Death by 1,000 cuts: Mainstream storage array suppliers are bleeding

Great beasts can be killed by a 1,000 cuts, bleeding to death from the myriad slashes in their bodies – none of which, on their own, is a killer. And this, it seems, is the way things are going for big-brand storage arrays, as upstarts slice away at the market with converged systems, virtual SANs, all-flash kit, hybrid devices, …

[Anonymous Coward]

It depends! =) Keep in mind that revenue and profit is not the same thing, and that it is easier to massively grow your customer base when you're small. However, the new guys do have some cool stuff!

So what do I think? I think the going will get tough in the smb space. When you talk about the enterprise space, it is always nice to have a multi-billion dollar company to choke. On a massive scale, it's also nice to buy from one vendor, to avoid islands of specific purpose built solutions from many vendors.

Even though the individual storage systems are cheap, once you start to add together the operations and support and training for one VDI stack from company 1, an object store from company 2, and enterprise array for enterprise application from company 3, and a AFA for a demanding data base from company 4, the TCO starts to point in favor of the giant, that does _not_ have the best systems, but good enough systems.

Looking forward to more of these articles!

[joejones]

Options are good, but cloud storage has a lot of issues the never seem to come up in these pieces

I agree. Little storage vendors are making it harder for the big ones to make as much money as they used to. For someone like me that works and purchases gear from these vendors, we are spending less on traditional storage than we used to. Excellent stuff.

The cloud is another issue...

There is a similar but less comprehensive and not-as-well reasoned piece on techcrunch today that speaks of cloud destroying the old storage guard, but what people keep seem to missing about cloud storage is how you get to it.

We had a director tell us to look into getting rid of our expensive mainstream backup solution (they hate paying for anything, which I understand) and use this new fangled cloud based backup some sales person called him about, and I obliged.

They sound like they know what they are doing. They support lots of products (but not Oracle, so already its a no). Then we asked about their bandwidth, and they claim to have OODLES of bandwidth. Then we asked the question if they had enough bandwidth to satisfy the demand of say 20% of their customers at once. Let's say the US northeastern seaboard power grid is down, and now we are all scrambling to restore our data at the same time. They didn't have a good answer for that one.

Then we calculated our costs required to upgrade our internet pipes to accommodate this - and the whole thing didn't make any sense, even if they did support oracle. Even with an incredible fast internet connection -it would never come close to a local 10Gb connection - let's not even get into latency over the internet compared to a local network.

And the rah-rah cloud storage piece is written by a VP of a cloud storage company. I guess its really paid editorial advertising, but they never mention that on their site.

My point is that there are all these other costs around cloud storage beyond the $/GB a month, but people just seem to accept that they aren't a big deal until they have to pay for them.

[Don Jefe]

Accounting

The following doesn't apply to very small companies, but becomes a huge issue once revenue is beyond a few million annually and for all publicly traded companies.

Once you're past that point, how you book expenses becomes more important than the cost of those expenses. Cloudy services go in the same column as utility bills and facility cleaning staff. Because of modern bookkeeping practices those things don't represent capex or other NRE's like licensing that you want to avoid on your books.

Super simplified, monies spent to actually buy something affect financials in a bad way, even when amortized across many years. But a whole bunch of small sums, even if the total is larger that an outright purchase of the same thing, has a very small effect on financials.

Accounting Macumba means that beyond a certain point it becomes cheaper to spend more money if you spend it in the right way. So, for small companies the 'hidden' costs of Cloudy things can be a killer, but for larger companies it's 'cheaper' for them to pay those costs than to buy things to do it in house. Ultimately, this all means that 'hidden' cloud costs will continue to rise, even if the 'simple costs' continue to go down.

On a lighter note, what this also means is that big array vendors are going to have to develop affordable, high performance products that target the small operations the Cloud providers are starting to ignore. Which is good news for small companies, but it also means I'm going to be able to put EMC kit in absolutely everything and not much would make me happier than telling my EMC rep that I'm using their once mighty arrays to record the drying of paint and the effects of gravity on large rocks for no reason other than it is cheap (we spend about $7m annual with EMC right now, so they are on my 'list of vendors who owe me' :).

[Anonymous Coward]

Re: Options are good, but cloud storage ...

Alternatively there are object store solutions out there implemented as a "private cloud" that would likely save you a fortune over outsourced cloud and your current back up strategy with all the resiliency benefits of erasure coding and dispersal.

[joejones]

Re: Options are good, but cloud storage ...

We have looked into object storage as well, since we have a large number of files stuck in file shares on filers that are pretty unwieldy - tens of millions of files per volume on several volumes for one app alone.

Vendor lock is always the biggest concern with object storage. EMC's offerings with their RAIN products are notorious. Customers move all of there data to the object storage (and its not cheap), and then realize how awful EMC and want to move it to another product. And how do you do that? There is a company in the Boston area that specializes in migrating people from one object storage appliance to something else. They make a killing because its so frickin' complicated.

