Data Center Tech Blog

Flash technology was a major touchstone for discussion at the recent IBM Edge2014 (EDGE) conference in Las Vegas. The topic came up again, and again – from the keynote by IBM Executive Vice President Adalio Sanchez, to the storage-focused morning session on Day One of the conference – and continuing in numerous breakouts and panels.

Why was flash such a high-priority topic at EDGE? For several reasons, really. IBM has an over-arching flash strategy for servers and storage – and it plans to apply the benefits of flash to enterprise computing, to cloud computing, to high performance computing (HPC) and to variable workloads that provoke random I/O access that simply cannot be predicted.

As in the industry model for Geoffrey Moore’s book “Crossing the Chasm,” early adopters of flash as an enterprise and cloud computing accelerator are gaining real-world results. But wider, and broader, adoption will be seen in the future, as flash-enablement of the servers and storage devices increases to meet the high processing demands for Big Data and Analytics.

There are specific benefits of flash for workloads that are especially demanding, such as financial services applications, real-time fraud detection, and business analytics. For these workloads, the reduced latency inherent in flash technology, and the higher IOPS rates, provide very real business results that impact everyday business operations.

IDC and Gartner, to name two market research firms, agree on the growth pattern that’s expected for flash – broader and wider and deeper adoption, across the tiers of the datacenter;  They expect growth in revenue throughout their forecast period, ending in 2017. It’s happening because many storage systems need to scale with more capacity – and to perform much faster —than HDDs alone would allow.

Flash As a Highlighted Topic at EDGE

In his session, IBM’s Adalio Sanchez spoke of the IBM eXFlash flash storage, which resides on the memory bus of IBM’s xSeries servers (e.g. IBM x3850, x3950), leveraging technology from SanDisk® Corp. and Diablo Technologies.

This innovative use of flash to accelerate business computing places high-capacity flash drives directly on a system’s DIMM slots. This approach allows the flash storage on the memory bus to scale up to 12.5 TB, and beyond, depending upon system capacity and workload demands.

To learn more details about this solution, you can view the presentation by Rob Callaghan of SanDisk “Utilizing Ultra-Low Latency Within Enterprise Architectures” here, if you are registered EDGE participant. For those who did not attend the event, you can learn more about this technology here.

Scaling linearly with ULLtraDIMM

Overall, IBM has a multi-faceted flash strategy, including the addition of the eXFlash in its xSeries servers, its 2013 acquisition of Texas Memory Systems – and IBM’s development of more flash-based server and storage products leveraging SSDs.

Clod Barrera, IBM’s chief technical strategist for IBM System Storage, speaking that first morning at the EDGE conference during the Storage Trends Futures and Strategy session, said that flash is coming into its own providing extreme performance that supports an improved end-user experience given faster results.

That’s going to become increasingly important, as Mobility and Social Media and Cloud Computing take off in the datacenter – and as these megatrends transform other enterprise data services and cloud data services.

Barrera predicted that there will be a “tipping point” for flash in 4-5 years (post-2018) – meaning that its deployments will accelerate, over time. “Flash is breaking new ground, and changing how we look at IT infrastructure and how we define performance.” He said flash would help to remove storage-related bottlenecks from data center systems.

Dealing with the Data Tsunami

What’s driving this adoption of flash? A need to close the storage bottleneck caused by pent-up demand for faster data delivery in a datacenter that’s seen rapid speedups for processor and network switching technology in recent years. Add to that the data tsunami hitting the data center itself – and the rapid adoption of cloud computing – all of which benefit from the addition of flash storage.

At EDGE, The voice of the customer about flash deployments was heard through breakouts and panels. Andy Walls, an IBM distinguished engineer who is CTO and Chief Architect of Flash Systems Hardware, outlined his vision for flash adoption – highlighting Analytics, OLTP and VDI as three workloads that benefit greatly from flash.

Walls moderated a panel of three IBM systems customers that had deployed flash – and all three saw improvements in system performance. One was an online dating site that has been able to scale its service quickly, using flash technology—with hundreds of TB in its data center infrastructure. The online company was able to reduce software licensing costs by consolidating workloads on flash.

Another was a consumer goods company that leveraged flash to do real-time data compression more quickly and effectively. A third company did a proof-of-concept (POC) with flash in an HPC computing environment to reduce I/O latency associated with using a large clustered file system, improving grid computing speeds.

All three customers in the EDGE panel said their organizations experienced performance improvements with flash technology, compared with hard-disk drive technology, and improved I/O operations per second (IOPS), making storage much faster. Importantly, all three saw the system reliability in their flash-enabled servers and storage. Downtime was not an issue, nor was endurance, given close monitoring of their flash-enabled systems.

Flash – Extreme Performance

racing car fast

These customers saw what even a small addition of flash in a cluster or storage array can do to transform overall performance of an important workload. In fact, many “flash adopters” are using flash in a hybrid environment, including both HDDs and SSDs in the same server cluster or storage array.

A single HPC workload or Hadoop workload can speed up with flash – reducing processing time and reducing operational costs related to data center space, power and cooling (see the SanDisk Hadoop Solution Brief).

For database workloads, the customers noted that I/O waits and SQL-query-related related waits were reduced dramatically in the flash-enabled systems supporting the database.

Key Takeaways

Leveraging flash as a workload accelerator is gaining traction in the datacenter, given its stronger support for service level agreements (SLAs) via longer usable product life, and its use in production applications.

The flash deployments described at IBM Edge2014 are typical of the assessments of many customers deploying flash technologies in the enterprise.

High on the list of workloads that get faster with flash: Virtual Desktop Infrastructure (VDI), Database, Analytics and HPC workloads are increasingly seen as candidates for new flash-enabled systems deployments. In that continuing process of workload identification, and proofs of concept for new flash deployments, the gap from the industry model of “Crossing the Chasm” is rapidly closing – and the era of greater flash adoption is upon us.

SanDisk at HostingCon 2014

Hadoop Means Business

Subscribe Today!

Get our latest posts via email and enjoy data center insights in a flash