HPE Scalable Object Storage with Scality RING on HPE Apollo 4200 Gen10 Object-based, software-defined storage at petabyte scale

Technical white paper

Technical white paper

Contents Executive summary ............................................................................................................................................................................................................................................................................ 3 Solution overview ................................................................................................................................................................................................................................................................................. 3

Business problem .......................................................................................................................................................................................................................................................................... 3 Challenges of scale ....................................................................................................................................................................................................................................................................... 3 Scality RING with HPE Apollo 4200 Gen10 servers ..................................................................................................................................................................................... 4

RING architecture ................................................................................................................................................................................................................................................................................ 5 RING components ......................................................................................................................................................................................................................................................................... 6 Scale-out file system ................................................................................................................................................................................................................................................................... 9 Scality’s S3 API .............................................................................................................................................................................................................................................................................. 10 Intelligent data durability and self-healing.......................................................................................................................................................................................................... 11 Multisite geo-distribution .................................................................................................................................................................................................................................................... 12 HPE value add for an object storage environment ..................................................................................................................................................................................... 14 HPE Apollo 4200 Gen10 server reference architecture for Scality RING .............................................................................................................................. 15 Sample bill of materials (BOM) for HPE Apollo 4200 servers and HPE ProLiant DL360 servers ................................................................. 18

Summary ................................................................................................................................................................................................................................................................................................... 18 Resources, contacts, or additional links ........................................................................................................................................................................................................................ 19

Technical white paper Page 3

Executive summary Traditional file and block storage architectures are being challenged by explosive data growth, fueled by the expansion of Big Data and the Internet of Things (IoT). Emerging storage architectures that focus on object storage are helping businesses deal with these trends, providing economical storage solutions that keep up with the demand to expand storage capacity while also providing improved data protection using erasure-coding technology at a lower cost per terabyte (TB).

Enterprise-class storage subsystems are designed to address storage requirements for business-critical transactional data latencies. However, they are not the most practical solution for unstructured data or for backup and archival storage at petabyte and beyond scale. In these cases, enterprise-class reliability is still required, but the need for massive scale-out capacity and lower solution investment per TB, while maintaining or improving the cost of data protection, have become the most important customer requirements.

Object storage software solutions are designed to run on industry-standard server platforms, offering lower infrastructure costs and scalability beyond the capacity points of typical file server storage subsystems. The HPE Apollo 4200 Gen10 servers provide a comprehensive and cost-effective set of storage building blocks for customers that wish to deploy an object storage software solution on industry-standard Linux®-based servers.

Target audience: CTOs and solution architects who are looking for a storage solution that can handle the rapid growth of unstructured data, cloud, and archival storage can reference this white paper. This paper also focuses on controlling licensing and infrastructure costs.

This paper assumes the reader is aware of the challenges that enterprise storage administration poses, and is familiar with data center best practices for storage systems.

Solution overview Business problem Businesses are looking for better and more cost-effective ways to manage their exploding data storage requirements. In recent years, the amount of storage required by many businesses has increased dramatically, especially in the areas of media serving, IoT data collection, and records retention. The cost per TB of storage and ease of data retrieval have become critical factors for choosing a hardware and software solution.

For an increasing number of businesses, traditional file and block storage approaches cannot meet the desired solution attributes. Organizations that have tried to keep up with data growth using traditional file and block storage solutions are finding that both the cost and the complexity of managing as well as operating them has grown significantly. Meanwhile, many organizations that have moved their object storage to a hosted cloud environment have also encountered cost or data control issues.

Challenges of scale There are numerous difficulties associated with storing unstructured data at petabyte and beyond scale to include:

Cost • Unstructured and archival data tends to be written only once and read very infrequently. This stale data takes up valuable space on

expensive block and file storage capacity.

• Tape is an excellent choice for achieving the lowest cost per TB, but suffers from extremely high latencies.

Scalability • Unstructured deployments can accumulate billions of objects. Traditional file systems limit the number and size of files and block storage,

restricting the capacity of presented devices. These limitations can become significant deployment challenges.

• Traditional block and file storage methods suffer from metadata bloat at massive scale, resulting in a large system that cannot meet service-level agreement requirements.

Availability and manageability • Enterprise storage is growing from single site deployments to geographically distributed, scale-out configurations. With this growth, the

difficulty of keeping all the data safe and available is also growing.

• Management silos and user interface limitations have made it increasingly difficult for businesses to deploy the additional storage capacity they need using their existing storage infrastructure.

• Unstructured and archival data may sit dormant for a while, but must be available in seconds rather than minutes when a read request is received by the storage system.

