Storage

Recently Coraid have been trying to make noise again about their storage arrays and ATA Over Ethernet. I’ve always dismissed it but never tried AoE  so I thought it was about time that I at least gave it a try before dismissing it.

I quickly brought up another Linux server on one of my lab boxes and loaded in the AoE modules; grabbed the vblade package (vblade is the package which allows you to serve AoE targets from a Linux server), a quick peruse of the how-to and set up the server instance to serve an AoE target. I must admit that it was ridiculously simple, a lot more simple than configuring iSCSI targets.

I then downloaded the Starwind software initiator for AoE for Windows, installed it, fired it up; it automatically  found the target that I had already configured on the Linux server and I added that in.

Job done; from start to finish, excluding the time to install a new Linux server instance; it probably took me fifteen to twenty minutes. Not bad considering I’d never looked at AoE before.

So if it is this simple, why are AoE and Coraid not making a big splash?

It’s certainly simpler than iSCSI and I get about 20-30% better performance than iSCSI from my initial tests from the same hardware.

It’s quite simple really; it’s just too late, in the same way Video 2000 was too late to make an impact in the video market despite being a superior format,  AoE and Co-Raid just don’t have the momentum and I just don’t see them ever getting it. Their best chance is to get some of the smaller home/SMB network storage vendors to pick it up and try to build ground-swell there but it’s going to be tough.

But if you have an hour to kill, you might as well give it a go…

Commodity hardware, commodity disk, commodity platforms are enabling interesting developments in the world of storage infrastructure; especially in what I like to think of as Boutique storage which is probably a posh and pretentious way of saying Niche!

Object Matrix with their MatrixStore product is one such development; aimed squarely at the video market with post houses, in-house video production teams and anyone who wants store and archive video assets in cost-effective and efficient manner; the team has developed an object store which has quietly been becoming a bit of a hit in the UK and is now spreading far from its roots in Wales.

Chatting with Nick Pearce, one of the co-founders, you find that he has a background in object storage spreading back into to the early days of Centera being responsible for…well, I’ll let him admit to his folly but suffice to say, he is a bit of veteran with all the scars required to mean that he knows of what he speaks and sells.

But most importantly, he knows that they don’t need to spend a fortune on hardware RnD and that they can concentrate on building a robust software product which runs on top of an open source operating system. They have decided to still ship with a hardware platform, Nick explained that this is to minimise the sheer amount of testing and verification that they have to do which allows them to keep support and development costs down. However it is still built on industry standard components which allows them to leverage performance and capability quickly and easily.

We can expect improvements in network capability, although it is not really a bottle neck for them at the moment; these are NL and archive solutions and not designed to support live streaming and editing.

We can also expect them to leverage the newer large capacity hard drives in the near future; further enhancing the value proposition; talk to Object Matrix if you want their costs but even list prices are actually very attractive and took me aback. These guys are not about building out of commodity components and sticking a crazy premium on top like some of the other broadcast specialist companies.

Supporting compliance, distributed search, off-site replication, application integration and an exposed API; MatrixStore might actually be the product that many post-production teams are looking for and the guys actually talk your language. They use the same jargon, they know the same people and they know the industry dirt…and they won’t turn up in suits looking like they are here to collect the rent. They have the local knowledge which is very important in this space but they also know the IT side; they know that Broadcast IT is more about ethernet cables than SDi cables these days but they do know what an SDi cable is.

More importantly, Object Matrix are no longer what I would class as a start-up and they dropped some seriously large names in conversation; they are quietly building themselves a market position. No, you won’t come across them in your IT department but if you look beyond, you might find that your in-house video production team has one or two stuck in cupboard pretending to be something innocuous.

And although they are rather cautious about moving beyond the video and broadcast arena, it might be worth having a look at them if you have an object storage requirement…

They also have some interesting ideas as to what to do with all these extra CPU cores; things which address some of the potential data-refresh and format challenges going forward.

[Disclaimer: Nick Pearce bought me a couple of drinks and ham, egg and chips but if you think that’d sway my opinion, I’ve got a bridge in the centre of London to sell you.]

After his no more tiers gaffe and I will still maintain it was a bit of a gaffe; Georgens has redeemed himself in my eyes with his statements on the IBM relationship, smart guy seems to really understand the OEM dynamic.

Lifted from Chris’ excellent article at the Register

“The desire by their [IBMs] internal groups to develop their own products makes the positioning very, very complicated. And are we happy with the positioning? No. On the other hand, our engagement with IBM’s customer facing groups, the people who actually have to put solutions in front of customers, that relationship is actually exceptionally strong. So I think that an approach to storage from a pure platform perspective and basically creating SAN products and NAS products and unified products in a very, very hardware point of view is interesting, but it’s just recreating the fractured product line that’s given us an opportunity to gain share … IBM has introduced products that are competitive with both Engenio and with the NetApp offerings over the years, yet this business still continues to grow, it still continues to be robust.

I think that [IBM] internal groups are looking to compete and develop competitive products and if they truly are competitive and they truly can compete with our feature set both from a hardware and a software perspective, clearly demand will shift in that direction. But if we continue to out-innovate them and have a higher development gains and introduce products to market faster, then we’ll preserve the business. It’s no different. It all comes down to innovation and execution excellence. And it’s been that way for the last five years and that’s the nature of the OEM business.”

It really represents the dynamic that I see being played out as an IBM customer; deep down IBM don’t really want to sell the NetApp stuff but whilst the NetApp stuff is better than theirs, they’ll take the margin and revenue. But there’s no illusion, if SONAS and v7000 can really gain a foothold; IBM will be off.

As a customer, it’s really interesting seeing this play out and if at the end of the day, IBM end up with viable, thriving and credible storage story of their own; they are going to owe NetApp big time for giving them the space and the time to do so. But Georgens’ confidence in his own product and its ability to keep its place in the IBM portfolio speaks volumes.

Now Val, about this roadmap you keep promising?

When I started this blog what seems years ago; people got used to me railing against the state of storage management tools and especially SRM tools.

Since then I’ve changed roles and moved away from the more traditional corporate back-office storage environment and I really do not have the same level of exposure to the SRM tools and we have a much more simple infrastructure.

But this is beginning to change, our environment is growing incredibly quickly as we store more and more content and as we start to roll out more traditional IT storage to support our Creative teams; with over 20 arrays, multiple clustered NAS environments, multiple tape-libaries and back/archive environments, I am starting to look for a tool to manage and report on this heterogeneous infrastructure.

So I am starting to have another look at the state of the art with regards to Storage Management tools again….And sadly I am finding that the storage of the art is still woeful; the tools appear to have moved on not one iota.

I still find talking to peers and colleagues that spreadsheets are almost universally used; the vendors have still not yet delivered the improved tools that they were promising 2-3 years ago. I still hear ECC derided on a regular basis; I hear people expressing disappointment that NetApp’s stewardship of SanScreen has not delivered and IBM’s TPC is still lacking in features.

It does seem to be pretty much beyond the storage industry to deliver a tool to manage storage in an heterogeneous and scalable fashion. It’s obviously just not sexy and we should all move to integrated stacks or perhaps the Cloud; oh well, back to Excel!