For EMC, it’s all about the rack-scale flash

DSSD does it all

Sakac said: "DSSD doesn’t require any of those file/block semantics between the flash read/write model. It can expose this via libHDFS or object semantics, or directly mapping to key value stores (with a PCIe/NVMe connection). If you want direct memory mapping over RDMA and over direct PCIe NVMe, it can do that too!"

This is building on the ideas of flash as memory espoused by Fusion-io, with its PCIe ioMemory flash. Fusion-io co-founder David Flynn said that applications could access data orders of magnitude faster if they didn't have to route access through the traditional OS's IO stack.

According to Sakac DSSD is "the 'ultra hot' edge – infrastructure that supports the in-memory database like SAP HANA, Gemfire, and key value stores such as memcached. DSSD changes the economics of in-memory databases as you can have far more memory at a different economic point. DSSD changes the envelope of IOPS and latency for the very low latency-dependent distributed NoSQL/SQL databases used for real-time analytics."

We're looking at HDFS access, block access, object access and in-memory use.

A thought springs to mind, namely: will one server access the DSSD flash array? With petabyte-plus capacity and microsecond-class latency surely lots of servers could use it, if the interconnect permitted multiple server access.

Perhaps, stretching credulity maybe, we should think of DSSD as the SAN reinvented as a quasi main-memory extension, with vastly faster network access and a memory-type access stack as well as HDFS. If this is realistic then we are looking here at an embryonic disk SAN killer tech.

EMC expects DSSD product revenues to grow faster than XtremIO's. That flash array has reached a $1.2bn bookings run rate this month. It shipped its first product in May 2013, 21 months ago. If DSSD fulfils EMC's hopes then we're looking at, say, a $1bn run rate in 18 months from first ship. Is that feasible? It seems incredible.

If it is true then that would give activist investor Elliott Management reason to pause its VMware-spin-off-based assault on EMC and wait to see if DSSD living up to EMC's expectations boosts its share price.

Let's give EMC the benefit of the doubt and say that with DSSD it has a real SAN-killing technology for IO-intensive and compute-intensive petabyte-class storage use-cases. Then what does that mean for other suppliers?

All in-memory application processing will be attracted to DSSD, meaning the SAP HANAs. EMC will have the alternative to Oracle's engineered systems.

Dell, HDS, HP, IBM and NetApp face being cut out from the in-memory app storage market and would need to buy or develop their own technology. But buy from who? Hey, didn't SanDisk buy Fusion-io, they might think? Let's give Sanjay Mehrotra a call.

Could Violin build a PCIe access route and have its array treated as a DRAM extension? Let's call Kevin DeNuccio. HGST's Skyera has the hardware density needed but what about the software? Maybe there's a partnering opportunity there. It could all get a bit frantic.

There is bound to be a huge hoopla of a DSSD launch by EMC. Prepare to be deafened by a DSSD deluge. Resistance will be futile. Give up now. ®

Note If you like the idea of marketing this product then EMC is looking for a senior marketeer to do just that.