Timing is everything in this game and following my last post (by mere minutes) on Pure Storage’s new FlashArray announcements, Chris Mellor at The Register wrote an article discussing Pure’s future evolution.
The inevitable comparison is made with NetApp and their “single” platform of 20 years, namely Data ONTAP. Forgive me for sounding like a broken record, but we know in fact that Data ONTAP is actually two platforms which were initially unsuccessfully melded together and eventually became harmonised by making the feature set consistent between the two incarnations.
NetApp was successful in terms of building a single-product company for a number of reasons including:
- Their focus was NAS, not SAN. NetApp introduced filers, not shared storage arrays and delivered CIFS/SMB & NFS support. Only over time did block connectivity evolve and that was due partially to NetApp providing iSCSI support for free.
- Data ONTAP was different. The platform had (writeable) snapshots and a “write-new” architecture that didn’t overwrite data in place, but rather stored all updated data as a new block, with cleanup/garbage collection for data that was invalidated. This made de-duplication and snapshot implementation elegant and efficient (however it’s also what made block device support more difficult). We shouldn’t forget the write technique of full stripe writes from NVRAM which made RAID protection more efficient, something we see emulated into today’s all flash arrays.
- They built an ecosystem. NetApp spent time building an ecosystem around their platform with products like Open Systems SnapVault (OSSV), which uses the NetApp filer as a target for backups.
- They built a community. NetApp made their product accessible. The NetApp simulator was made freely available and anyone could practise and write scripts and code, tested against a simulator machine.
- They created lock-in. Technology such as SnapVault deliver highly efficient backups through snapshot consolidation but are proprietary and can’t easily be moved elsewhere without data re-hydration.
Data ONTAP and the “single” platform is looking a little old in the tooth. The legacy component, ONTAP 7-mode is being retired in favour of clustered ONTAP, which has its genesis in the Spinnaker acquisition. We now see NetApp creating new platforms, but still struggling to move away from their core cash generating business.
Chris talks about a number of options for Pure Storage:
- Out-develop all other AFA companies. This is a tricky one. NetApp chose a different market route (NAS) which meant they had little competition in the early days. The current flash market is oversubscribed with companies making flash products, all with different niches; scale out (SolidFire), high performance (Kaminario), O/S integrated (Violin), plus the big incumbents who have re-engineered their products or are developing new platforms.
- Add Scale-out to Scale-up. This is a logical approach and we’ve seen the release of the FA-405, a “baby” flash array, which could potentially be the genesis of a scale out solution. As already discussed, this is just another string to Pure’s bow – a scale out product would need to offer the same feature set (and crucially, physical layer compatibility) as the scale-up options.
- Upgrade to 19nm flash. This is merely a step along the evolution of flash and not a game changer. All the vendors claim “special relationships” with the flash manufacturers, but they’re just friends with everyone. There’s no competitive advantage here.
- Add PCIe interface. This is an interesting middle ground we don’t really have yet. EMC tried to develop a product, but claimed it wasn’t required by customers, but perhaps they just did it wrong. As we’ve seen with the acquisition of DSSD by EMC, this market is set to come alight again and in fact, with products like RONNIEE, PCIe connected storage may have a chance.
- Add File interfaces. I don’t think this one is a relevant discussion. File support can be easily achieved by deploying a gateway in front of existing hardware – EMC, HDS, HP and others all do it. This wouldn’t be a game changer. File has a much lower I/O demand than block, so using all flash for file support would be a waste.
- Add back-end bulk capacity. This seems like another non-starter. SolidFire have added cloud backup to their platform, but I’m yet to be convinced of the real benefit of this feature.
- Explore TLC cold storage. This is a possibility; but the difficult issue will be around the cost/benefit calculation. TLC would have lower power demands than say 6TB drives, however drive capabilities are increasing continuously with 10TB on the horizon and tape capacities of hundreds of terabytes a real possibility. TLC flash for cold storage probably won’t stack up as a viable option on its own.
- Grow and Extend Software Capability. This is the most likely area for opportunity. Whilst we saw functionality move up to the hypervisor, with scale out server deployments that use NoSQL and Hadoop-style distributed storage, there’s the option to bring functionality (and intelligence) back into the array. NetApp created one or two killer software features. Pure needs to do the same.
- Be leaders in product support. This is another table stakes discussion. Excellent support should be assumed rather than be a value add. NetApp, HP, HDS and EMC have massive support organisations. Matching these is simply keeping up with the competition.
- Continue extrovert marketing. Well this process works for EMC, who have the most “creative” storage marketing teams on the planet. However even that isn’t working as well as they would like, as recent results have shown.
This list doesn’t include acquisition and doesn’t discuss partnerships. NetApp have a wide partner ecosystem and are trying desperately to give their products relevance by forging alliances with cloud vendors. Acquisition must surely be an option for Pure, even if its not their primary one. Assuming they really, really don’t want to be acquired, what’s left?
One area that Chris hasn’t discussed is diversification. Initially Pure could diversify across the market and release scale out, NAS and even object support. However the big question is whether a new storage company can ever survive selling just storage hardware. The decade that started at the turn of the new millennium (from 2000 onwards) were the golden years for siloed IT functions (server & O/S, storage, networking), but virtualisation changed that. Software Defined Data Centres (SDDC) as they evolve will change that again and being a single tower vendor doesn’t look like a long term business. EMC, HP, HDS, IBM & Dell are all diversified businesses. NetApp isn’t and although they are big, that lack of real diversification is still their weak spot.
The Architect’s View™
Pure will do well for a number of years as they take business from the ailing high-end storage market. However that market is shrinking as lower cost hardware becomes more reliable. Platforms like VSAN are in their early years and they surely will develop, as will the open source market, as companies like Red Hat promote Ceph and GlusterFS. Diversification and innovation will be the key. I for one am looking forward to seeing what Pure announce this time next year in their attempt to continue the growth of their business.
Pure Storage sets out stall: We want to rule the storage world (The Register, 19 May 2014)
Copyright (c) 2007-2021 – Post #F5D2 – Brookend Ltd, first published on https://www.architecting.it/blog, do not reproduce without permission. Photo credit iStock.