TheCUBE hosts David Floyer and Dave Vellante of Wikibon caught up with Ron Asher, Director of XIV WW Development for IBM, during a live broadcast from IBM Edge in Las Vegas this week. Ron is responsible for leading product innovation, development and support.

In the early days, the design of XIV was taking low cost disks and SATA-based disks and making them Tier 1 storage devices, optimizing their performance and turning them into very cost-effective tools. When XIV started, only 1 TB drives were available. There wasn’t a very big difference in the capacity of fast-spinning drives compared to the slow spinning ones. Nowadays, XIV boasts that they are able to decrease power consumption and boost up performance.

The DNA of the architecture allows seamless spreading across all the resources. In the past one could boost up performance of slow spinning drives through caching, but Flash allows to take this to a bigger level. For instance, an XIV system without Flash can do 40K IOPS in random workload (which is like SPC-1), but with Flash it can do 180K IOPS. This is almost delivering 1K IOPS on a slow spinning drive that can physically do only 200. That is a huge improvement.

The architecture of XIV and Flash allows for 15 Flash devices working for the same workload at once. With traditional architecture, that would be difficult to achieve.  See Ron’s entire segment on flash-driven architecture below.

There is no write front-end at the moment, only a read Flash. The huge benefit of Flash is for random read. In the future, it is intended to develop towards write, and maybe towards an internal Flash storage Tier (hidden from the user). The biggest challenge in terms of storage tiers is that at the end of the day you want to have predictable results for your workload; that means predictable latency and predictable IOPS. With two layers of storage, HDD and Flash, that is difficult to attain. XIV is attempting to make that as smart as their caching, and it’s going to be completely transparent.

The system will be automatic, deciding which data is stored where, but if a client needs something on Flash-only, this service will be provided.

XIV is using techniques to leverage low cost disks, to minimize and mask the bottleneck of spinning disks. With the introduction of Flash, some will argue that the t4echnology was not designed for this use. Asher clarifies that using the Flash media is different than using the spinning media. There are a lot of optimizations that are applied only to the spinning media (such as sequentializing the workload and caching). But the fundamental architecture of workload being broken down into small pieces and spread out across all the resources, and cached, that can also be applied to Flash. The advantage of using Flash is that if a piece of information fails, there is no need to recover the entire media, just the piece that failed. This is the advantage of the XIV architecture, that has nothing to do with spinning drives.