Technical white paper Page 4

Advantages of software-defined storage • Today’s data centers have adopted a new software-defined storage (SDS) model as part of their overall strategy to provide efficiently

scalable infrastructure services. By decoupling software from the underlying platform, enterprises can build solutions with the greatest flexibility spanning the entire portfolio of industry standard HPE ProLiant DL servers, including future hardware offerings. This provides a decisive step forward in reducing the cost of ownership for data center deployments to come.

Scality RING with HPE Apollo 4200 Gen10 servers The Scality RING (RING) running on HPE Apollo 4000 storage servers provides a SDS solution for petabyte-scale data storage that is designed to interoperate in the modern software-defined data center (SDDC). The RING software is fashioned to create a scale-out storage system, which is deployed as a distributed system on a minimum cluster of three storage servers. This system can be seamlessly expanded to thousands of physical storage servers as the need for storage capacity grows (see Figure 1). To match performance to the deployed capacity, the RING can independently scale out the access nodes (connector servers) to meet a customer’s growing I/O throughput requirements.

The RING software requires no specific certification for a customer’s HPE ProLiant server configuration of choice, and supports new generations of hardware as they are released. The RING requires no kernel modifications, eliminating the need to maintain hardware compatibility lists beyond the constraints imposed by the specific Linux distributions running on the server.

HPE Scalable Object Storage with Scality RING utilizes the HPE Apollo 4000 storage servers for high-density storage capacity. HPE Apollo 4000 storage servers enable enterprises to deploy Scality RING with efficiency at the scale their business needs, ranging from an HPE Apollo 4200 Gen10 server with a minimum of 10 hard disk drives (HDDs) to the HPE Apollo 4510 Gen10 server containing up to 60 HDDs.

The HPE Apollo 4200 Gen10 is the latest generation of Hewlett Packard Enterprise’s server to provide maximum storage capacity in a 2U server footprint. The HPE Apollo 4200 Gen10 server continues the storage-density footprint of the original 2U product, and enhances it with leading processor and solid state technology for both memory and storage.

Figure 1. Scality RING SDS high-level architecture

Technical white paper Page 5

The software-defined architecture of RING addresses a number of key customer challenges:

• Massive capacity growth—supplies virtually unlimited scaling of storage capacity and performance to meet today’s and tomorrow’s requirements

• Legacy storage silos with high costs—provides broad support for a large mixture of customer storage workloads, to simplify storage management with fewer silos

• Always-on requirements—designed for 100% uptime, with self-healing, and the highest-levels of data durability

• Cloud-scale economics—compatible across the HPE portfolio, permitting customers to leverage the low total cost of ownership (TCO) of a proven and reliable HPE server platform

• Multiprotocol data access—enables the widest variety of object-, file-, and host-based applications for reading and writing data to the RING

• Flexible data protection mechanisms—efficiently and durably protects a wide range of data types and sizes

• Self-healing—expects and tolerates failures and automatically resolves them

• Platform agnostic—affords optimal platform flexibility allowing for mixed server configurations, and eliminating the need to migrate data when refreshing the underlying hardware

The storage-optimized HPE Apollo 4200 Gen10 server provides the greatest flexibility for deploying the RING architecture.

• Extreme storage capacity in a 2U form factor, with 28 Large Form Factor (LFF) disk bays

• Highly configurable options for both bulk storage HDDs and solid state drives (SSDs)

• HPE Integrated Lights Out (iLO) 5 for remote server management, with the latest innovations in security and performance

• Instantly deployable in new and existing RING configurations

RING architecture To scale-up both storage capacity and performance to massive levels, the Scality RING software is designed as a distributed, fully parallel, scale-out system. It has a set of intelligent services for data access and presentation, data protection, and systems management. To implement these capabilities, the RING provides a set of fully abstracted software services. It includes a top layer of scalable access services (Connector processes installed directly onto the storage servers), providing storage protocols such as SMB, NFS, and S3, for a wide range of application access, as shown in Figure 2.

The middle layers comprise a distributed virtual file system, a set of data protection mechanisms to ensure data durability and integrity, self-healing processes, and a set of systems management as well as monitoring services. At the bottom of the stack, the system is built on a distributed storage layer consisting of virtual storage nodes and underlying I/O daemons that abstract the physical storage servers and disk drive interfaces.

At the heart of the storage layer is a scalable, distributed, key-value object store based on a second-generation, peer-to-peer routing protocol. This routing protocol ensures that store and lookup operations scale-up efficiently to very high numbers of nodes. These comprehensive storage software services are hosted on all RING servers with appropriate processing resources and disk storage. They are connected through standard IP-based network fabrics such as 10/25/40/100 Gigabit Ethernet (GbE).

