Upload
trannguyet
View
222
Download
1
Embed Size (px)
Citation preview
1
Session 1726 - NetBackup 7.6 Best Practices: Improving Recovery Times
George Winter, Technical Product Manager
Reneé Carlisle, Sr. Product Manager
Session 1726 - Improving Recovery Times
SYMANTEC VISION 2014
Sample Agenda
Session 1726 - Improving Recovery Times 2
Recovery Challenges 1
NetBackup Tools That Help Overcome Challenges 2
Deep Dive on VMware Recoveries 3
Final Thoughts 4
Questions 5
SYMANTEC VISION 2014
Recovery Challenges
Session 1726 - Improving Recovery Times 3
SYMANTEC VISION 2014
Recovery Challenges
4
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
• Recovery needs to be quicker
• Backups need to happen more frequently
Tight Backup Window
• Backup can’t complete in the allotted time
• Data has outgrown the backup window
• Recovery can’t happen without a backup
Optimize Type of Recovery
• Single File Restores
• Full System Restores
• Full site restores
The larger the recovery scope, the
greater the cost
Disaster Recovery requires data to be available
• Trucking Tapes Has Risk and High Operational Cost
• Array Based Replication Increases Storage Cost and is the most Expensive
• Data has to be available where you need it
Session 1726 - Improving Recovery Times
SYMANTEC VISION 2014
CPU
• How much compute
power do you have?
• Is the load too high
on the client?
CPU
• Offhost Backup
• Accelerator
• Replication Director
• Use Appliances
I/O
• How fast can you read from
your source?
• How Fast can you write to
your target?
I/O
• Client Direct
• Accelerator
• Snapshots/RD
• OptDupe/AIR
• Use Appliances
Data Growth
• How much data do you have?
• How fast is it growing?
• Is it all equally important?
• Do you need instant access to
it?
Data Growth
• Data Archiving
• Deduplication
• Accelerator
• Replication Director
• Use Parallel streams
Your Recovery Time is dependent on:
5 Session 1726 - Improving Recovery Times
Network Bandwidth
• How much bandwidth do you have?
• Do you need to send it all offsite?
• Where do you need to recover?
Network Bandwidth
• OptDupe
• AIR
• Accelerator
SYMANTEC VISION 2014
Leverage NetBackup to reduce Recovery Challenges
Session 1726 - Improving Recovery Times 6
SYMANTEC VISION 2014
Meet RTO & RPO objectives with Appliances
• 30% Faster Peak
Client Dedupe
Performance
• 155% Faster Peak
Target Deduplication
• 50% Faster Restore
Speed
Leveraging the Appliance to meet your RTO Reduce RTO, Improve DR, Decrease Backup Window
7
5230 – 2.5.2 5220 – 2.5
Backup - Peak Throughput Client Deduplication 100 Streams
30.85 TB/hr 23.66 TB/hr
Backup – Peak Throughput Target Deduplication 100 Streams
8.33 TB/hr 3.85 TB/hr
Restore – 8 streams, 80 GB of data 331 MB/s 213 MB/s
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Tight Backup Window Optimize Type of Recovery Disaster Recovery requires
data to be available
Session 1726 - Improving Recovery Times
SYMANTEC VISION 2014
Meet RTO & RPO objectives with Appliances
• 29% Faster Peak
Client Dedupe
Performance
• 46% Faster Peak
Target Deduplication
• 39% Faster Restore
Speed (4 streams)
• 38% Faster Restore
of 100th Backup
Upgrade to take advantage of performance improvements
8
Description 5230 – 2.5.x
5230 – 2.6
Backup - Peak Throughput Client Deduplication @98% 144 Streams
30.8 TB/hr 39.9 TB/hr
Backup – Peak Throughput Target All-In-One Deduplication@98% 366 Streams
8 TB/hr 11.7 TB/hr
Restore – 4 streams 331 MB/s 538.98 MB/s
Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Tight Backup Window Optimize Type of Recovery Disaster Recovery requires
data to be available
SYMANTEC VISION 2014
Regarding my monthly weekend full backup of a Linux client that mounts 13 TB SAN storage that backs up to MSDP and then duplicates to tape. Before the upgrade the duplication to tape process would not complete till mid week. After the upgrade it completed over the weekend! Very satisfied with v7.6 duplication improvement.
