In February 2021, IBM announced the FlashSystem 5200, a 1U entry-level storage system with all the capabilities and resilience of the whole FlashSystem family. In this post we look at the hardware architecture and explain why the 5200 is arguably different from other solutions on the market.
Around 12 months ago, we started a series of posts looking at the capabilities of the FlashSystem range of storage solutions from IBM. The entire product line had been through a refresh that transitioned from the Storwize technology to use San Volume Controller (SVC) or Spectrum Virtualize as the underlying storage operating system. We looked at the hardware, software, and new approaches to sharing pricing online with prospective customers.
- IBM FlashSystem Review – Part 1 – Hardware
- IBM FlashSystem Review – Part 2 – Software
- IBM FlashSystem Review – Part 3 – Ease of Use
These posts (linked above) have recently been refreshed with details of updates to the FlashSystem line. One new system that stands out is the FlashSystem 5200, a 1U storage platform that offers the same features and resiliency as the higher capacity solutions in the FlashSystem family. The 5200 model is worthy of additional discussion as this offering can be used in a wider set of use cases, which we’ll discuss in a follow-up post. Here, we’re going to look at the hardware and software capabilities, including with reference to the entire portfolio.
If you take a look across the storage appliance market today, solutions generally fall into one of two form factors. Most enterprise systems start at 2U (2 rack units in height), with some models built in 4U and then combinations thereof. The reason for this design is two-fold. Firstly, enterprise systems must accommodate a minimum of two controllers to meet an acceptable level of availability. Second, system design dictates either vertically mounted 2.5” drives, which require 2U, or horizontally mounted drives that can accommodate 2.5” or 3.5” models. Although the “ruler” format we discussed four years ago could offer a better density, we’re not yet seeing widespread adoption in the enterprise.
In the 1U segment of the entry-level market, solutions from vendors including Synology and QNAP are generally targeted at SMB/SME customers. Most are single-socket, single controller designs that can’t match the availability requirements of the enterprise.
The FlashSystem 5200 enclosure design accommodates dual controllers, or in IBM terms, canisters. The architecture is shown in figure 1. Each canister has access to all 12 2.5” NVMe drives, which can be FlashCore modules (4.8TB to 38.4TB) or traditional NVMe flash drives, with capacities from 800GB to 15.36TB. Storage Class Memory (SCM) devices from 375GB to 3.2TB are also supported. Obviously, the NVMe protocol is only available on flash and SCM devices so in the base configuration, FlashSystem 5200 does not have a hard drive option. Drives are dual ported, so connected independently to each canister.
FlashSystem 5200 supports extension enclosures for both hard drives and flash drives, either as a 2U or 5U system. However, these drives are connected via SAS expansion rather than NVMe.
Delivering high availability means providing redundancy and the capability to hot swap failed components. The 5200 has dual hot-swappable canisters, redundant hot-swappable batteries in each canister, hot-swappable PCIe adapters and redundant hot-swappable power supplies, DIMMs and fans. A single canister has 2x Skylake Intel Xeon CPUs at 2.3Ghz and from 32GB to 256GB of DRAM. Every component is effectively hot-swappable, with integral batteries used to commit data to disk in the event of a power failure.
The use of on-board batteries is now possible within storage systems due to the lower power demands of SSDs compared to hard drives. Additionally, with a maximum of only 12 drives in the main enclosure, the power demand needed to commit in-flight data to persistent media is much lower than required in the past.
Although a single system provides a high degree of availability, all FlashSystem systems offer clustering. The 5200 can be clustered with two, three or four enclosures, either locally or across distance using HyperSwap. The ability to cluster is important when we discuss edge and remote configurations in the next post. Obviously, the cluster capability represents both an increase in performance scaling as well as increased availability.
