There is a fight going on for the soul of server storage and HP and EMC are in the front line trenches getting ready for all-out marketing war. HP's strategy depends upon smartening up its ProLiant servers. The eighth generation ProLiant is turbo-charging its storage with a Smart Array containing solid state drives and Smart …
So what matters: servers or storage?
So HP is big in servers, they'd want to perpetuate that situation, wouldn't they? EMC is big in storage and virtualisation, that's their land grab.
My crystal ball suggests servers don't matter that much anymore. You certainly need somewhere to look after your data though.
All this server cache stuff is transitory. If there's money to be made from apps that are disk IO bound then they'll migrate into a high performance, consolidated compute & storage platform. That only needs to offer a hypervisor on top of we know as a storage system. Presentation & session management can stay out there on the network.
Only my tuppence worth, t'ra.
Re: So what matters: servers or storage?
I think the point is that the increases in CPU & memory now available mean that simply loading those elements up to max capacity results in i/o bound systems - i.e. cpu & memory have progressed faster than storage.
Put another way, storage capacity has increased faster than its access ability. In the home sector, we've gone from 40Gig to 3000Gig still running at 7.2k rpm. That's fine for a home archive of movies, but it doesn't work when consolidating business loads.
Moving applications onto the storage array is simply reverting to DAS. It provides bandwidth at the expense of flexibility.
What we can do is sideways scale. More cables & switches & smaller disks. The requirement is to separate the data usage so that different users get different spindles and thus better speed. That is a clever trick and requires lots of expertise and analysis software. Its complex and even if you understand where your problems are, getting business to migrate data around for a system which is "working" is non-trivial.
Who knew that in March 2012 a company would be willing to come out and say that SSD is a game changer. Maybe their motto should be "Too big to innovate"
Slightly more serious (and unfortunately only slightly), but maybe HP should take a look at how disjointed their storage portfolio is. They have HDS rebrand (not for long eh), LeftHand, 3Par, and their own. Maybe they should consolidate that into workable brands they can engineer, innovate, and support. Or will it end up like their PC business (do they have one this week, can't be sure)
The most telling bit is that the majority of the storage in HP managed datacenters are.....not HP. They can't even sell it to themselves.
Re: About time!
(Full disclosure: I work for HP Storage)
You're speaking from a place of not really knowing all your facts. HP "managed datacenters" are moving to 3PAR storage in a big way; even before we had 3PAR, many of those included XP and EVA. Still in the end, HP Enterprise Services offers customers the storage they want - and if what they want is non-HP, they accomodate them. Also note that the XP is not a rebranded HDS product; HP has a direct engineering relationship with Hitachi Japan and in fact we have worked with them to deliver features that HDS does not have access to (e.g. Application Performance EXtender or APEX).
To my knowledge, 100% of HP IT data centers are HP Storage.
We've had 3PAR in our portfolio for just over a year. We have shipped over 100,000 EVAs and boatloads of P9000 XP Disk Arrays - we'd be stupid to drop a bomb on those existing customers by killing those products in the short term.
Lastly, I've worked for HP Storage since 1990; there's never been a time when HP was so focused on winning in the storage industry than now - the competition is noticing and we are in far more deals than ever before. Gartner's latest MQ on the disk array market is also a testament to the changes being made in HP Storage. That is what is most telling, not the competitive generated FUD you've offered.
Re: About time!
"we'd be stupid to drop a bomb on those existing customers by killing those products in the short term."
Exactly, "in the short term", but you know those XP boxes are out as soon as some critical mass number of customers migrate to 3PAR or something else. HP isn't pushing XP at all. It is true that XP isn't straight HDS, but it is mostly HDS. HP isn't going to continue shipping a large portion of the profit to Hitachi when they can put it entirely on now HP owned IP. Another re-start from HP.
Re: About time!
It's been 18 months and we still sell the XP today and nothing has been announced that we're going to stop selling it. The XP goes places that 3PAR doesn't and probably never will (like mainframe connect).
Call it a restart if you like but I'd call it HP has a solid strategy, a great product portfolio, and it's paying off.
Re: About time!
Yes, FICON is the official HP reason for XP staying in the portfolio. How many IBM mainframe customers use XP? 99% of those guys use IBM DS8 or EMC.
Maybe a solid strategy for HP. Not so much if you are an XP or EVA customer.
I think servers will becomes less and less important and will eventually only be reserved for key applications. Looking at some of the massive scale out architectures, such as Isilon, we're seeing storage clusters sitting there with a ton of memory, compute and network. They're basically hugh clusters of servers just waiting to be used.
Why do I need a server with VMs for most applications when I can run them directly on my storage array? That way I don't need PCIe cards in servers to being the data closer to the server or smart movement of data - the compute and stoarge are ALREADY close together and I don't need to buy servers AND storage.
This whole PCIe cache card stuff is a stop gap and I couldn't care less about how well servers and arrays commuicate - I'm looking for the two together.
Does EMC make SAN bootable LUNs today that require no CPU, or only the CPUs in the controllers, which can run VSphere, or whatever, directly from the storage array?
dont kid me ...
HP brought 3PAR. yes.. good move and I like the technology in 3PAR. Other than APEX... which uses Qlogic as its HBA ... what other innovation does HP does with XP? 99.9% Hitachi technology and 0.01% HP??? mmm .. fantastic ... and I didnt see HP selling more XP with APEX then HDS. funny ...
100,000 EVA .. since when? how many of the 100,000 EVA got swapped out by likes of HDS, Netapp, 3PAR (now HP), EMC and etc etc???
Stop dream and get real .... when HP do get real .. then they are serious.