Technical white paper Page 6

Figure 2. Scality RING architecture

RING components The RING software comprises the following main components: the RING Connector servers, a distributed internal database for metadata called MESA, the RING storage servers and I/O daemons, and the supervisor web-based management portal. The MESA database is used to provide object indexing and manage the metadata used by the Scality Scale-out file system (SOFS) abstraction layer.

Connectors The Connectors provide the top-level access points and protocol services for applications that use the RING for data storage. The RING Connectors support a family of application interfaces, including object-based Connectors (the S3 Connector is based on de facto industry REpresentational State Transfer [REST] standard Amazon Web Services [AWS] S3), as well as file system Connectors (NFS, SMB, and FUSE) to suit a rich set of applications and a wide variety of data types.

Connectors allow application access for objects or files stored into the RING. Applications may make use of multiple Connectors in parallel to scale-out the number of operations per second or the aggregate throughput of the RING for high numbers of concurrent user connections. The system may be configured to simultaneously provide a mix of file access and object access (over NFS and S3, for example) to support multiple application use cases. In addition, files written using S3 may be read back using NFS, and vice versa.

Technical white paper Page 7

Connector processes are most commonly installed directly on the storage servers, as displayed in Figure 3. Some use cases may benefit from choosing to run Connector processes on dedicated hardware. This is an available architecture choice when needed.

Figure 3. RING software processes: RING Connectors, storage nodes, and I/O daemons

The application data I/O path flows from applications through the Connectors. The S3 and sfused connections are also responsible for implementing the configured data protection storage policy (replication or ARC), as described in the Intelligent data durability and self-healing section. For new object writes, the Connector servers may chunk objects that are above a configurable size threshold before the object data is sent to the storage servers. Multiple interface protocols are supported through the Connector processes. See Table 1 for a complete description.

Table 1. External application interfaces supported by Connector servers

Type Connector Strengths

Object Amazon S3-compatible AWS S3 compatible REST Application Program Interface (API), supports AWS Identity

and Access Management (IAM), Active Directory (AD), Bucket/object ACLs, scale-out metadata, and fast listing

REST (sproxyd) API RING’s native interface for geo-distributed deployments provides a flat object storage namespace with direct access to RING objects

CDMI (SNIA Cloud Data Management Interface) REST API namespace compatible with SOFS—NFS, SMB, FUSE—data

File NFS NFS v3 compatible server supports Kerberos, advisory-locking (NLM), and user/group quotas FUSE Scality Sfused Local Linux file system driver, great for application servers and fast

for big files, provides parallel I/O to multiple back-end storage nodes SMB SMB 2.x and a subset of SMB 3.x compliant server

Technical white paper Page 8

Storage nodes Storage nodes are virtual processes that own and store a range of objects associated with its portion of the RING’s “keyspace”. Each physical RING storage server is typically configured with six storage nodes. Under each storage node is a set of storage daemons that are responsible for data persistence across the underlying local disk file system. Each daemon is a low-level process that manages the I/O operations associated with a particular physical disk drive, maintaining the mapping of object indexes to the actual object locations on the disk. The typical configuration is to have one daemon per physical disk drive (see Figure 4), with support for up to hundreds of daemons per server.

Servers hosting the storage nodes should have a small amount of SSD for metadata operations.

Figure 4. RING software deployment

Systems management The Supervisor is the web based GUI for graphical RING management, operations, monitoring, and provisioning. The RING also provides a Command Line Interface (RingSH), and an SNMP MIB and Traps, for use with popular monitoring consoles such as Nagios. The RING supplies a monitoring daemon that is used to efficiently scale statistics collection and monitoring from a large set of storage nodes and storage daemons to the Supervisor.

RING 7.0 expands the user GUI introduced in 6.0, delivering more comprehensive component alerting and a new graphical dashboard. RING 7.0 introduces the Scality Cloud Monitor, which provides 24/7 health and activity monitoring of a RING deployment using easy to visualize graphs for global performance metrics. RING statistics are accessible through a REST API making it possible to integrate with a wide variety of visualization software options.

Technical white paper Page 9

Scale-out file system The RING supports native file system access to RING storage through the file Connector servers and the integrated Scale-out file system (SOFS), a POSIX-compliant virtual file system that provides file storage services without the need for external file gateways as is commonly required by other object storage solutions.

