CDOT relatively crap for flash, hyperscalers crap for constant storage
NetApp CTO lays it on the line
Chief NetApp techie Jay Kidd had some strong words for the flash and cloud crowd at a Wells Fargo event for investors.
Among other things, he covered Clustered Data ONTAP and flash and the customer spending slowdown. Not a great fit and a transitory slowdown were the take-away messages.
The event was the Wells Fargo Tech Transformation Summit and the NetApp CTO answered questions there, starting off with the cloud. How much of a threat was it?
Kidd thought that cloud storage economics didn't stack up for mainstream enterprises except for transient workloads like test and development. He said recent price cuts by Google and Amazon "didn't really change the dynamic at all". Google and Amazon cost 2.5 cents/GB for archive class (S3) storage. Over three years that equated to 80 to 90 cents/GB: "You can buy and manage enterprise-class storage for well less than 90 cents/GB" for this use case.
Cloud block storage like EBS is around 10 cents/GB, which equates to $4/GB over three to three-and-a-half years: "You can certainly buy enterprise storage for well less than $4/GB."
His view is that Amazon and Google - the hyperscalers - storage clouds are a good fit for transient and spiky requirements but not for constant or slowly growing storage needs. He said: "Amazon and Google reduce the cost of failure. They significantly increase the cost of success."
Disaster recovery workloads could go to the cloud but "I don't think there is a permanent workload move from enterprises to hyperscalers because the economics don't work."
Asked about a slowdown in customer spending, Kidd said the cloud decision had become more complex and customers were spending less while evaluating the many cloud alternatives, such as full public cloud, SaaS, various flavours of hybrid cloud with MSP contribution, and full on-premise private cloud IT. It used to be a binary decision – cloud or not cloud – but it's not any more.
"Before customers refresh on-premise (storage) they'll evaluate keeping data on-premise or sending it to the cloud. [This] slows down purchasing."
What about competition from smaller guys [like Nimble Storage and others]? "Competition from smaller guys is generally in the smaller accounts. ... We don't see this in the $10m - $20m accounts ... where the bulk of storage is bought. ... I don't think there's a share shift in the market... IDC says we're gaining share."
"We're a 15 - 6 per cent share player. When we make our storage more efficient customers move to NetApp." It would be different if NetApp was a 60-70 per cent share player.
This implies that, although cloud considerations are slowing purchases by NetApp customers, they will come back because cloud storage economics work in NetApp's favour. Kidd is not seeing NetApp's growth hobbled by storage startups so there's no need to respond in any special way to them.
Also NetApp sells to providers of enterprise-grade virtual private clouds, having some 200 of these customers and Clustered Data ONTAP (CDOT) can be used to manage on-premise storage and storage in the virtual private clouds, providing a single data management environment. The hyperscalers can't do this.
NetApp and flash
The discussion then moved on to flash, which Kidd says is a nascent market (meaning NetApp isn't behind, I think). The main customer moves here are a transfer if latency-sensitive database apps, typically loaded up on a Symmetric-type large array, to all-flash arrays (AFAs) for faster speed. Kidd said: "We've had great success with the EF540/550 [and] we've sold a good number if all-flash ONTAP systems [in one case winning against XtremIO] ... We're excited about FlashRay ... a great $/GB offering ... with MLC flash ... and hundreds if thousands of IOPS. ... FlashRay will be icing on the cake."
Then Kidd said something very interesting to us NetApp watchers: ONTAP has been extended to support a wide range of workloads but it would stretch it too far to cover flash, as its primary focus is disk. FlashRay and the EF540/550 systems can run and use flash better than ONTAP.
Our take is that, although an all-flash ONTAP array goes faster than a disk ONTAP array, it goes slower and performs overall than either the EF540/550 SANtricity OS systems or the coming FlashRay. Our understanding is that FlashRay will have ONTAP "goodness" but not be an ONTAP system. It will be a new architecture and, we expect, perform and manage its flash better than the EF540/550 systems as SANtricity is, at heart, a disk-focused OS.
Analytics and engineering focus
Kidd was also asked about data analytics: should NetApp sell data analytics products as well as storage infrastructure? He said NetApp partners with Teradata, which buys the V-series, for data analytics. What used to be a big data market is becoming an analytics market and it "doesn't lend itself well to our traditional salesforce. ... We think we are continuing to grow in storage and data management [infrastructure]" and won't move into data analytics.
Lastly, Kidd talked about how engineering budgets would be modified following the recent restructuring of 600 positions. More attention (spend) would be paid to FlashRay, CDOT, cloud investments, virtual storage appliances (CDOT as SW-only) and object storage.
For NetApp, the future crudely looks like disk arrays for capacity and flash arrays for performance – both on-premise and in enterprise virtual private clouds. We think this will be accompanied by a wary and calculated co-operation with public cloud hyperscalers like Google and Amazon. If customer spending comes back, then this should be fine. If it doesn't, then it's back to the drawing board to devise Plan B. ®