HPE Storage Users Group

hdtvguy wrote:

afidel wrote:- 3par has true IP based replication, Compellent is either FC or iSCSI

I don't consider the castrated 1GB RCIP ports on the 7000 series "true IP"....provide some SFP+ ports and let us choose to use 1GB or 10GB and then we can talk. It's still has an FC heart on the backend.

Maybe so, but you can still get 125MBs of replication through them.

Dears,

Just to update we are doing 7450 poc now, we had performed 2.5 million IOPS with 4k block size and 70/30 read write less than 5ms avg host response time....the San infrastructure is of 4 Gb/s. 7450 is with 2 nodes and 24 disks and 4 host ports .

Tested dedupe and IO performance in oracle and SQL Databases, for SQL dedupes are almost 93 % for second copies ,but oracle data dedupe not that good so far eventhough no compression or encryption enabled in dB side ....the IO performance we didnt notice any performance degrade from application comparing thin volumes and dedupe volumes ...we are testing using 8tb oracle 11g r2 dB by stimulating heavy IO dB queries....

Noticed dedupe values reporting from volumes seems incorrect from CLI...but from cpg and system wide the values looks good...our current FW is 3.2.1.

We are expecting similar tests with EMC xtreme IO soon .

Sincere Regards,
Aneesh

if your looking at XtremIO make sure you evaluate more than just performance and dedupe. Take a proper look at online expansion and firmware upgrades as well as the hardware roadmap. Plenty of people already bitten by the former two (which aren't yet fixed BTW) and the latter refresh to a proper V1.0 hardware platform won't be far off. Whatever you do don't accept their offer of a bench marking tool kit

I have to agree with johnmh on this one. There are a number of things that scare me about the xtremeio platform. For starters:

1) no online expansion
2) has had more than one destructive update. This is huge. We're not talking about an offline upgrade. This is destructive, as in free array wipe included with the upgrade. And they've done this more than once.

Emc claims that in order to improve, they needed to change the layout of the data on disk. That's fine, however having done it to to their customers suggests that the data is not laid out very well from the outset. They seem to not be able to create a layout that is scalable enough to meet the future needs of the platform.

I'm not in the market for all flash, in fact I don't currently have any flash at all, but if I was, I would fine pure or hp a much better all flash play

True the destructive upgrade on XtremeIO is scary stuff.

Plus I see the big advantage of the 3Par is your not creating another island of storage. If the data on your flash array 7450 get cooler and no longer requires flash, you can peer motion it to another 3Par. Plus you have the 7440c to choose from now which has the best of both worlds capacity and power.

Yes , we heard about the destructive upgrade in XtremeIO due to dedupe block size change from 4k to higher . We already requested more information and details about the further Xtreme node upgrade plan from EMC , still awaiting their response and box for POC.

Anyway so far we got amazing results from 7450 and matched most of our objectives except the data recovery points , which we are trying to setup Dataprotector ZDB/IR with HP 3PAR 7450 , but some of our challenges are 3PAR Inform OS 3.2.1 is not listed latest HP DP 9.01 support matrix though our backup environment is DP 7.03. HP storage team is working with backup support guys to get the modified support matrix to qualify the latest 3PAR firmware , also we need to upgrade our existing backup cell manager version to be minimum of DP 8.01 to get 3PAR zdb integratiion libraries.

The key for EMC in data recovery part is they can build the solution using Recovery Point appliance which have point in time recovery for volumes along with the Continuous replication ( CDP & CRR).

Sincere Regards,
Aneesh

see here's the thing, if the truth is that they had to change the dedupe block size, I'm not sure why that required a destructive upgrade. In theory, all they had to do was re-hydrate the data, change the block size, then re-run a dedupe process, and voila, done! I had heard that that was the reason, but it doesn't sit right.

Also, they have stated that the update from 2.4 to 3.0 will result in much greater performance overall. If that is true, then there must be more than just a dedupe block size being changed.

Not being able to expand the system online is a huge issue for me as well. I don't feel that I should have to take my system offline to add an x-brick. I mean seriously EMC, ever heard of a system tune to re-balance the data?

More and more, I don't understand why EMC gets included in the conversation when looking at 3PAR. EMC has been unable to demonstrate any ability whatsoever to execute a modern, high performance, scalable storage system. VMAX is the same as it's been for 20 years. VNX, again, hasn't changed drastically since the clariion days except for adding flash support, adding support for different drives, and going multicore. XtremeIO is a brand new system, that is still shackled in a number of ways that IMO a modern architecture should NOT be.

As far as recoverpoint, I still have issue there. As far as I know, in order to make that work, you would use XtremeIO on the back-end, front end your xtremeIO with VPLEX, and then attach recoverpoint to VPLEX. That whole thing just makes me go "ugh". I've not found HP data protector to be very worth my time (don't use it. I'm a veeam/AppAssure guy), but that cobbled together solution with recoverpoint to me doesn't seem a whole lot better.

aneezhcl wrote:Dears,

Just to update we are doing 7450 poc now, we had performed 2.5 million IOPS with 4k block size and 70/30 read write less than 5ms avg host response time....the San infrastructure is of 4 Gb/s. 7450 is with 2 nodes and 24 disks and 4 host ports .

2,5 million IOPS on a 2 controller 7450?! I presume you've used 24 SSD's? HP spec says it'll top at +/- 900k IOPS @ 0,7ms latency (4 controller). Sadly, I couldn't find any details on what blocksize or read/write ratio they've used.

Did you test a 7450c or the 7450?

We're thinking about switching from our Lefthand storage to 3PAR and starting our PoC with an 7200c. Because we sell pure SSD storage to our customers, the 7450 was our first target. However, the 7200 is now also being placed as an "AFA" due to it's AFA starter kit and is a lot cheaper. I was only worried about the max. IOPS.

HP now claims a 7200 can handle 300k IOPS @ 1ms, but it doesn't say how many disks are needed to get that much power.