To render file system semantics and views, the RING utilizes an internal distributed database (MESA) on top of the RING’s storage services. MESA is a distributed, NewSQL database that is used to store file system directories and inode structures to provide a virtual file system hierarchy with the guaranteed transactional consistency required in a highly available file system infrastructure. Through MESA, SOFS supports sparse files to offer highly efficient storage of very large files using a space-efficient mechanism.

SOFS file systems can be scaled-out in capacity across as many storage nodes as needed to support application requirements, and can be accessed by a scalable number of NFS, FUSE, or SMB Connectors to meet application load requirements. The RING provides the concept of “volumes”, which may be used to easily configure file system services through the Supervisor. The RING can present up to 232 volumes, and sustain billions of files per volume, with no need to preconfigure volumes for capacity. The RING effectively supports thin-provisioning of volumes. Volumes will utilize the RING’s storage pool to expand as needed when files are created and updated.

A volume gives a view into the file system that may be accessed over one or more Connectors simultaneously with a global namespace. SOFS supports full performance scale-out access within a folder, facilitating multiple file system Connectors of any type (NFS, SMB, FUSE) to concurrently write and read data in a common folder at the same time. To implement safe, high-performance, and steady sharing of folders across multiple Connectors, the RING includes a shared folder cache, which can be accessed by all participating Connectors, to ensure they see the latest view of the folder. This enables consistent (cache coherent) cross-Connector updates and listings even during concurrent update operations.

The RING includes integrated file system load balancing and failover capability. This provides the ability to configure Virtual IP (VIP) addresses, which are accessed externally by applications to mount a file system (or Share from SMB). A load balancer can then route requests into the VIP across multiple physical file system Connectors to spread the load evenly, as well as to route across potentially highly loaded Connectors. In addition, this provides failover capability across multiple Connectors if one becomes inaccessible due to a network or process failure. In conjunction with the full folder scale-out feature described previously, this provides a comprehensive global namespace feature across the RING and its file system folders, with load balancing and failover for all file system Connectors.

The RING supports SMB Multichannel I/O, allowing file servers to use multiple network connections simultaneously to increase throughput and continued operation despite the loss of a network connection.

Beginning with RING 7.0, versioning with Write Once Ready Many (WORM) change control is supported for both File (through SOFS) and Object (through S3). More detail on the Scality S3-compatible protocol can be found in the Scality’s S3 API section.

Technical white paper Page 10

Scality’s S3 API For enterprises and service providers, the S3 Connector offers the most secure, scalable, and portable storage services for custom and packaged cloud applications on the market today, as demonstrated in Figure 5.

Figure 5. Accessing data through an S3 interface

Comprising a set of distributed software-driven services, the S3 Connector offers three complementary functions: S3 API (S3 Server), security (S3 Vault), and a purpose-built metadata engine that is optimized to manage Buckets at cloud-scale (S3 Metadata). These key services are delivered as Docker containers for ease of deployment.

S3 Server The central component of the S3 Connector, S3 Server supports standard S3 API command URLs, enabling SSL certificate configuration for secure HTTPS access and encryption over-the-wire. It supports S3 headers, response codes, and error codes. Its processes are stateless. Therefore, applications can access any Bucket or Object resources through multiple servers for requests to ingest (PUT), or access (GET) object data, using standard load balancers.

S3 Vault S3 Vault provides identity management and access control security functionality for the S3 Connector. Adhering to the IAM model, the service affords a comprehensive AWS-compatible security service that is able to integrate with external enterprise directory and security services—for instance, Active Directory—through Security Assertion Markup Language (SAML) version 2.0.

S3 Metadata A distributed metadata database service, S3 Metadata stores system metadata in an ultrareliable, scalable, and high-performance manner (for example, data related to user’s Buckets and Object keys, as well as such S3 Vault security-related items like accounts, users, and ACLs).

The S3 API is delivered as a function of the Scality RING product. For development purposes, a free and open-source version of the API is also available at scality.com.

Technical white paper Page 11

Scality’s S3 functionality provides multiple services, as listed in Table 2.

Table 2. Services provided with Scality’s S3-compatible API

Rich AWS and Enterprise Security

Support for the full complement of AWS security services, such as multitenant accounts, IAM for users, groups, and roles, AWS-style access keys and secret keys, the latest Signature v4 Authentication mechanism, and data encryption. Also featured is interoperability with such existing enterprise security services as LDAP and Microsoft® Active Directory servers.

S3 API Compatibility Notwithstanding rapid AWS advancements, a high-degree of S3 API coverage is assured, including core data APIs for Bucket and Object access and Multipart-Upload (MPU) for efficient ingest of large objects. S3 Connector development is based on Continuous Integration (CI) and agile delivery of features when ready, which allows Scality to introduce new S3 methods shortly after their AWS publication. This functionality is provided by the S3 Server, which is supported by Scality as an open source project on GitHub.

