Stratoscale – A true Hyper-converged solution similar to VMware Evo:RAIL?

I’ve been exploring the partner solution offerings during Cisco Live 2015 in Milan, and came across this stand from Stratoscale which I found quite interesting. So I had a chat with the guys and thought it would be worth  mentioning a little bit about what I learnt of them.

Stratoscale is a small Israel IT start up (offices in Israel, USA and expanding out globally), with funding from Cisco investments (understandably) that provides a true, large scale hyper converged infrastructure software that can run on BYOH (Bring your Own Hardware). I’ve had a lengthy (technical) chat with one of the founders of the company,  Etay Bognar (etay@stratoscale.com) and they seems to have a very interesting proposition at hand. Unlike other popular converged infrastructures vendors such as Nutanix, Simplivity….etc. Stratoscale seem to offer just the software elements that can work with commodity hardware from any server vendor. But the key here is that they actually own and provide all the software elements including the hypervisor which is a customised version of KVM which is optimized to work based on their hyper converged offering. For example, as a part of this customisation, they claimed that they’ve completely re-written the memory management module within KVM to optimise it and as a result of such optimisations, that they could migrate a running VM from one host to another within a second, regardless of the size of the VM which sounds very interesting (though I haven’t seen it in action myself, but kind of make sense as they claim to use post copy operation unlike VMware VMotion that uses pre-copy).

As a part of the software solution offering, they are also providing a fully customiseable management platform (similar to VMware vCenter in an ESXi cluster), something that seems to be lacking in a generic KVM based hypervisor cluster deployment without heavy scripting…etc. This management platform apparently also offer intelligent algorythms managing and maintaining VM placements within the cluster (similar to VMware DRS). According to what I was told, they key here is that the level of control & visibility they have within the compute and storage layer is very granular & limitless, as they own the hypervisor too, much like VMware in their Evo:RAIL offering.  This is very different to, say a Simplivity or a Nutanix solution that don’t have their own hypervisor, therefor having to rely on vCenter / ESXi API to interact with compute platform and therefore be restricted to what you can see and do to the contents of the API. To me it make sense that Nutanix and Simplivity are more like aggregators rather than true Hyper-Converged solutions where as VMware Evo:RAIL is probably the only other true Hyper Converged solution that truly converge everything.

So these guys at Stratoscale looks like a direct competitor to VMware’s Evo:RAIL solution going forward from what I could see as they too have true Hyper-Convergence in that they own the whole software stack in their offering and all aspects of the solution, compute and storage are all baked in directly to the hypervisor, much like VMware Evo:RAIL. Their solution at present appear to scale out to 64 nodes at present with distributed storage in all nodes (SSD backed, similar to VMware Evo:RAIL) and to be honest I quite liked their pitch. It sounds like a very powerful offering which could potentially be a very good solution for some of those non VMware customers, that are more Linux orientated that do not want to spend too much money for a VMware solution (I do not know how expensive it is compared to a VMware solution but I was assured that its a lot cheaper in comparison). Being a start up, there’s no doubt that things would need improving, and its probably no match right now, for the almost fully complete VMware offering with all its eco system, but it would be interesting to see how they fare, especially given the fact that owning their own hypervisor giving them somewhat similar capability to VMware to produce an optimised Hyper-Converged solution. If they do well, I can see naturally, they could end up being a very good acquisition target for the likes of Cisco, Intel who may want to strategically enter the Hypervisor play (current Cisco funding is pretty self explanatory here I think). Worth keeping a close eye out, I know I will.

Comments are welcome, especially from existing Hyper-Converged solutions users

Cheers

Chan