Session 1726 - Improving Recovery Times 9
SYMANTEC VISION 2014
Meet RTO objectives with Advanced Disk
• Allows for fast staging
area for increased RTO
• Use commodity disk that
can span physical
boundaries
• Allows classification of
data for increased
protection of more critical
data
Advanced Disk Reduce RTO, Decrease Backup Window
10
Adv Disk Pool Gold Adv Disk Pool Silver
Disk Volumes
NFS Mounts
NetBackup Media Servers function as both storage servers and data mover
Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Tight Backup Window Optimize Type of Recovery
SYMANTEC VISION 2014
• Automatically move
data through its
lifecycle
• Ensure data is stored
in the right place for
the right time
• Ensure that you
always have a copy of
data available that
meets your retention
requirements
Storage Lifecycle Policies Reduce RPO/RTO, Decrease Backup Window, Improve DR
11
backup job 2
backup job 3
backup job 1
backup job 4
bronze
lifecycle
policy:
backup to
tape, retain 6
months
silver
lifecycle
policy:
backup to
appliance,
retain 2
months
duplicate tape,
vault offsite &
retain 6
months
gold
lifecycle
policy:
backup to disk,
retain 3 weeks
Write to
appliance
retain onsite 2
months
write to tape,
vault offsite,
retain 6
months
Meet RPO & RTO objectives with Storage Lifecycle Policies
Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Tight Backup Window Optimize Type of Recovery Disaster Recovery requires
data to be available
SYMANTEC VISION 2014
Meet RTO & RPO objectives with Auto Image Replication
• Data is off site as soon
as your backup policy
completes
• Data and applications
are backed up, deduped,
and replicated
immediately
• Data and applications
are available for restore
- NOW
• No tapes to search for
and load
Auto Image Replication (A.I.R.) Reduce RPO/RTO, Decrease Backup Window, Improve DR
12
Production Data Center #1
Domain A
Remote office Domain C
Domain D Production Data Center #2
Domain B Branch office
Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Tight Backup Window Optimize Type of Recovery Disaster Recovery requires
data to be available
SYMANTEC VISION 2014
AIR jobs from MSDP would normally take all weekend, with jobs still running on Monday morning. After the 7.6 upgrade, the first weekend AIR jobs were all complete by Sunday noon. This is due to the MSDP rehydration performance improvements in 7.6
Session 1726 - Improving Recovery Times 13
SYMANTEC VISION 2014
Meet RTO objectives with NetBackup SAN Client
fc fc fc fc fc
fc fc
Adv Disk Pool
load-balanced
media servers os os
fc
OpenStorage
Devices
NetBackup
Appliance
fc fc fc
SAN Clients
fc fc
• Fast SAN Backup–
150 MB/sec backups,
up to 500 MB/sec
aggregate through a
media server
• Remove Backup
Impact from the LAN –
dedicated SAN backup
• High Availability – can
configure redundant
Fibre Channel paths
SAN Client Reduce RTO, Decrease Backup Window
14
fc
Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Tight Backup Window Optimize Type of Recovery
SYMANTEC VISION 2014
Meet RTO & RPO objectives with Accelerator
• Decreased backup
time allows for more
frequent full backups
• low I/O and CPU cost
on client, network
bandwidth and
storage cost
decreases CapEx
• Reduce RTO by
recovering from a full
rather than a series of
incremental backups
Accelerator: Files and Folders and VMware Optimize Recovery, Reduce RPO/RTO, Decrease Backup Window
15
Media Server
Client
Backup
engine 3
4
Application
File System
Synth
esis En
gine
Track Log
Master Server
NBU
Catalog
5
NTFS Change Journal
De
du
pe
Engin
e
1 2
Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Tight Backup Window Optimize Type of Recovery
SYMANTEC VISION 2014
Accelerator had actually meant it’s signature
quote of speed up backup with 100X, indeed it
is here in our case (182%). Seeing at
improvements NetBackup had made, would
strongly recommend to all existing customers
to upgrade their environment on this release.
