'Zilla talks about various things here but what caught my eye is his commentary on the Exchange DAS vs Storage Array conversation.
It's an interesting one; he feels that Microsoft are trying to maintain their margins putting more and more traditional storage functionality into their software and then suggesting that the end-user saves money by putting it on DAS. DAS being cheaper, this allows the TCO for Exchange to fall.
As end-user; as long as my service is maintained and perhaps even improved, I don't care. This is not a religious debate of SAN versus NAS versus DAS, this is more fundamental than that. This is about saving money. This approach could save me both capex and opex; it needs careful modelling but more and more traditional storage functionality is going to end up in applications.
We are going to see an almighty tussle between the storage vendors and the application vendors; EMC find themselves in an interesting position in this tussle, how much traditional storage functionality could find it's way into the hypervisor?
High-speed interconnects such as 10 GbE and Infiniband don't just speed up server-storage connectivity; they speed up server-server connectivity as well enabling fast block-level replication at the hypervisor layer in a way which is transparent to the operating-systems and the storage (could be DAS or SAN).
So who'd be an array-vendor? Margins are going to be squeezed and differentiators are going to need to be found.
I’ve seen very tight budgets swing folks to DAS. That’s a shame. DAS has a number of downsides, to suggest a few:
– poor utilization
– poor management
– poor DR
– poor VMWare fit
Thought about fleshing those out, we know the issues. A LeftHand seems to be a blend. Take your local storage and attack the downsides mentioned above. We’ll see how much traction they get – I’m not excited. Think about why SAN/NAS took off in the first place, wasn’t as if folks were bored with DAS (solved the top 3 downsides mentioned above, VMWare came later). I believe we are in for another wave of storage chaos before it sorts itself out. Vendors pulling in storage functionality (Exchange, Oracle ASM, VMWare) and leaving others to fend for themselves. Had a client meeting and the Oracle folks are doing their thing with ASM and they have budget. They are causing “issues” because they are steering an upgrade in a bad direction because they have their onw little storage playground. I pointed out the problem they are causing much to the delight of the others in the meeting. Finally, because of tight budgets and strange departmental-think versus enterprise-think [1], we’re going to see more storage segmentation. Maybe more services for cleanup, hmmmm… maybe not a bad thing!
[1] We’ve been at this a while people, the money comes from the same place, why run off and do your own thing?? Get a clue!
//We are going to see an almighty tussle between the storage vendors and the application vendors//
You are not alone in that prediction:
http://searchstorage.techtarget.com/news/article/0,289142,sid5_gci1332234,00.html
http://searchstorage.techtarget.com/generic/0,295582,sid5_gci1341963,00.html
Well firstly, blimey..Beth Pariseau commenting on my blog!
But secondly unholy alliances between application and hardware vendors leave me feeling that I will continue to pay a premium for my storage, just not to the traditional storage vendor.
And to take Rob’s comments
1) Poor utilisation – arguably we have that with SAN and NAS anyway. We just pay more for the wasted disk.
2) Poor management – in what world do we have good storage management? Most of the storage vendors are in tacit agreement on that.
3) Poor DR – not if you build replication into the application. In theory it could be better; applications understand their transaction models alot better than a bunch of spinning disks.
4) poor vmware fit – it could be a fantastic vmware fit. vmware could abstract the storage, manage/implement the replication/snaps etc.
Networked Storage will not die but it should not be the default answer. Your data could be on DAS, NAS, SAN or in the cloud.
It’s quite exciting, some thought will have to go into storage design. It’s not just spinning rust storing 1s and 0s; it’s the cornerstone of building the information mangement of the future.
re the 4 points listed above :-
1) Poor utilisation – who cares? Its the cost of the utilisation that’s the key thing… The capex & opex cost of DAS ‘can’ be an order of magnitude cheaper than SAN/NAS etc so the ‘utilisation’ can afford to be half and still be less cost.
2) Poor Mngt – agree, but until we have functioning NAS/SAN estate SRM at a viable TCO this get’s a nod but no more.
3) Poor DR – information needs to be replicated not data. Handle the information instances at the app/middleware/db info layer not the disk data layer.
4) Poor VMWare fit – how? Look at 4 and see what interesting stuff is being done.
Some very interesting changes in the application layer are definietly re-validating the DAS approach – as your apps and information go horizontal / sharded then so can your infrastructure 🙂 (heck the fastest DWH infra around has just gone DAS…)
Don’t get wrong – DAS must not be your default answer for business data (in fact it’s the last answer on my decision tree), but is rapidly becoming a very powerful option in the storage TCO/TLC game when combined with the correct app architecture. (in the same way object stores will rule in the next storage wave)
Re vendors eating eat other’s offerings? Agree it’s complex today and getting more overlapping, but at the same time simpler in some areas. I totally believe that the more functionality that moves from the infrastructure data layer into the application information layer the better, so then the ‘commodity storage’ infrastructure manufacturers really will have to start delivering real proven value… Always will be a place for big iron arrays but market will be more niche – thinking server market with ‘commodity horizontal x86’ Vs ‘scale-up Superdome/E25’ etc, value in the horizontal market is on service, predictability, quality and manageability not made-up feature widgets…
Good points all. We probably won’t agree and talk past each other. I had a long-winded reply and ditched it. Maybe another long-winded go..
ianhf touches on why DAS is making a comeback, even though poor utilization (and is it bad if it costs much less? Sure. More to keep track of more to break, encourages spatial abuse and then more to backup, etc.) it is growing again and ianhf hits the nail as apps can do their own HA. I think it only takes a few production outages to rain storage back in again – I’ve seen that (swing back to Enterprise).
Regarding Beth’s links:
http://searchstorage.techtarget.com/generic/0,295582,sid5_gci1341963,00.html
“Microsoft Exchange 2007 is the application leading the direct-attached storage (DAS) revival, which began earlier this year after Microsoft recommended that users, especially those with large mailboxes, use DAS, rather than networked storage to simplify management. A key feature, Cluster Continuous Replication (CCR), has been added to that version of the email application that allows Exchange to handle high-availability clustering so server nodes can have the same redundancy as SAN storage.
Matt Lavallee, director of technology for MLS Property Information Network, followed Microsoft’s advice. He put Microsoft’s SQL Server and Exchange on a Hewlett-Packard DL785 server directly attached to external Modular Smart Arrays (MSA). ”
But is it DAS? Doesn’t have to be, you can hang those MSA’s off switches and they run just as fast. Now it’s a SAN. It’s a SAN anyhow, isn’t it?
“Migrate DAS data to new MSA Connected SAN Array”
At least HP refers to it as a SAN array. The point is BUs within companys get budget, buy these low-end arrays and build their own redundant applications. Fine you say? What a waste when they could be purchasing storage in their Enterprise arrays that are already on the floor. But what it gains them is:
– Their own storage
– No sharing
– No interacting with that painful storage admin
Wins in the short run in
that department.
But I left a couple MSAs behind at a client site and the *only* reason they were there was OS support. When they break, someone will be scrambling because I’m sure no one knows a thing about them.
But why DAS?
– Mom and Pop shops
– No budget
– Very narrow support reasons
Mid-commercial and larger are setting themselves up for support nightmares if they keep letting a bunch of DAS (local to server RAID arrays, or low-end arrays) come in the door. My opinion of course.
Nobodys mentioned the all important lack of backup benefits on DAS….
The hit to storage vendors bottom line will most likely be lack of investment from the SMB sector who can “make do” with less RTO/RPO in 2009 with more plain jane DAS and associated bottlenecks.
This could be a chance for the likes of Lefthand (good buy HP) and dare I say it SUN with there unified array.