ORIGINALLY POSTED 15th December 2007
7,792 views on developerworks
Back in August I posted about the ‘Stealth’ announcement of Invista 2.0. It seems that our friends over at EMC decided that after almost 6 months, they’d try again and actually tell the world this time. I thought I’d wait till the end of the week to see if the rest of the marketing, analyists and bloggers out there were as under-whelmed as I was when I actually saw the announce which we’d heard was coming on Monday. Are they going to embed Kashya in the product, are they going to provide some amazing way of overcoming the problems with Switch Based solutions and provide true FlashCopy facilities (instead of the Cloning that they claim as point in time) … No. They reduced the price, upped some limits and can support Kashya, but you still need yet another expensive line card to get you there. I have to say I was hoping for something a bit more interesting and something to shake up the market a bit more, but yet again it looks like – and it seems a lot of the commentors out there agree – they’d rather ride the SRDF and DMX cash cows for a bit longer rather than actually provide what customers are looking for. An all in one box that breaks the ties on vendor lock-in.
Anyway, I’m not going to dwell on this, it really was a bit of a wet fish announce. Oh, and VMware ESX certificaiton is coming, must be a bit embarrassing that SVC was supported before their parent companies product. I’m sure that won’t always be the case…
Here’s what I’ve read and found amusing out there :
CBR : EMC tries again with Invista – Here a couple of points, also noting that this is nothing new, been out there for 6 months. A couple of points that I’d like to address from this article though :
IBM’s SVC was harder to discount as an alternative to Invista, Rubesch said. But the university accepted EMC’s argument that Invista would present less risk of being a bottleneck on throughput, because it runs directly on a smart SAN switch, while the SVC runs on a cluster of Intel servers.
Anyone who has EMC coming at them will recognise as Sales talk – however speak to IBM, let us do a Proof of Concept simulating your actual environment and you will see its nothing more than FUD.
If a virtualization system fails, access to data is lost. So the virtualization systems such as the SVC, Invista, and USP are all protected by high levels of component redundancy. But unlike the SVC and USP, a pair of Invista smart SAN switches and their supporting control appliances can be split apart into separate buildings or floors.
While this is true for USP, it is not for SVC. We have always supported SVC nodes, up to 500m apart (Thats 200m more than Invista now claims) There are some caveats, as always, but it can be done.
Linux Magazine : EMC Advances SAN Virtualization – Hmmm, hardly really an advance, more like a step backwards. Its taken 2 and a half years and it still doesn’t have any more features than its first release. Hardly an advancement, and it would seem customers are voting with their feet. With only 200 of them actually trying it out, and some of these are likely to be the ones that have told IBM that EMC gave them one to try, but it was so complicated they didn’t install it.
Doc D’Errico, EMC Vice President and General Manager, Infrastructure Software Group, said, “EMC believes in delivering the right storage tool for the right job. When it comes to the non-disruptive movement of data across heterogeneous arrays with the ability to virtualize from the server through to the storage…
So it would seem that EMC are picking on the ‘killer app’ of Virtualization, but nothing more. It can deliver many more things than just migration. But as I stated above, that would eat into the big chunk of profits from DMX and SRDF licenses.
The Register : EMC and VMware perform virtualization upgrade together – Which concentrates more on the VMware side of the story. It doesn’t really tell us anything more, but does get a quote in from SVC’s marketing manager which links you to the Byte & Switch article.
Byte & Switch : EMC Overhauls Invista – More of the same, but for those that noticed, and for the many many questions and comments I get regarding ‘Thin Provisioning’ on SVC you may notice Chris’s comment in this article. So there is your answer.
be very very interested to know if, and how, Invista could ever support
Thin Provisioning, without a ‘major’ performance impact or re-design.
Its not a clustered solution, so bolting on a failover node isn’t really
the same as SVC which is built from the ground up as a cluster.
Remember that even if EMC provide Thin Provisioning on DMX or any other
platform, unless you put it in your Virtualizer you need every single
one of your heterogenous controllers to support it. Stick it in your
Virtualizer and now even the cheapest of the cheap box can be thinly
provisioned. It also becomes the only place you need to monitor and
manage. Not ten different vendors, with ten different terminologies, ten
different warning and threshold levels etc etc. This is where you start
to see that unless your Virtualization device is flexible, powerful,
can cache I/O, can cluster efficiently (so that global meta-data can be
maintained), then you are limited to only a subset of features. (For
example, next years must have feature hasn’t been thought of yet – but I
bet you we can implement it in SVC if we want – can the same be said
for a ‘packet-cracking’ filter box like Invista…)
Anyway I’ve rambled on enough, come on guys, you must be able to come up with something more interesting than that lot in 2 years – are you taking this market seriously, or is it as I’ve always assumed, you have to just ‘have something in your portfolio’ to silence a few awkward customers that want to break out of the EMC mould.