Session 1726 - Improving Recovery Times 16
SYMANTEC VISION 2014
Meet RTO objectives with FlashBackup
• Combines the speed of
raw-partition backups with
the ability to restore
individual files
• Supports multiple data
streams
• Best for file systems that
contain a large number of
files where most of the file
system blocks are
allocated and have a high
change rate
FlashBackup Optimize Recovery, Reduce RPO/RTO, Decrease Backup Window
17 Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Tight Backup Window Optimize Type of Recovery
SYMANTEC VISION 2014
• Protect entire volumes
of data with Hardware
Snapshots, while still
maintaining GRT
• Application
consistency for
supported workloads
• Manage replication
and long-term copies
from a single policy so
data is where you
want it, when you
need it
Replication Director Optimize Recovery, Reduce RPO/RTO, Decrease Backup Window, Improve DR
18
Meet RPO & RTO objectives with Replication Director
NAS File Services VMware on NFS
MS Exchange, SQL Server (on VMware)
Oracle on NFS
DB2 SAP
Catalog
Hyper V
NetBackup Admin
Console
SnapMirror/SnapVault
NDMP Tape Backup
Tape
Streaming Tape Backup
Snapshot copies on primary offer low impact
Storage efficient block-level incremental Snapshot replication
Snapshot copies and replication fully integrated into backup data life-cycle
NDMP Disk Backup
Disk
Leverage Accelerator with final disk backup using Windows Policy
Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Tight Backup Window Optimize Type of Recovery Disaster Recovery requires
data to be available
SYMANTEC VISION 2014
• Recover entire
database or inidividual
components from a
single backup
• Eliminate 2-step
recovery process
required by database
dumps
• Leverage transport
and storage options
that match your RPO
and RTO SLAs
Application Protection with NetBackup Agents Optimize Recovery, Reduce RTO, Decrease Backup Window
19
Meet RTO objectives with Application Agents
Storage Target
Off Host Instant Recovery SAN
Application Protection Policies Backup Transport
Network
Local Snapshot
Application Integration API
Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Tight Backup Window Optimize Type of Recovery
SYMANTEC VISION 2014
• Meet recovery SLAs
by enabling
application instant
recovery
• Leverage array-based
snapshots or Veritas
Storage Foundations
• Increase reliability by
removing manual
process
Instant Recovery with NetBackup Snapshot Client Reduce RPO/RTO, Decrease Backup Window
20
Meet RPO objectives with Instant Recovery using Snapshot Client
Enterprise Client & Data base Agents
Primary Snapshots
Media Servers
Disk
Tape
Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Tight Backup Window Optimize Type of Recovery
SYMANTEC VISION 2014
Meet RTO & RPO objectives with Accelerator
• Consistent and functional system recovery
• Integrated, easy to manage and administer
• Scalable
• High degree of recovery automation
• Supports Dedupe and Accelerator
• Provides recovery flexibility
Bare Metal Restore Optimize Recovery, Reduce RTO
21
Step
3.
2.
1.
Reboot
Click “Prepare
to Restore”
Repair
hardware
Serv
er R
ecovery
Tim
e
Step
9.
8.
7.
6.
5.
4.
3.
2.
1.
