Ok – once the marketing panache settles (view it as good, as bad, as fun, as evil – your call – but it does what marketing does, and does it well) – what’s the TECH behind what was announced today?
There are 4 big parts:
- Part 1 – Next Generation VNX.
- Part 2 – Important “Halo” releases around VNX – Appsync, XtremSF
- Part 3 – ViPR General Availability
- Part 4 – “Project Nile”
During the Speed2Lead launch – while certain things were anticipated, a couple were not. One was that we announced that ViPR will GA before the end of the month. Just as importantly, we announced the pricing.
ViPR is an ambitious purely organic EMC innovation. If you want a refresher, read this post…. Ok, back?
ViPR has two VERY distinct parts – the controller, and ViPR data services. Let’s talk about the controller first.
- The ViPR Controller is an important thing for the storage industry as a whole. There is no other (that I’m aware of) effort to create a “software only something” that does what VMware NSX does for networking:
- Pool physical storage assets (in many forms) into a logical set of capabilities (this is a “virtual array” in ViPR-speak that can then overlay…
- …Abstracted logical “virtual storage pools” that advertise their services, and can be subdivided into logical tenant domains. Certain multi-tenancy ideas have to still be in the data plane in storage (notably IP spaces) – and you can see what we’re doing about that here [and there’s more too – we will keep extending the ideas of VDMs natively which can have distinct routing tables now (fully multitenant), but still lack the ability to have overlapping IP spaces without a NSX-like thing – but we know we could do it]. These abstracted pools can then be….
- … Automated. This last part is key. Open APIs (that we will be publishing that can harmonize northbound integration with things like vCO/vCAC, Openstack, Puppet, Chef, CIAC, whatever). For the whole storage industry potentially (!). Like the northbound APIs, we will be publishing code examples of southbound array integration modules, and will make it CAKE for anyone to integrate.
- BTW – to be very clear and transparent, it’s a VERY, VERY important thing for EMC (as one party in the storage industry). We FIRMLY believe that the reason there is insane architectural diversity in storage land is an intrinsic reflection of workload diversity. Few folks still truly believe that a single architecture can service the market as a whole. This is one of the broad ideas I sketched out in STO5420 at VMworld (you can download it here), and also in this blog post here (that also has a whackload of demos). In a nutshell, it’s this picture:
- The observation of our customers is they dig the choice we give them – delivered in a bunch of ways:
- purely organically (next gen VNX, VMAX, Atmos, XtremSW)
- through organic improvement on IP acquired a while ago (Data Domain, Isilon, VPLEX, Recoverpoint, etc)
- through inorganically recently added IP (XtremIO, ScaleIO)
- … there IS a BUT, however. A customer visiting us at HQ yesterday (thank you!) yesterday told me – “I understand the need for diversity, but there comes a point where the operational complexity outweighs the advantages. I want to have my cake and eat it too!”. While it is true that this is an industry-wide problem – as the broadest player in the industry, this challenge affects EMC more than anyone else.
This whole dialog puts aside the OTHER part of ViPR – which are software-only data services. In the 1.0 release, there will be a robust, exascale, next-generation object data service. It will support S3, Atmos, and Swift APIs. You’ll be able to just “add it” to the storage you want (first release will depend on an underlying persistence layer).
Both the controller and data services are designed for “simple, but cloud scale” – stateless, horizontal scaling, no centralized metadata or services handling. It’s designed by a team that built Azure, and a ton of the advanced code in VMware and other software-only stacks.
Look – it’s a 1.0 product. It’s been deployed at many customers through a “early access program” for months. It will have limitations. But that doesn’t change that…
…This is a BIG idea. Could we bork it? Sure. But – damn the torpedoes, we’re investing to change the storage world here.
Why is this BIG news?
- People thought that when we put the idea out at EMC World in May, it was “vision”, unicorns and rainbows. WE WILL GA in weeks.
- People tried to lump this in as “storage resource management” – it’s not. People will still need tools for reporting, performance, analytics, management (just like they will still be needed with NSX in networking land). It’s a PLATFORM. We announced a go-to-market model that looks like a platform, more than a product. It will have price that is designed to make it a simple no-brainer. $0.01/GB/month type pricing for the controller and for the data plane object service. If you think about the power of pooling, abstracting, and automating all storage – for VMware, for KVM/Xen/Hyper-V and for Physical environments – that’s killer. I can also tell you that we WILL make it broadly available to play, innovate, automate, integrate, learn – i.e. the linux economic model of (“play for free, contact RedHat if you want enterprise support model”s) . I can also tell you that we will publish everything that we possibly can.
THAT is big news. The “NSX for storage” is coming in weeks. It will be super open. It will be nearly free. It will provide big value.
As always – input, constructive critiques, alternate world views welcome!
Comments