Any-to-Any Scale-Out Applications can access any Bucket or Object from any Connector, thus allowing for parallel and multiuser access to data and scaling to billions of buckets and objects. Performance can be scaled-out simply by adding more connectors.

High-Performance Buckets Support for low-latency response times and high throughput of reads and writes of Objects in Buckets. Also, performance is optimized for fast Bucket listing operations, including fast partial-path search for selected objects by path prefix.

Geo-Distributed Capabilities S3 Connector provides integrated geo-replication capabilities for storage across multiple data centers, supporting active/active stretched deployments for site disaster protection with continuous data availability.

Ease of Deployment Delivered as a set of easy-to-deploy Docker containers, installation of the S3 Connector is simple, with zero-configuration across the customer’s choice of physical, virtual, or cloud environments.

Intelligent data durability and self-healing The RING is designed to manage a wide range of component failures involving disk drives, servers, and network connections within a single data center or across multiple data centers. The RING provides data durability through a set of flexible data protection mechanisms optimized for distributed systems, including replication, erasure coding, and geo-replication capabilities, illustrated in Figure 6, that allow customers to select the best data protection strategies for their data. The RING automatically manages storing objects with the optimal storage strategy. Replication and erasure coding may be combined, even in a single Connector, following user-defined policies. Small objects are stored more efficiently and at lower storage cost using replication. Large objects are stored more efficiently using erasure coding, avoiding the cost of replicating very large datasets.

Figure 6. Scality classes of service

Replication class of service To optimally store smaller files, the RING employs local replication with multiple file copies. The RING will spread these replicas across multiple storage servers using a unique disk drive for each copy in order to separate them from common failures.

The RING supports six Class of Service (CoS) levels (0–5) for replication, indicating that the system can maintain between 0 and 5 replicas, or 1-6 copies, of an object. This allows the system to tolerate up to “5” simultaneous disk failures while still preserving access to the object.

Replication is typically used only for “small objects” as defined by a configurable value. By default, objects less than 60 kilobytes will be replicated.

Technical white paper Page 12

Advanced resiliency configuration erasure coding Scality’s advanced resiliency configuration (ARC) provides an alternative data protection mechanism to replication that is optimized for large objects and files. ARC implements Reed-Solomon error correction1 techniques to store large objects with an extended set of parity “chunks” instead of multiple copies of the original object. The basic idea with erasure coding is to break an object into multiple chunks (m in number) and apply a mathematical encoding to produce an additional set of parity chunks (k in number).

The resulting set of chunks, (m+k in number) are then distributed across the RING nodes, providing the ability to access the original object as long as any subset of at least m data or parity chunks are available (see Figure 7).

Figure 7. Scality ARC: example of ARC (10/4) schema

Self-healing rebuilds performance under load The RING provides self-healing operations to resolve component failures automatically, including the ability to rebuild missing data chunks due to disk drive or server failures, and the ability to rebalance data when nodes leave or join the RING. In the event that a disk drive or even a full server fails, background rebuild operations are spawned to restore the missing object data from its surviving replicas or ARC chunks. The rebuild process is complete when it has restored the original CoS by restoring either the full number of replicas or the original number of ARC data and parity chunks.

Self-healing supplies the RING with the resiliency required to maintain data availability and durability in the face of the expected wide set of failure conditions, including multiple simultaneous component failures at the hardware and software process levels. For many customers, self-healing has eliminated the requirement for maintaining external backups which, in turn, reduces infrastructure and operational expenses.

Multisite geo-distribution To enable site-level disaster recovery solutions, the RING can be deployed across multiple geographically-distributed sites (data centers) with native support for synchronous (stretched) deployments across sites, with all the nodes participating as if they were local to one site. Either SOFS (File) or S3 (Buckets) may be used with the geo-distributed configurations. Stretched RINGS may be deployed with erasure coding, providing full-site failure protection, active/active access from all data centers, and dramatically reduced storage overhead compared to data mirrored across sites. A three-site stretched RING utilizing erasure coding can protect against one complete site failure plus one additional server failure, with just 70% of overhead.

1 See Reed-Solomon error correction

Technical white paper Page 13

Multisite stretched RINGs A Scality RING can be stretched across two or three sites within a Metro-Area-Network (MAN), as shown in Figure 8. These active/active configurations provide automatic full-site failover and recovery should the latency between sites rise above 10 milliseconds. The minimum physical topology is two storage servers per site.

