Working within R&D at VMware means you typically work with technology which is 1 – 2 years out, and discuss futures of products which are 2-3 years. Especially in the storage space a lot has changed. Not just innovations within the hypervisor by VMware like Storage DRS, Storage IO Control, VMFS-5, VM Storage Policies (SPBM), vSphere Flash Read Cache, Virtual SAN etc. But also by partners who do software based solutions like PernixData (FVP), Atlantis (ILIO) and SANDisk FlashSoft. Of course there is the whole Server SAN / Hyper-converged movement with Nutanix, Scale-IO, Pivot3, SimpliVity and others. Then there is the whole slew of new storage systems some which are scale out and all-flash, others which focus more on simplicity, here we are talking about Nimble, Tintri, Pure Storage, Xtreme-IO, Coho Data, Solid Fire and many many more.
Looking at it from my perspective, I would say there are multiple phases when it comes to the SDS journey:
- Phase 0 – Legacy storage with NFS / VMFS
- Phase 1 – Legacy storage with NFS / VMFS + Storage IO Control and Storage DRS
- Phase 2 – Hybrid solutions (Legacy storage + acceleration solutions or hybrid storage)
- Phase 3 – Object granular policy driven (scale out) storage
<edit>
Maybe I should have abstracted a bit more:
- Phase 0 – Legacy storage
- Phase 1 – Legacy storage + basic hypervisor extensions
- Phase 2 – Hybrid solutions with hypervisor extensions
- Phase 3 – Fully hypervisor / OS integrated storage stack
</edit>
I have written about Software Defined Storage multiple times in the last couple of years, have worked with various solutions which are considered to be “Software Defined Storage”. I have a certain view of what the world looks like. However, when I talk to some of our customers reality is different, some seem very happy with what they have in Phase 0. Although all of the above is the way of the future, and for some may be reality today, I do realise that Phase 1, 2 and 3 may be far away for many. I would like to invite all of you to share:
- Which phase you are in, and where you would like to go to?
- What you are struggling with most today that is driving you to look at new solutions?
Because the hardware cycle is coming to a close in the next year and we have plenty of space left we have gone full circle. First we were looking at replacing our entire storage infrastructure (or migrating to tiered storage solutions and reallocating existing storage for non-critical use). We interviewed many vendors, sat through umpteen solution briefs and technical deep-dives, and came up with a recommended solution. Then the latest European crisis occurred affecting our company’s financials and we were tasked with holding steady on current inventory or significantly altering our strategy.
About that time we learned about software solutions that accelerated our VMs without replacing storage, allowing us to use the remaining available disk space for the next year or two. We are currently demoing those products and are anticipating a successful experience showing the value. The cost is a fraction of replacing all storage and could potentially even direct where we go when we actually have to replace the storage due to end-of-life products and support contracts.
So we are probably in phase 3 but we skipped phase 2 because we never considered using Storage IO control to be a viable solution given how many systems are sharing a single cluster of disks. Unless phase 4 were to somehow show us how to improve performance even more than caching technologies promise to provide I don’t see how we’d have the need for Storage IO control or even phase 4. But technology changes so never say never.
I am in Phase 1 with plans to move to Phase 2 or 3 for DMZ, VDI and Management Clusters. Business challenge is to complete the ROI for Phase 1 infrastructure, which is why server-side flash acceleration is quite cool to extract a few more years out of my monolithic investment. Article I wrote recently on same subject: http://vcdx133.com/2014/05/13/evolution-of-storage-what-is-your-strategy/
We are currently in Phase 2. Our two major sites each have its own storage array dedicated to VMware. We are using SDRS and Storage IO control to manage the VM’s.
We would like to move to Phase 3 and look at tiering solutions. Currently we are managing which VMDK can be placed on slow disks and which can’t. Would be awesome to automate this more to push more performance out of our arrays.
Phase 2, but w/o Storage DRS simply because of NFS datastore based snapshots. Phase 3, we’re starting w/ vSAN for our management cluster, but waiting for more vSAN passthrough compliant storage controllers. Mainly due to the operational pain associated w/ a failed disk.
I don’t think I agree with your assumption in these phases that one is implicitly better than the previous. By your definitions, we’re living at Phase 0. While I find Phase 3 very interesting (part of what keeps me coming back to your blog), I don’t see it as inherently superior. Storage has too many variables to make such a broad statement. While Phase 3 might some day serve my company’s needs, it will only do so if it delivers the performance at the right price. Right now, our SAN is pretty overbuilt, so Phase 3 is really just interesting at this point.
I understand that, and of course I am greatly over simplifying the world here and I also do understand that the next phase also does not necessarily for all folks will mean things are better. They are different that much is certain, but if they are better very much will depend on what you are running etc.
Thanks for your comment, much appreciated.
You have to remember also not everyone has Enterprise Plus on there systems and may never get Enterprise plus. So some of those Enterprise Plus features at some point need to move down the Enterprise to Standard licenses before you may see wider adoption. At least with the older technologies that have been around for a while only as apart of the Enterprise Plus license. aka vDS, or Storage DRS etc would be nice to have roll down finally.