Performance and Capacity
As highlighted earlier, a single FlashSystem 5200 can support up to 12 NVMe drives, with a range of supported configurations including NVMe SSD, FlashCore modules or SCM (based on Intel Optane or Samsung zNAND). With 12 FlashCore modules, a single 1U system supports up to 403TB of usable (460TB raw) capacity or close to 1PB if a typical 2:1 compression ratio is achieved in the FlashCore modules. IBM claims a single system can deliver up to 1.5 million IOPS with latency as low as 70µs and 21GB/s of throughput. (Note: latency figures generally refer to 100% read I/O, so it’s worth asking IBM for more information on performance capabilities).
With the advent of flash storage, all storage system vendors have published so-called “hero numbers” which show throughput, performance, and latency in the best light. Whilst we caution taking these numbers on face value, modern storage systems provide an unprecedented performance capability that generally exceeds the needs of most applications, especially those in edge and remote office locations.
FlashSystem 5200 supports a wide range of connectivity and performance options. This includes four 10GbE iSCSI ports as standard (2 per canister), with expansion options for 16/32Gb Fibre Channel and 25Gb Ethernet. Storage protocol support includes iWARP and RDMA over RoCE and NVMe-oF.
In an edge or remote location, systems are more likely to use Ethernet-based storage protocols, simply because this design reduces complexity and offers more consolidated deployments. However, having the option to deploy additional protocols provides the ability to use Fibre Channel tactically, if required. Ultimately, the option to choose protocols is available, if needed.
In our previous posts, we used the following table to summarise the capabilities of each member of the FlashSystem family.
|FlashSystem 5015||FlashSystem 5035||FlashSystem 5200||FlashSystem 7200||FlashSystem 9200|
|IBM Spectrum Virtualise Software||⊕||⊕||⊕||⊕||⊕|
|IBM Storage Insights||⊕||⊕||⊕||⊕||⊕|
|VMware & Red Hat Openshift Container Integration||⊕||⊕||⊕||⊕||⊕|
|Local & remote replication||⊕||⊕||⊕||⊕||⊕|
|IBM Easy Tier||⊕||⊕||⊕||⊕||⊕|
|Transparent Data Migration||⊕||⊕||⊕||⊕||⊕|
|Data Reduction Pools||⊕||⊕||⊕||⊕|
|HyperSwap high availability||⊕||⊕||⊕||⊕|
|NVMe flash and FC-NVMe host connections||⊕||⊕||⊕|
|High -performance compression & encryption in FCM||⊕||⊕||⊕|
|External storage virtualisation||⊕||⊕||⊕|
As we can see, the FlashSystem 5200 is fully featured, with no compromise in capabilities compared to the higher-capacity models. “Compromise” is perhaps the key word here, as the use-cases for 5200 may include less secure locations where data encryption is essential.
Equally important is the ability to remotely support systems using standard tools like Storage Insights and to have the same levels of application support and integration available in core data centres. At scale, consistent supportability is a must and lack of standardisation can represent significant additional cost implications.
The Architect’s View™
In the early days of centralised storage, many vendors took the decision to build and support multiple products within their portfolios. Each solution required a different method of configuration and operation. At scale and across many geographic locations, the need to change platforms just to gain increased capacity or performance represented an operational issue for growing businesses. The leading storage vendors in today’s market have realised that customers want consistent look, feel and operations from their storage platforms, with the ability scale from tens of terabytes to petabytes without forklift upgrades.
We see the FlashSystem 5200 as an interesting solution because the hardware doesn’t compromise on features, functionality, or availability. At the same time, IBM customers can use the solution in a range of scenarios where centralised storage is either impractical from a cost, support, or environmental perspective.
In our next post, we will explore exactly what the “no compromise” scenario means when looking at a range of deployment scenarios, such as edge, remote office and core data centres.
For more details on FlashSystem 5200, click here.
Copyright (c) 2007-2021 – Post #c7e4 – Brookend Ltd, first published on https://www.architecting.it/blog, do not reproduce without permission. This work has been produced with sponsorship from IBM.