CX600 vs VA7410

[Lacrosse]

I am looking for good white papers or third party reviews comparing the VA7410 to the EMC CX600 CLARiiON unit we are wanting to use the VA7410 but our management has EMC bad mouthing the performance and reliability and I need some good docs to prove otherwise.....any help would be much appreciated

[Dave Wherry]

I can't find a specific document. Here are some numbers to look at taken from the HP and EMC web sites. (Hope it lines up resonably well).
VA7410 CX200 CX400 CX600
IOPS 34,000 40,000 60,000 150,000
MB/sec 330 200 680 1,300
Logical Units 1024 256 512 1024
Ports 4 4 4 8
Max Cache 4GB 1GB 2GB 8GB
Drives 105 30 60 240

So you can see the CX600 does have better performance numbers than the VA7410. It should. It's not really targeted at the space the VA7410 plays in. The CX600 is really targeted at Symmetrix customers. Enterprise class storage.
You'll see that the VA stacks up very well against the CX200 and decently against the CX400.
None of the vendors make truly "like" systems so it is always difficult to compare.
When EMC announced the CX600 it positioned it against the HP EVA. That's placing the CX in the high end, not the mid range. So the CX600 better outperform the VA. How does it stack up against the EVA 3000?

EVA3000 CX600
IOPS 162,000 150,000
MB/sec 335 1,300

Sorry I do not have the rest of the numbers at hand. You can see we're talking about the same class of systems. I can't explain the difference in MB/sec throughput. Numbers this disparate are usually based on different reporting methods.

So it really comes down to what are your requirements? Does the VA meet them or do you need to move up to the Enterprise class systems? If so, it's not fair to compare the VA. You need to look at an EVA or even a Hitachi system. The 9500 line is very nice too.
As for the reliability claims. I've heard nothing negative about the VA or any of the HP products. Before letting EMC drive the whole process you need HP, or an HP reseller in there to clear the air. That's not to say the CX600 is not a good product and would not do a good job for you. You're just not getting a fair comparison.

[Karsten_4]

Hi
the Problem is not as easy.
These Numbers are all "Marketing Values".
They have nothing to do whith the real World in your Datacenter.
Lets explain IOs :
These Value represents the max. read IOs whitch these arrays can archive OUT OF THE CACHE !
This dosent realy interest you, because 95% or more of your read IOs come from the Disks and not from the Cache. And even worth : most of these Numbers are archived whith Blocksizes that you never ever will use.
For Example :
VA 7410 : 34.000 IOs Read out of the Cache at 8k Block Size (This could perhaps match)
EVA 5000 : 168.000 IOs Read out of the Cache at 0,5k Block Size (0,5k ?? never used and you ?)
CX 600 : 150.000 IOS max. THEORETICAL possible IOs (This is a calculated Value and not measured, not very usefull)

Its the same when someone tell you : My FC Port does 200MB/s : Its only theoretical, in real Life Values between 160-170MB/s are very good.

If you want to know more lets talk later.
Im on the way ....

[Peter Mattei]

The other contributors have talked a lot about feeds and speeds.

For mr the major difference is Virtualization.
The VA7410 is a virtualized array the cx600 is not!

Now, what does that buy?
The VA puts all available disks into 2 pools (Redundany Groups or RGs). LUNs are created in either RG.
The LUNs are automatically striped over all disks in the RG.
If you add drives to an RG the VA redistributes all LUNs to all disks in the RG thus improving LUN performance and avoiding Hot-Spots. If you now create new LUNs they are again optimally striped! Automatically.

In a traditional array like the cx600 you manually define your disk groups and carve LUNs out of them. When you add capacity you have to manually alter the disk configuration. In a living environment you have to spend quite some time to prevent Hot-Spots and to keep your array in an optimal state.

So the difference is Virtualization which drives down management cost!

Regards
Peter

[Tim D Fulford]

I can't talk too much about the EMC products as we do not use them. But I can say what we see on the VA.

We have a VA7410 with 30x 15,000 rpm disks. It supports a OLTP database (Informix), we have 2GB cache (1GB per controller). It is configured as RAID1+0.

We get 60-70% cache (as shown on armperf), 100% write cache & 20-40% read cache.

The service times vary from 1.5-2.5 with a consistent average of 2.0 ms. Our AVERAGE block size is 2.5kB (Just over 1 Informix page size).

What does the above mean.. well 2ms per LUN equates to 500 IO/s /LUN, we have 13, so 6,500 IO/s shifting about 15.8MB/s. This might not seem too impressive when compred to the HP specs, BUT

o There system had 90 or more disks (not stated, but I'm sure this is the case)
o They used 4GB cache (not stated, but I'm sure this is the case)
o Our block-size is V small & configured ON PURPOSE for throughput & NOT BANDWIDTH.

If you had a system with 90 disks as 39 RAID1+0 LUNs and caching of, say, 80% on 4GB memory. The service time would be about 1.9ms, over the 39 LUNs this would give about 20,500 IO/s. If the block size was a respectable 8kB you could see 160 MB/s. I believe these are achievable, but it would depend on the type of system you you were trying to cater for.

Bottom line EMC do have a very quick (& expensive) product. Do you want to engineer a solution to save ???????????? or do you want the fastest system available?

Tim

I'm a simple man & just go on service time