Figure 8. Three-site stretched RING

For the two-site stretched RING, exhibited in Figure 9, when using file access protocols (NFS or SMB), two witness servers will be needed.

Figure 9. Two-site stretched RING

Technical white paper Page 14

Two-site asynchronous replicated RING For situations with higher latency between sites, a Scality RING can be replicated asynchronously across two sites, as presented in Figure 10. The RING supports a “Full Diff” mechanism that can compare at scale the content of the two sites to ensure the data is effectively replicated. Should a site in this configuration suffer permanent data loss (flood, fire), Scality RING provides a procedure to fully reconstruct the lost site.

Figure 10. Two-site asynchronous replicated RING

Asynchronous geo-replication for disaster recovery across WAN sites With RING 7.0, multisite replication is expanded to provide full data and metadata asynchronous geo-replication. This Cross Region Replication (CRR) for WAN and hybrid cloud alleviates latency issues between very long-distance sites, and can effectively be used to maintain a geographically remote disaster recovery site which stays in sync with the main data center. CRR utilizes S3 asynchronous bucket replication and supports replication to either remote enterprise RINGs or to public clouds (AWS and Microsoft Azure).

Versioning and security Beginning with RING 7.0, versioning with WORM change control is supported for both File and Object. RING 7.0 also introduces S3 Vault, which implements IAM multi-tenancy control for accounts, users, and groups and is interoperable with Microsoft Active Directory. More security is now available with bucket-level encryption utilizing an External Key Management Service (KMS). With S3 Bucket Location Control, it is now possible to designate the physical location of a bucket. In addition, low latency performance improvements are possible with S3 stretched sites, enabling read/write access to buckets from every metro-city site.

Entry-level configurations Beginning with RING 7.4, single-site entry-level configurations can be built with just three storage servers. This reduces the hardware investment needed for entry configurations by as much as 40%, as compared to six-server configurations with similar amounts of capacity. Entry-level three-server configurations support business applications using a single type of access connector (S3, NFS, or SMB). Three-server configurations protect data against two simultaneous disk failures and allow for maintenance needs where one server may be taken off-line for short periods time. When more availability is required such as multisite protection, or multiple access protocols will be used, continue to use six-server or larger configurations.

HPE value add for an object storage environment Software-defined storage running on Linux servers can be deployed on a variety of hardware platforms. However, clusters built on a white box server infrastructure may work for businesses at small scale, but as storage needs grow, the complexity and cost makes these white box systems harder to maintain than enterprise-grade hardware-based solutions. With white box server infrastructure, IT has to take ownership of standardizing and integrating platforms as well as supported components, and support escalation becomes more complicated. Without standardized toolsets to manage the hardware at scale, IT must chart their own way with platform management and automation. Often the result is the IT staff working harder and the businesses spending more to support a white box hardware infrastructure than the one-time capital expenditure (CAPEX) savings realized in buying the white box servers.

Technical white paper Page 15

Using an HPE hardware and software solution provides advantages that reduce operational expenditure (OPEX) spending not available in an infrastructure built on white box servers. Key OPEX savings from using an integrated HPE solution are:

• Platform management tools that scale across data centers

• Server components and form factors that are optimized for enterprise use cases

• Hardware platforms where component parts have been qualified together

• A proven, worldwide hardware support infrastructure

Customized factory integration HPE also offers customers a Factory Express program to integrate software with hardware and verify RING cluster operation before shipment. This program is available for RING configurations based on either the HPE Apollo 4510 or HPE Apollo 4200 servers. Customers may specify multiple custom parameters, such as networking IDs and admin access, they wish HPE to set up before the solution is shipped. The RING solution arrives preconfigured and customized, which streamlines the final on-site deployment and helps customers quickly get their object storage system ready for production.

Disk encryption In addition to the benefits of using the HPE platform as listed in the HPE value add for an object storage environment section, all HPE Apollo 4000 server configurations include an HPE Smart Array card capable of HPE Secure Encryption providing enterprise-class encryption. Secure Encryption is Federal Information Processing Standard (FIPS) 140-2 certified, and has been verified to have a low impact on IOPS for spinning media, in addition to being transparent to the operating system. This means data for any drive on the server can be encrypted, providing users with encryption, giving much more flexibility than encryption on drive solutions at a reduced cost. Keys can be managed either locally on the server or through an enterprise key management system.

Multigenerational RING support RING clusters support mixing multiple generations of x86 server storage nodes. For example, the HPE Apollo 4200 Gen10 server can be used to expand existing storage RINGs based on HPE Apollo 4200 Gen9 servers.

