Original URL: http://www.theregister.co.uk/2013/04/09/blind_spot/
Mutant array upstarts feast on EMC, NetApp's leavings
Should've rewritten the software lads...
Blocks and Files Mainstream storage vendors have a potentially huge blind spot in their product strategy and hybrid array startups are now eating away at their customer base. Nimble Storage, Tegile and Tintri, specifically, appear to be benefiting at the expense of mainstream storage giants EMC and NetApp.
When the storage array mainstream met flash they did one (or more) of three things. First, they stuffed SSDs into disk drive slots in their arrays (EMC). Secondly, they flash cached their arrays (NetApp) so that, in both cases, data access requests to hot data took less time. Neither tactic meant rewriting the array's operating system.
The third tactic was to add an all-flash array to their product portfolio. EMC bought XtremIO to do this and willingly added a new HW/SW silo to its line-up. NetApp initially flashified an Engenio drive array and used existing SANtricity software to drive it, but is also developing its brand-new FlashRay all-flash hardware and new software.
A dividing line here is that using SSDs and/or a flash cache in existing arrays required no basic software rewrite, whereas building (or buying) a new all-flash array needed ground-up designed software. The option they decided against was to add flash to an existing disk drive array and re-write the software.
This was the hole that Nimble, Tegile and Tintri detected and are now pouring products through.
These startups decided that customers would rather buy hybrid flash/disk drive arrays than invest in expensive all-flash arrays, using flash for speed and disk for bulk data, with new software to use both to best effect. Where EMC, NetApp and others built hybrid arrays by adding flash to their legacy disk drive array HW and SW, hybrid startups re-invented the hybrid array with new software.
Tintri did it by focusing on virtualised servers, with its unit of VM-aware storage and point of admin being the virtual machine (VM) and not LUNS, volumes or RAID Goups. Nimble Storage focussed on lengthening flash's endurance by sequentialising writes and simplifying array management by instrumenting its software and having it send logs multiple times a day to a central Nimble cloud vault; a Nimble array can send in 30 million data points a day.
This data is analysed so that performance optimisation suggestions, capacity upgrade recommendations and failing components can be identified without customers needing to do any work themselves.
Tegile's Zebi array uses both DRAM and flash as cache layers with SAS disk used for nearline capacity. Its MASS software carries out deduplication and compression with metadata stored in flash to cut processing time.
These three suppliers say they are growing fast, with product that is faster than legacy hybridised arrays and also cheaper. Tintri had more than 150 customers, who had deployed 35,000 VMs, at the end of 2012. Now, at the end of the first 2013 quarter it has 53,000 deployed VMs, 18,000 more and we expect it to pass the 100,000 deployed VM mark in the third quarter.
Nimble Storage is growing fast too:
- Jan 2011 - few systems sold and 66 customers
- Jan 2012 - 506 sold and 276 customers
- Jan 2013 - 2,025 systems sold and 1,107 customers (around 400 per cent growth in each category)
All three basically use the same hardware mix as flashified VNX and FAS arrays. Admittedly the flash hardware can differ, but the big difference is the ground-up designed software. Get this:
- Neither EMC nor NetApp can readily retrofit VM-aware storage abstractions to their storage SW. Score one to Tintri.
- Neither EMC nor NetApp can retrofit Nimble's sequentialisation of flash writes or its software sensing ability to their storage SW. Score one to Nimble.
- Neither EMC nor NetApp can retrofit efficient primary data deduplication to their legacy storage. Score one to Tegile.
The only technical way EMC, NetApp, Dell, HDS, HP and IBM, can compete with these three startup hybrid array vendors is by re-writing their incumbent array software; a massive task that will take years to accomplish. They could gain new hybrid array technology by acquiring one of the three startups, or they could stick their heads into the sand and say that, although the three are growing fast, they are tiny minnows compared to the great whales that are EMC, NetApp et al.
But disruptive technology startups are always minnows. Then they become sprats, then bigger fish, then even bigger fish, and then whales themselves.
Of course, it's easy to recognise disruptive technology with hindsight. Detecting it when it's happening, especially when zillions of wannabee startups fail, is far more difficult. Incumbents never willingly shift their fat product arses; it's uncomfortable, expensive, and a gamble.
But when they say such-and-such isn't truly disruptive technology, is it their wise-far-seeing technology vision that has spoken, or are they blowing smoke out of their arse? We'll see. ®