ORIGINALLY POSTED 20th August 2007
6,891 views on developerworks
Over at The SAN Technologist Steven is making some wild and mostly inaccurate statements, particularly about SVC, as well as some curious associations between startup company failures and storage virtualization.
He asks :
So why has storage virtualization failed?
I wasn’t aware it had, and with SVC exceeding targets quarter on quarter it seems nobody has has told us – nor our customers.
Steven goes on to say :
IBM gives SVC away in hopes to get customer traction, however, most deployments of SVC never have been used in production and are quickly getting dusty.
While I understand blogs are about personal opinions, usually these are expected to be backed up by some kind of evidence or at least reasoning. I am not aware of SVC nodes having been “given away” It is true that SVC is often included in deals with storage, but thats because of customer buying behavour. Virtualization is often bought at the same time as storage for several reasons. One, that’s when customers are thinking about updating their SAN infrastructure. Two, virtualization is a great way to introduce new storage purchases, especially when it can remove vendor lock-in. However, just because SVC is in the same deal as the storage, and is relatively low cost by comparison, this does not mean it is being given away. I would also question that there are SVC nodes sat gathering dust, other than the dust in customers datacenter getting sucked through the box by the fans. Few customers would pay for something and then leave it on the shelf to gather dust. That would simply be foolish.
SVC currently has more than 3300 active clusters, IBM has deployed over 9000 individual nodes. This includes over 100 reference customers crossing all walks of life.
As with any new ‘invasive’ technology that radically changes the way you think about SANs, disk provisioning and control, it take time for such technologies to sink down from the early adapoters of new technology to the mainstream user. I can assure you that many customers I have spoken to are very happy with the performance, reliability and money savings they are getting from SVC and storage virtualization.
Steven’s comments about Brocade and McData failing, and EMC are true, but then as I have discussed at length in my Storage Virtualization part work, the only proven approach to deploy true network based Virtualization that provides all the benefits and promises is with an in-band appliance model – and to a lesser degree a storage controller based model (though this does still have drawbacks and limitiations – certainly in the case of USP and its clones). My take on EMC and Invista for example is that they have to have something in their portfolio so they can talk to their customers that ask about virtualization, but its certainly not being pushed as a key product – probably as Steven says because it detracts from the value of their big iron, and would eat into their profits – IMHO.
As for the comparisons being made with the three products / companies cited. They fit into two categories.
- Those that have used commodity hardware AND operating systems to try and produce a virtualizaton appliance. In a world of speeds and feeds a basic OS is not setup to be the center of the SAN. I would certainly not think about using Windows at the heart of my SAN – and that may well be the major contributer to their lack of impact in the enterprise / mid-range.
- Those that chose to create custom hardware. As I discussed a few days ago using custom hardware is great from a performance point of view at the time of its architecture and design – you may be able to look maybe six months into the future and plan for whats coming, but at some point you have to commit to ASIC designs, hardware and bus / memory speeds. Unless you are really slick, or short-cut the testing you will be lucky if you product makes it out the door 18 months later. For a startup that’s a long time. If you did cut corners, your product may get a bad name in the field and bang goes your chance of v2.
Again this simply re-iterates the reason we chose the SVC design point of pre-tested, high performing Intel servers, running as little Basic Operating System (BOS) as possible. Is it any co-incidence that SVC is the only of the cited products growing in revenue and market acceptance – maybe FalconStor, Datacore and co will be along soon to make similar ‘adjustments’ to Steven’s thinking.
There are many reasons for startups in the IT industry hitting a brick wall, or failing completely. I certainly wouldn’t associate failure of startup companies as Steven has done, its quite a big leap to get from “why are startups failing” and end up saying “Virtualization has failed” !!!! Storage virtualization is a strategy still in its relative infancy (as far as SAN based virtualization is concerned) and we have come a long way. Certainly at IBM and at our customers, storage virtualization is a key player in our overall storage strategy and its growth continues to accelerate. Oh and SVC provides ‘PPRC’ functions too, so you can indeed expect to see lots of data still flowing over PPRC links.