Traditional
Recovery
Reboot
Reboot
Reboot
Reboot
Load
tape(s) and
restore
Reload OS
Reload
backup
software
Collect all
media
Repair
hardware
Bare Metal Restore
Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Optimize Type of Recovery Disaster Recovery requires data to be available
SYMANTEC VISION 2014
Meet RTO & RPO objectives with VMware P2V
• Eliminate need for
stand by hardware –
decrease CapEx
• Have instant access to
servers without
recovery
Automated Physical to Virtual Conversion Reduce RPO/RTO
22
Media server
VMware ESX or
vCenter server
Master server Storage containing backup
Virtual Instance Convertor
(NB-Proxy Host)
Data store that contains
converted virtual instances
Virtual Environment Setup
Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Optimize Type of Recovery Disaster Recovery requires data to be available
SYMANTEC VISION 2014
Meet RTO & RPO objectives with Accelerator
• Instantly power on
any protected VM
from disk backup
target
• No need to
restore VM first
• Uses standard
NetBackup backup
images
• Once powered on,
VM is 100% available
VMware Instant Recovery Optimize Recovery, Reduce RPO/RTO
23
ESX/ESXi
NAS
SAN
NetBackup NFS Datastore
Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Optimize Type of Recovery Disaster Recovery requires data to be available
SYMANTEC VISION 2014
Meet RTO & RPO & DR objectives with combined technologies
• AIR + P2V
• RD + Accelerator
• VADP + Accelerator +
Appliances + AIR +
VIR
Leverage the power of combining solutions Optimize Recovery, Reduce RPO/RTO, Decrease Backup Window
24
Production Data Center
NBU Master Domain A
DR Domain
NBU BMR Master Domain B
NBU Media Server
NBU Media Server
OST Appliance or PureDisk
Device Notifies NBU
OST Appliance or PureDisk
Image
Image
OST Optimized Duplication
NBU Clients Physical/Virtual
Backup
Import image Client System Configuration
backup
Import client system info
NB client which drives conversion
Create Clients Virtual Instances
Virtual Environment Hypervisor Server: (VMWARE ESX or
HyperV)
Session 1726 - Improving Recovery Times
Data Loss Downtime
Slow Recovery Time (RTO) Tighter Recovery Points (RPO)
Tight Backup Window Optimize Type of Recovery Disaster Recovery requires
data to be available
SYMANTEC VISION 2014
The nice thing about NetBackup 7.5 is that with
deduplication, optimized image replication,
and NetBackup Accelerator, we have some file
servers that have 600 gigabytes of SharePoint
data on them and a full backup is done in 40
minutes that would take two days before.
Session 1726 - Improving Recovery Times 25
SYMANTEC VISION 2014
Deep Dive on VMware Restores
Session 1726 - Improving Recovery Times 26
SYMANTEC VISION 2014
VMware Restore Considerations
Session 1726 - Improving Recovery Times 27
SYMANTEC VISION 2014
General Restore Performance Thoughts
• I/O – Spinning Disk
– Reading data from disk is easy
– Writing data to disk is hard
– Full disks (> 80%) slow this process
• I/O – Tape (not VTL)
– Reading data from tape is hard (multiplexed)
– Writing data to tape is easy
Session 1726 - Improving Recovery Times 28
SYMANTEC VISION 2014
VMware Restores With vStorage API for Data Protection
• Backups based on VADP
– VMDK itself is *not* backed up – common misconception
– Data *inside* VMDK is backed up
– This provides ability to reformat VMDK provisioning at restore
– Backup more efficient – skips unused space (NBU adds efficiency too)
• NetBackup with VADP enables additional restore capabilities
– Single file (e.g. Word doc) restores from image (VMDK) backup
– Database object level restore from image (VMDK) backup
– Physical to Virtual
– Virtual to Virtual (somewhat manual)
Session 1726 - Improving Recovery Times 29
SYMANTEC VISION 2014
Why Are VMware Restores Slower Than Backups?