HPE Apollo 4200 Gen10 server reference architecture for Scality RING The base HPE Apollo 4200 Gen10 server is a perfect fit for customers looking for smaller fault domains in comparison to hyperdense storage. These boxes offer less data loss in the event of a node failure—rebuild time is decreased at the server-level. Although you can use the HPE Apollo 4200 servers at any scale, it is more likely to be used for storage capacity requirements under 2 petabytes or for enterprise customers that require a standard 2U form factor. Additionally, customers using colocation data centers can achieve higher density than general purpose servers, but are still able to utilize standard depth racks. The base architectures can be customized using the HPE sizing tools for Scality to build RING configurations with the ideal amount of bulk storage, metadata capacity, and memory performance. Work with your HPE account team to customize a RING configuration.

The HPE Apollo 4200 Gen10 server is the densest 2U server in the market today at up to 336 TB per server, all in a telco-compliant 1000 mm deep rack. It is a great choice when customers need to standardize on one server for many use cases. The HPE Apollo 4200 Gen 10 server has strong configuration flexibility, allowing customers to achieve a balance between expansion options and storage density.

This paper describes a base reference architecture with external Connector nodes, Storage nodes, and a Supervisor node. Each layer can be sized up or down, independently. External connector nodes are optional, and may be added to provide load balancing or separation of networks subnets. Your field engineer will advise you when you need to add these separate nodes. Figure 11 illustrates a typical I/O usage scenario in which there is one external connector server per three storage servers.

Technical white paper Page 16

Figure 11. Sample Scality configuration using HPE Apollo 4200 servers

Networking for the cluster is recommended to be 10GbE or faster. Application read-write throughput performance through an NFS connector can be significantly enhanced by using faster cluster networks to reduce the time needed to rebuild erasure-coded files. Hewlett Packard Enterprise has measured 2X or greater NFS application performance improvements for heavily utilized RINGs when upgrading the RING cluster from a 10GbE to a 25, 40, or 100GbE network.

HPE Apollo 4200 systems The HPE Apollo 4200 Gen10 server, is a storage dense platform recommended for use as Scality RING storage servers (see Figure 12). Key attributes of the HPE Apollo 4200 Gen10 server include:

• Chassis

– The HPE Apollo 4200 Gen10 is a 2 RU server that fits in a standard 1000 mm depth rack

– Uses HPE Gen10 Flexible Slot Power Supplies, which provides support for 800 W, 48 VDC, and 277 VAC environments, in addition to AC environments for 800 W and 1600 W Platinum and 800 W Titanium hot-plug power supply kits

• Processor

– Intel® Xeon® Scalable processors with Intel UPI, up to 24 cores, 150 Watts (8100, 6100, 5100, and 4100 series)

– Up to 16 DIMMs with 266MT/s DDR4 SmartMemory, up to 1 TB with 64 GB LRDIMMs

– Up to 2 NVDIMMs per processor

• OS drive controller/drives

– M.2 flash devices can be used for the OS drives

• Storage

– Supports a 4 LFF rear drive cage, which provides for up to 26 data drives (24 front-accessible, 2 rear-accessible) and 2 rear-accessible read drives for the OS

– Maximum storage capacity is 312 TB (26 x 12 TB) with two bays used for OS

– Integrated HPE Smart Array S100i

– HPE Smart Array Gen10 controllers, 3 maximum (optional)

• Networking

– Embedded dual 1GbE NIC

Technical white paper Page 17

• PCIe slots

– With the 4 LFF rear drive cage, supports up to 5 PCIe slots:

2 x PCIe 3.0 x 24 slots

2 x PCIe 3.0 x 16slots

1 x PCIe 3.0 x 8 slots

• Power Supplies

– Dual 800W or 1600W Flex Slot Power Supplies (AC/DC/277V AC)

• On System Management

– HPE iLO 5 Management Engine with 2 dedicated iLO NICs

– Optional iLO licensed capabilities:

HPE iLO 5 Advanced Premium Security Edition for premium security and automatic recovery of HPE Gen10 servers

HPE iLO Advanced for intelligent system tuning of HPE Gen10 servers

HPE iLO Amplifier Pack for discovery, inventory, and updating HPE servers at scale

• Data center support

– HPE OneView Advanced (optional)

• Cluster Management

– HPE Insight Cluster Management Utility (CMU [optional])

Figure 12. Front view of an HPE Apollo 4200 Gen10 system

HPE ProLiant DL360 Gen10 server The HPE ProLiant DL360 Gen10 is a low-cost, 1 RU server platform that is a perfect fit for the compute and memory requirements of the Scality manager and connector servers when external connectors are desired for load balancing or separation of network subnets. This is an optional component. Contact your HPE solution architect to help you determine whether you will need these systems in your RING configuration.

