. It stores the properties of managed objects such as libraries, drives, and media and updates the

database whenever an administrator or an application makes a change in a computer running Removable Storage. This database is used internally by Removable Storage and is not directly

accessible to administrators or client applications. Administrators should back up the database on a regular basis.

Using Removable Storage

The Removable Storage snapin, shown in figure 9, is located beneath the storage node in the Computer Management snapin. You can also start it directly by running Ntmsmgr.msc. You

use it to add Removable Storage objects, view and modify properties of Removable Storage objects, insert and eject media, perform inventories, mount and dismount media, and check

12/8/2014 Fil d P i t S i T h i l O i


25/39



status information.

Figure 9: Removable Storage is located under Computer Management

The four Removable Storage nodes shown in the screenshot in figure 9 are described in the following four subsections:

Physical Locations: Libraries and Offline Media

Media Pools: Logical Media Collections

Work Queue

Operator Requests

Physical Locations: Libraries and Offine Media

Removable Storage manages two classes of physical locations:

Libraries.Removable Storage libraries include both media and the means to read and write to them. A CDROM drive with a disk inserted is a standalone library with one drive, noslots, and no transport a transport is a robotic device that moves a medium from its slot to a drive and back again. A standalone libraryis one in which the medium must be

manually placed in the drive. A more complex example is a roboticbased tape library, which holds several up to several thousand tapes, has one or more tape drives, and has a

mechanical means to move tapes into and out of the drives a transport. A robotic library, also called a changeror ajukebox, has either a door or an insert/eject IE port. The

physical location of media in an online library is the library in which it resides.

Offline media.Removable Storage offline media are media that are not in an online library but are physically located elsewhere, such as on a shelf. When a tape or disk is taken out

of an online library, Removable Storage records that it now resides in the offline media physical location.

Media Pools: Logical Media Collections

Removable Storage organizes all the media in a library into logical groups called media pools. A media pool is a collection of removable tapes or disks. The tapes or disks in a media poolhave the same management properties that is, administrators define the properties that apply to a set of media. A given media pool can hold only one type of tape or one type of disk.



26/39



Data management applications use media pools to gain access to specific tapes or disks w ithin libraries managed by Removable Storage.

Media pools control the selection of media and media type, let media be shared across applications Removable Storage moves media between media pools to provide the amount of data

storage an application requires

, and track such sharing. Four types of media pool existthree types of system