Having said that, I find that my company is at a cross-road as we are looking at different storage possibly and looking at vSAN potentially to be our next storage platform for our vm’s in replace of a traditional San. I think the big thing I heard from you is make sure the SCSI card is on the HCL. Another question is what’s the best way to migrate all the VMs to a vSAN? can you put an old server in a new vSAN cluster so that it can see both Datastores and then migrate the VM’s that way? Hate to shut them all down to do a cold migrate.
I think you are a bit on the right track on progression of different storage technologies. What comes to my mind from what I have seen recently are :
1) old san way of doing things,
2) hybrid – san + a local SSD/Flash disk for read/write cache for the SAN
3) all local flash/ssd/sas|sata drives + scale out hosts for replicas and redundancy like vSAN or PernixData
Having never had the licenses to do #2 on your list so I skipped that one. By legacy I assume you mean SANs. By Object driven & policy storage is a bit abstract for me so not sure if you are talking local scale out storage or also meaning that could be more SAN based but has more capability to designate what policy aka tier of disks you need to run VM’s for performance / certain number of iops etc.
My list above is more just storage considerations but I can see more of the future having compute and storage being 1 sku like a Nutanix. Thinking that someone out there could come up with solution like Oracle provides with Exadata but based more on VMware / vCloud / vSAN priced per VM per month and one that can be rolled into a on-prem datacenter. And can sale from one to thousands of VM’s if needed. Just roll in another rack for another set of servers. I think that can be done but haven’t seen it sold that way for an on-prem cloud solution yet.
I know not everyone runs E+, and I can see many folks skipping that whole phase… but for many out there it was part of the evolution hence I added it.
Number 3 would include solutions like: VSAN, VVOL capable arrays, Nutanix etc.
Thanks for your comments, and I agree with a lot you have said.
Hey Duncan. We are currently on phase 0 and working/testing phase 2. I asked around and I am surprised that a lot of folks are on phase 0. Basically, it all boils down to cost. However, I cannot wait to do POC on VSAN.
Thanks, I was just reading your post and links about VVOL, that’s very interesting to do it on a per VMDK level for sure.
Honestly I don’t see the connections between the different features. Some are different by the distributed architecture (scale-up vs scale-out), then by media type (hybrid, so the others should be all disks and all flash, even if not listed) and then by data services (you call it policy driven, others say QoS, and Storage DRS is a feature of the hypervisor, not of the storage).
Let me be clear, I love scale-out design for example, but you can have QoS/VVOLS/policies even on a scale-up all-disks architecture, or a hybrid one. vvols+policies over a Netapp nfs storage is a 0+3 phase? Not clear…
Nah, I should have added brackets to the phase 3. And it is just the way I see things evolving. You could look at it like that:
legacy storage –> legacy storage + basic hypervisor extentions –> hybrid solutions with hypervisor extensions –> fully hypervisor / OS integrated storage stack
May even be a better way of putting it 😀
I like that a bit better.
Oh, now I like it 🙂
Yep, the evolution of data services integration between the hypervisor and the underlying storage is indeed going into changes. And as always, the more integration, the better (with a bit of unavoidable lock-in…)
Left my old company two years ago, where we had moved through the phase from 0 with VirtualCenter 2.5 through phase 1 and ended up in phase 2 around 3 years, ago. For us it was a smart moving with the technology, by no means were we first movers, but every phase had it pros.
Phase 0 – Made virtualisation possible
Phase 1 – Simplified our way of handling and thinking of storage
Phase 2 – Was simple for us, as it came with a loss of support on our currently and the price of extending support, payed for the new array and it gave us better performance (flash/ssd cache and tiering).
In my current company we haven’t moved that fare, just deployed an array which does tiering. I think a lot how we move through phase has to do with how companies are structured. In the first company I was VMware and the storage guy – The second one we have a dedicated storage team (aswell as network), so it take a lot more work to influence the decision (If you even know what coming).
So currently we are some were between phase 1 and 2, with a way before we get to phase 3.
I’m not sure phase 2 makes a lot of sense to the enterprise – at least not the way it being sold. It made a lot of sense in the first company example because we only had 300 VM or so – The overhead of managing storage was some what highere than in an enterprise.
But what we want and need is consistent storage performance, what tier storage brings is even less consistency – Let me give an example of a customer meeting I was in:
The customer say; We are not sure the SAN can keep up, after initial testing, we believe we will hit the storage 10 times harder when we move to production, and it seems to be struggling a bit.
Storage guy says; Don’t worry, we have auto tier, if you just keep at it, it will be moved to SSD.
So no consistency and predictability seems to be what the enterprise is getting – Some of this is of course “cured” with large caches.
I think what is needed is Phase 3, on AFA AND an archive tier or vSAN/Nutanix where storage moves as close the CPU and memory – in the same way we still have types around.