Technical white paper Page 18

Sample bill of materials (BOM) for HPE Apollo 4200 servers and HPE ProLiant DL360 servers

Sample HPE Apollo 4200 Gen10 BOM

Quantity Product Description

1 P07244-B21 HPE Apollo 4200 Gen10 24 LFF CTO Svr

1 P07943-B21 HPE Apollo 4200 Gen10 Rear Cage Kit (provides 4 LFF rear drives)

1 P08046-B21 HPE XL420 Gen10 Xeon-S 4114 Kit

1 P08046-L21 HPE XL420 Gen10 Xeon-S 4114 FIO Kit

4 815100-B21 HPE 32GB 2Rx8 PC4-2666V-R Smart Kit

1 727055-B21 HPE Ethernet 10Gb 2-port 562SFP+ Adptr

2 JD096C HPE X240 10G SFP+ to SFP+ 1.2m DAC Cable

1 813546-B21 HPE 2nd Cage FIO Ctlr Mode for Rear Strg

2 P04499-B21 HPE 480GB SATA RI LFF LPC DS SSD

1 877825-B21 HPE 1.6TB PCIe x8 MU HH DS Card

26 881787-B21 HPE 12TB SATA 7.2K LFF LP He 512e DS HDD

2 865414-B21 HPE 800W FS Plat Ht Plg LH Pwr Sply Kit

1 822731-B21 HPE 2U Shelf-Mount Adjustable Rail Kit

Sample HPE ProLiant DL360 Gen10 BOM (external connector)

Quantity Product Description

1 867958-B21 HPE ProLiant DL360 Gen10 4 LFF Configure-to-order server

1 860657-L21 HPE ProLiant DL360 Gen10 Xeon-S 4114 FIO Kit

1 860657-B21 HPE ProLiant DL360 Gen10 Xeon-S 4114 Kit

8 835955-B21 HPE 16GB (1x16GB) Dual Rank x8 DDR4-2666 CAS-19-19-19 RDIMM

1 727055-B21 HPE Ethernet 10Gb 2-port 562SFP+ Adapter

2 JD096C HPE X240 10G SFP+ to SFP+ 1.2m DAC Cable

1 804326-B21 HPE Smart Array E208i-a SR Gen10 Controller

2 861691-B21 HPE 1TB 6G SATA 7.2K rpm Gen9 (3.5-inch) SC Midline 1yr Warranty Hard Drive

1 875595-B21 HPE 800GB NVMe x4 MU SFF Scn DS SSD

2 865408-B21 HPE 500W FS Plat Ht Plg LH Power Supply Kit

1 789388-B21 HPE 1U Gen9 Easy Install Rail Kit

Summary With rapid growth of unstructured data and backup/archival storage, traditional storage solutions are lacking in their ability to scale or efficiently serve this data from a single unified storage platform. For unstructured data, the performance capability of traditional SAN and NAS vendors is often less important than the cost per gigabyte of storage at scale.

Scality RING running on HPE ProLiant and HPE Apollo hardware combines object storage software and industry-standard servers to provide low cost, reliable, flexible, centralized management that businesses need for large-scale unstructured data. The HPE Scalable Object Storage with Scality RING creates a solution with a lower TCO than traditional SAN and NAS storage vendors, while providing greater data protection for current and future large-scale storage needs.

Technical white paper

Share now

Get updates

© Copyright 2015-2017, 2019 Hewlett Packard Enterprise Development LP. The information contained herein is subject to change without notice. The only warranties for Hewlett Packard Enterprise products and services are set forth in the express warranty statements accompanying such products and services. Nothing herein should be construed as constituting an additional warranty. Hewlett Packard Enterprise shall not be liable for technical or editorial errors or omissions contained herein.

Intel Xeon is a trademark of Intel Corporation in the U.S. and other countries. Microsoft is either a registered trademark or trademark of Microsoft Corporation in the United States and/or other countries. Linux is the registered trademark of Linus Torvalds in the U.S. and other countries. All other third-party marks are property of their respective owners.

4AA5-9749ENW, January 2019, Rev. 6

Resources, contacts, or additional links HPE ProLiant DL360 Gen10 server hpe.com/us/en/product-catalog/servers/proliant-servers/pip.hpe-proliant-dl360-gen10-server.1010007891.html

HPE Secure Encryption hpe.com/servers/secureencryption

HPE Integrated Lights Out (iLO) hpe.com/info/ilo

Learn more at hpe.com/storage/scalableobject