I have bitched and griped in the past about how some storage companies waste their customer’s time, money and resources by not leveraging the Intel/Commodity CPU architecture that some of them tout so heavily.
Someone commented on here in response to my HP SPC-2 results pointing out that the new Oracle 7240 ZFS system has some new SPC-2 results that are very impressive, and today I stumble upon an article from our friends at The Register which talks about a similar 7240 system being tested in the SpecSFS benchmark with equally impressive results.
The main thing missing to me with the NFS results is the inability to provide them over a single file system(not just a global name space as NetApp tries to advertise but truly a single file system), oh, and of course the disclosure of costs with the test.
This 7240 system must be really new, when I went to investigate it recently the product detail pages on Oracle’s own site were returning 404s, but now they work.
I’ll come right out and say it – I’ve always been a bit leery of the ZFS offerings for a true high availability solution, I wrote a bit about this topic a while ago. Though that article focused mainly on people deploying ZFS on cheap crap hardware because they think they can make an equivalent enterprise offering by slapping some software on top of it.
I’m also a Nexenta customer for a very small installation (NAS only, back end is 3PAR). I know Nexenta and Oracle ZFS are worlds apart but at least I am getting some sort of ZFS exposure. ZFS has a lot of really nice concepts, it’s just a matter of how well it works in practice.
For example I was kind of shocked to learn that if a ZFS file system gets full you can’t delete files off of it. I saw one post of a person saying they couldn’t even mount the file system because it was full. Recently I noticed on one of my Nexenta volumes a process that kicks in when a volume gets 50% full. They create a quota’d file system on the volume of 100MB in size, so that when/if the file system is full you can somehow remove this data and get access to your file system again. Quite a hack.
I’ve seen another thread or two about existing Sun ZFS customers who have gotten very frustrated with the lack of support Oracle has given them since Oracle took the helm.
ANYWAYS, back to the topic of exploiting x86-64 architecture. Look at this –
ZFS Storage array base specifications
Clearly Oracle is embracing the processing and memory power that is available to them and I have to give them mad props for that – I wish other companies did the same, the customer would be so much better off.
They do it also by keeping the costs low (relative to the competition anyways), which is equally impressive. Oracle is a company of course that probably likes to drive margins more than most any other company out there, so it is nice to see them doing this.
My main question is – what of Pillar ? What kind of work is being done there? I haven’t noticed anything since Pillar went home to the Larry E mothership. Is it just dieing on the vine? Are these ZFS systems still not suitable for certain situations which Pillar is better at supporting?
Anyways, I can’t believe I’m writing about an Oracle solution twice in the same week but these are both nice things to see come out of one of the bigger players out there.