Session 1726 - Improving Recovery Times 30
SYMANTEC VISION 2014
VMware VMDK Restore Transport Modes
• NBD Transport – Restores
– VMkernel port bandwidth limit
– First stream always fastest
– First stream ≈ 100 MB/sec (10 GbE)
– Subsequent (simultaneous) streams slower
– No way around this with NBD transfers
• SAN Transport – Restores
– No VMkernel port QoS limitation
– Can be fastest traditional restore performance
• HotAdd - Restores
– VMkernel port not used
– Can be similar to SAN restore speeds
– Dependent on available ESXi host resources
Session 1726 - Improving Recovery Times 31
SYMANTEC VISION 2014
What Happens During VM Image (VMDK) Restore?
1) VM image restore initiated
2) New VM registered in vCenter
(Ever try to boot the VM at this point?)
3) New VMDK is created
4) All space inside VMDK must be zeroed
(Note that at this point zero data has been restored)
5) NetBackup begins restoring VM data
This explains why a tape may be quickly mounted with delay before data written to tape (or any storage unit)
Session 1726 - Improving Recovery Times 32
SYMANTEC VISION 2014
How VMDK Provisioning Impacts Restore Times
Session 1726 - Improving Recovery Times 33
SYMANTEC VISION 2014
1. An empty 16MB VMDK is first created
2. This 16MB chunk is “zeroed”
– This step mandatory
3. Data restored to this chunk
4. Process repeated until all data restored
– Can’t determine final size of VMDK
• Choice of thin or thick provisioned VMDK impacts restore time
• Restores take twice as long as backups (approx)
• SAN restores impacted by busy vCenter
– All restore instructions channeled through vCenter
The Unvarnished Truth: Thin Provisioned VMDK Restore
Thin VMDK
Restored
Data
Restored
Data
Restored
Data
Restored
Data
Restored
Data
16 MB
Session 1726 - Improving Recovery Times 34
SYMANTEC VISION 2014
EagerZeroedThick Provisioned VMDK Restore
• VMDK is created using 100% required space
• Entire VMDK is “zeroed”
• Data restore process begins
• Thin or Thick Provisioned? Which is faster?
• Small percentage of restore VMDK data
– Thin provision faster
• Large percentage of restore VMDK data
– Thick provisioned probably faster
• Common to have choice dictated by VM admin Thick VMDK
Restored
Data
Session 1726 - Improving Recovery Times 35
SYMANTEC VISION 2014
An Alternative To This Restore Process
Session 1726 - Improving Recovery Times 36
SYMANTEC VISION 2014
Instant Recovery For VMware (IRV)
• Best possible RTO - VM or group of VMs instantly available
• Reverse traditional restore process:
– Traditional restore: 1) Restore VM (hours) 2) Boot VM (minutes)
– IRV restore: 1) Boot VM (minutes) 2) Restore VM (hours)
• Engineering tests indicate 30 second boot times (YMMV)
• Works with any disk based NetBackup target
– Includes MSDP and NetBackup Appliance
• No change to NetBackup disk backup image is required
Session 1726 - Improving Recovery Times 37
SYMANTEC VISION 2014
VMware Instant Recovery in NBU 7.6
38
• Instantly power on any protected VM from disk backup target
– No need to restore VM first
• Uses standard NetBackup backup images
– No need to change any backup process
• Support with all Symantec disk based solutions
– Basic disk, Advanced disk, PDDO, MSDP, NetBackup appliance
• Once powered on, VM is 100% available
– After power-on, VM disks transferred to ESXi storage (Storage VMotion)
– Storage VMotion ensures no disruption of service
ESX/ESXi
NAS
SAN
NetBackup NFS Datastore
Session 1726 - Improving Recovery Times
NetBackup Instant Recovery for VMware Process Overview
Session 1726 - Improving Recovery Times 39
SYMANTEC VISION 2014
• Zero modifications require to either NetBackup or VMware environments
• Works with NetBackup 7.5 images and vSphere 5.0 (or later)
• All ESXi Datastore types are supported (NFS, SAN, iSCSI, DAS)