Switching to an old guard nas vendor with dedupe and compression has saved us tremendous amounts of space (75% in most cases), and while its no fun to backup, its possible, and if we decide we want to switch to another NAS vendor, we can with tools like Richcopy and some downtime, and not have to rewrite the application because there is a new API.

[Jim O'Reilly]

ODMs will enter the fray, too!

Commoditization is already moving at a fast pace, but it will accelerate as the ODMs who service the big CSPs start selling branded product and whiteboxes in the US. The barrier to entry on storage is reaching the Linux level, and we all know what happened to mainframes!

[joejones]

Re: ODMs will enter the fray, too!

Mainframes still run the majority of financial business in the world. They really didn't go anywhere, they just aren't the only option anymore. They aren't cool, but they are still used pretty widely. There is a real shortage of programmers for them.

[Philip Lewis]

Re: ODMs will enter the fray, too!

Now, where did I put my BAL manual?

[Mikel]

Bring on the Storinator!

Seriously - Google it. 180TB raw in 4U, good perf, price is under $14k with drives. Add some PCIe attached SSD and some 10Gbps ports, some virtual SAN software and it's off to the races.

[joejones]

Re: Bring on the Storinator!

The Storinator looks very cool. Still need to buy a bunch of disks for it and have some level of 24 x 7 hardware support.

What virtual SAN (I hate how the world has changed the meaning of the acronym SAN) software would you suggest placing on something like this?

One would assume you would need to buy a few of these storinators with redundant storage on both so there is no downtime, and the virtual SAN (sideshow bob grimacing and moaning) would need to take care of that redundancy.

[Mikel]

Re: Bring on the Storinator!

It is ~$7k sans drives. The drives are the other half of the $14K. Some assembly required - which is how enterprise SANs come normally anyway.

All the VSAN virtual appliances offer network striping and redundancy features. That is kind of the point after all. A 3-node site-redundant SAN can be had for under $50K.

HP's VSA will work with this, as will various open source solutions, which have specialist consultants offering support levels all the way up to "permanent onsite engineer".

[dsfgdsd]

Joejones says there is problem with storage in the cloud, because it would require a really fast low latency 20% full bandwidth access.

Indeed that is true if you have the data on premises, but why wouldn't you move everything into the cloud? If done that way, you'd only need enough bandwidth to allow for some human viewable terminal output and very low bandwith keyboard/mouse input.

[nullgenome]

Cloud and Object

@dsfgdsd Control unless you've created the cloud yourself? Plus what do you do if the cloud company goes bust, etc etc. I still don't trust it fully but it has it's place for me, i.e. personal crash plan.

@joejones regarding object storage I know nothing of EMC's offering but with the two other vendors we tried (we're now using one) getting data in and out is pretty easy using standard HTTP GET and PUT commands. Admittedly you have to build your own system for keeping track of file to object ID. But I guess if you need to look at object storage then you'd probably have the skills in house anyway.

[joejones]

I guess my comment wasn't clear enough.

I understand that cloud storage is a good tool for some companies to look at. If I had an SMB shop, cloud backups would be a superb solution for offsite storage and DR purposes.

The cloud industry (and its salespeople) have been trying to sell the notion of cloud to anyone for everything. My point is this is often not the case.

We had a cloud backup company trying to convince us (a nontechnical director at my company who just sees cloud = job promotion and relevance) to look at their product. This company has their own datacenter with storage to hold the backups, and customers connect to this from the internet to store and retrieve their backup data.

My point: Let's say the northeast region of the US was down due to a power grid problem. This has happened, and is likely to happen again.

If all of their customers affected by this outage started restoring their data to their DR site in Nebraska or wherever, would they have enough bandwidth to accommodate it without all of their customers complaining that the restores are too slow? The people I spoke with never thought about that, or didn't want to discuss it. I think that's a pretty substantial issue, and a good reason not to use their service.

Our company has contacts with customers that pay us based on availability. I think anyone else in this situation needs to look long and hard at cloud solutions before jumping into one - especially if they already have the systems in place that meet their needs.

[random_graph]

It's those dang VmWare guys

Diversity in networked storage solutions was always a function of workload diversity. But with virtualization reaching 100%, the Hypervisor is the only client that midmarket storage vendors need to design for these days. Of course there are still differing performance profiles for different guest workloads, but the point is that the host environment (networking, provisioning, protocols...) has gotten much simpler. So these 3 things represent the bulk of all cannibalization of the legacy array market:

1) Practical adoption costs for using public cloud are much reasonable once you're virtualized

2) Startups can optimize strictly for the virtualized use-case (thus creating differentiation)

3) The hypervisors can handily disintermediate the value chain (EVO-Rail!) and put everyone else out of business (or force them into OEM servitude)

So what's left for the storage vendors? Probably just the vertical workloads (Web2, M&E, HCLS, GIS, analytics, PACS, etc etc).