The days of expensive enterprise arrays within smaller businesses is limited (possibly). Traditional storage vendors like HP, EMC, NetApp, etc., have provided some awesome products in this area, and the competition between them has evolved their solutions even further (dedupe, snap mirrors, geo clusters, flash cache, blah lab).
But suddenly the world of software defined storage (SDS) is taking markets share from these vendors – or threatening to. SMBs are not often awash with capital, and their accountants (normally running the business) are not keen on just letting IT “get on with it”. They want to know where their money is going, and are not sure why they have spend £500k on a new multi site storage array that they “only put in 7 years ago”.
Often I find a lack of foresight in companies where no provision is made on an annual basis to accrue the funds they will no doubt need to replace ANY IT service. So these infrequent large capex requests are not welcome when they do come. What’s my point? Well, I also don’t like traditional enterprise array series of events where we as IT feel compelled to hugely over provision storage requirements for fear of getting any scale up rejected 24 months down the line.
So, what does SDS give us? Well, first it is an easier solution to scale out IMO. Nice linear server by server scaling for compute and storage… Like it. More predictable and palatable capex requirements… Like it. Grow-as-we-need services… Like it. Tigghter integration with my Hypervisor… Like it. Lower skills sets required (remember not all SMBs can afford two dedicated storage gurus!)… Like it.
I’ve been looking at HP’s StoreVirtual VSA and VMWare’s VSAN (version 6 just released), and in principle I like them both. HP’s solution is “easier” from the perspective that there’s no hardware compatibility list (HCL), and it has nice features like adaptive optimisation for auto tiering hot blocks. But this model seems to me to have a limitation for random IO payloads …. It’s too late tiering my random IO after I needed it. However, with VSAN there’s a horribly restrictive HCL and whilst I understand the reasons for this (intense cross vendor hardware checks on disk and controller capabilities like queue length to check their suitability on our behalf) I feel potentially backed into a corner as an architect due to the current lack of supported hardware. For instance right now as I speak there’s NO supported disk controller for HP’s Gen 9 Proliant DL 380. Hmmm. That might change, but trying to covince your CIO to buy into something that might not support our hardware in the next generation is not an easy sell. I don’t like EVO:RAIL for our size environment either, and VSAN Ready Nodes are not available in the UK though good to use as a template build.
But the VM centric structure of VSAN does give it powerful advantages over HP VSA… I love that full integration breaking down the barriers between storage and management. VMWare also have VVOLS now too for extension of that Vm centric model into the traditional storage array.
Anyway, I don’t know where we’ll end up, but SDS has a powerful place in SMBs and I see it taking market share increasingly. Take a look yourself.