• Works with NetBackup Appliance, MSDP and Adv Disk
Instant Recovery for VMware Configuration Notes
VMware ESXi
NetBackup Appliance
ESXi Datastores
LAN
Session 1726 - Improving Recovery Times 40
SYMANTEC VISION 2014
Let’s See How Instant Recovery for VMware Works…
Instant Recovery for VMware Configuration Notes
VMware ESXi
NetBackup Appliance
ESXi Datastores
LAN
Session 1726 - Improving Recovery Times 41
SYMANTEC VISION 2014
Instant Recovery for VMware Process
VMware ESXi
NetBackup Appliance
ESXi Datastores
Temp NetBackup
Read Only
Datastore
LAN
NetBackup disk storage is provisioned as read-only NFS Datastore Note that VM1 is currently located on NetBackup disk
1
VM1
Session 1726 - Improving Recovery Times 42
1
SYMANTEC VISION 2014
VM is automatically created and registered in vCenter
Instant Recovery for VMware Process
VMware ESXi
NetBackup Appliance
ESXi Datastores
LAN
2 2
VM1
Temp NetBackup
Read Only
Datastore
Session 1726 - Improving Recovery Times 43
SYMANTEC VISION 2014
Instant Recovery for VMware Process
VMware ESXi
NetBackup Appliance
ESXi Datastores
LAN
REDO 3
VM1 is now powered on REDO location automatically configured All changes to VM1 are captured in REDO
VM1
3
Temp NetBackup
Read Only
Datastore
Session 1726 - Improving Recovery Times 44
SYMANTEC VISION 2014
Instant Recovery for VMware Process
VMware ESXi
NetBackup Appliance
ESXi Datastores
LAN
REDO At this point VM1 is 100% accessible to all users All changes that occur are safely captured in REDO This entire process can take less than 60 seconds
VM1
Temp NetBackup
Read Only
Datastore
Session 1726 - Improving Recovery Times 45
SYMANTEC VISION 2014
Instant Recovery for VMware Configuration Notes
VMware ESXi
NetBackup Appliance
LAN
REDO
Storage VMotion is now initiated VMDK(s) are copied to final destination During this process, VM1 is still 100% accessible
4
VM1
4
Temp NetBackup
Read Only
Datastore
Session 1726 - Improving Recovery Times 46
SYMANTEC VISION 2014
Instant Recovery for VMware Configuration Notes
VMware ESXi
NetBackup Appliance
LAN
REDO
5
REDO is automatically consolidated into VM1 All changes that occurred during this process are automatically retained
VM1
5
Temp NetBackup
Read Only
Datastore
Session 1726 - Improving Recovery Times 47
SYMANTEC VISION 2014
Instant Recovery for VMware Configuration Notes
VMware ESXi
NetBackup Appliance
LAN
Instant Recovery process is now complete Temporary Datastore is removed
VM1
6
Temp NetBackup
Read Only
Datastore
ESXi Datastores
Session 1726 - Improving Recovery Times 48
6
SYMANTEC VISION 2014
IRV Restore Method Comparison – 6 TB VM
• Standard restore method = 25h:01m
• IRV restore method = 3m:22s
446x faster restore with NetBackup IRV
Session 1726 - Improving Recovery Times 49
SYMANTEC VISION 2014
Final Thoughts On Improving Recovery
• Know what kind of recovery you need to do
• Understand the bottle necks in your environment
• Use Appliances
• Keep up with NetBackup upgrades
• Use the right NetBackup feature for your recovery SLA
• Restores are typically slower than backups - plan accordingly
• Virtualization provides additional restore options over physical backups
• Be aware of VMware created limits for certain restore types
• VMDK provisioning and restore transport selection will impact restore performance
Session 1726 - Improving Recovery Times 50
Thank you!
51
YOUR FEEDBACK IS VALUABLE TO US!
Please take a few minutes to fill out the short session survey available on the mobile app—the survey will be available shortly after the session ends. Watch for and complete the more extensive post-event survey that will arrive via email a few days after the conference.
To download the app, go to https://vision2014.quickmobile.com or search for Vision 2014 in the iTunes or Android stores.
Session 1726 - Improving Recovery Times