PS5500E / PS5000E

[admin]

Im looking for some feedback from existing users. We are thinking of implementing the PS series arrays in our environment. It (to start) and they (down the road) would be used for our virtual environment (200+ VM's) as well as storage for our SQL servers and file servers (30tb or so combined). The heavy IO sql servers would likely remain using their current das for the time being. I have some quote requests out and waiting to hear back on the PS5500E from Dell (they are our vendor). I think the 5500e fully populated would be good for us. But am also considering the 5000e as well to get us started.

What I am interested in is the following: How large of an environment are your supporting with the PS sata array's. How many array's/units do you have connected? Are you using them solely for your virtual environment or does it also share as a storage repository for your other large data servers? Do you have SAS units as well as SATA and how well does this work assigning some servers to virtual pools created on the SAS arrays?

We were really considering going with the CX series but from what I have been hearing from a lot of people the PS does just as well excluding the fact that you cannot just add a couple drives as you need them, you instead have to be a new unit either half or fully populated and you cant mix sas and sata within the same shelf. The fibre of the cx really isnt a selling feature for us and the additional cost for every piece of software always keeps the price extremly high. It does however offer considerable granularity and scalabiliy on an as needed basis.

The other concern I have is, if we went with the SATA disks with a raid10 array are we going to see decent enough performance to warrant going with sata vs sas? As I understand the SATA in the PS arrays offers surprisingly good performance. I just need some reassurance from others running larger environments on these arrays.

Thanks

[admin]

Really looking for some input.

Thanks

[mikeddib]

We have been an EqualLogic customer for some time now and have three arrays, all with SATA drives, and we run our ESX / Exchange / SQL all backended to this system. We are smaller than your environment from a server perspective with about 50 VMs. We do have consistent data growth despite the smaller number of servers and our SQL DBA is happy with performance. During the initial implementation we live converted from RAID 10 to RAID 50 on their system and no one noticed, and of course no down time. Especially if FC is not a driver to consider other solutions, I would absolutely say EqualLogic is right there from a performance perspective. We also didn't like the EMC model where snapshots are provided through another box hooked into the array, replication is provided through another box hooked into the array. We preferred all the snapshot / replication technology built in, and it's included with all arrays. The downside of that model of course is you could end up paying for functionality you might not need. But as far as performance goes, we have never seen an issue in our environment and we do use the array for more than just ESX vmdk access. Hopefully you hear back from some other people with larger environments as well.

Mike

[azn2kew]

I've implemented the Equallogic PS5000E houses about 80 VMs with all kinds of OS (linux, windows, solaris) for testing and coding different applications development. Equallogic has very good performance and stability and also takes 20-30 minutes to configure especially free management tool built in. Its scalable as well if you want to expand or growth in the future, its easier to do so.

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

Regards,

Stefan Nguyen

iGeek Systems Inc.

VMware, Citrix, Microsoft Consultant

[admin]

Anyone else?

[Cooldude09]

Hi Scott, we are using PS5000E in our environment for appr 4.5 TB storage and we are using SAS drives of 450 GB 10K rpm. In case of production server SATA drives are not good as you will not get a good thruput for I/O...hope this thread helps...http://communities.vmware.com//thread/187692?tstart=0

Also using 5500E can scale upto 48 disks...so using SAS drives on that will defnitely give u better results

Regards

Anil

Save the planet, Go Green

if you found my answer to be useful, feel free to mark it as Helpful or Correct.

[happyhammer]

scott

ive implemented a number of EQL with ESX from 10-15 vms upto around 80 vm's including a 1500 user exchange environment. Do you have any idea on the number of I/O's you require ?

If its from a physical estate id recommend using platespin recon for a couple of weeks to get your stats and from here it makes the SAN planning very easy to see what I/O is required over a 24 period. You can then check this against what the SATA or SAS version iops to ensure you have sufficient arrays to cover the iops peaks

[malaysiavm]

Personally, I had not own any of the EQL box in my environment. Previously, I had 1 EQL box which loan from DELL, for our POC test in VMware. Here is what I found. Performance wise, the SAS hard drive is really perform and able to be tune up to the peak performance, which I will said is the best performing ISCSI storage I ever experienced. If you plan for bigger capacity, SATA definately is the option for you and if we do compare the cost and performance, you can always increase the numbers of SATA disk and make it as RAID 10 to improve the I/O as you need, and the cost of SATA is really cheap nows day.

Malaysia VMware Communities -

[Azriphale]

I love the PS series boxes. I use a 5000E for a test/dev network, consisting of two PE2950.IIIs running some 40 odd VMs. We restore around 100GB of exchange db at least once a week. The last restore showed a throughput of 2303MB/min and took 55mins. The bottleneck in this process is the backup server - a four year old 2650 with a 600GB RAID10 array providing the data partition for the backup file containg the exchange db.

My plan is to combine the 5000E with a 5000XV I'm using to build our live environment. I haven't got enough load on the XV yet to get a proper idea of the difference in performance. If my understanding of the Equallogic tiering process is correct, I should be able to combine them both into a storage group and let the arrays work out where to store the data based on what data gets used the most. I believe you can set preferences for each LUN but I figure the arrays have probably got a better idea of what's going on than me - like DRS.

Also take a look at the auto-snapshot manager and site recovery manager tools. Seriously cool.

HTH

Azi

[BenConrad]

I'd be careful about locking in on the PS5500E unless you don't have issues with purchasing more I/O down the line. The PS5500 will give you about 3 x PS000E I/O for the price of 2 PS5000E's, that's great! 38TB usable storage, yay! We've seen that the EQL SATA disks can do 100-120 IOPS of various size with <20ms latency, equallogic does a very good job with their SATA units. The 15K RPM units are terrific.

You need to be careful that you don't overdo the I/O on the SATA disks, if you do you will have consistent slow performance for ALL volumes. 200 VMs is a decent size environment, it will generate at least 200-400 I/O when the VMs are doing nothing, add load and you will go higher. Add too much load and you've got a swampy SAN.

You might want to start with a PS5000E and a PS5000XV (450GB drives) to start, that way you have some options if one of your apps kills the SATA disks. That is a major downside with EQL, you can't segment a heavy app onto it's own disks (unless you buy another array...).

Ben

[_VR_]

The PS5500E will actually outperform the PS5000XV as far IOPS go.

[admin]

That kind of seems to go against what one would expect.

[_VR_]

It really comes down to iops per disk and usable spindle count. RAID level selection also plays a big factor, but I'm assuming both systems are using the same config.

[admin]

I agree, it would. Im trying to wrap my head around the effecitvness of the sata array vs. sas when we are talking 14 x 7200rpm sata 2 vs. 14 x 15k rpm sas drives in the same raid configuration though. to be that seems impossible.

[_VR_]

The PS5500E comes out on top because it uses 48 SATA disks compared to 16 15K SAS disks in the PS5000XV

In theory, you would get 50%-60% less iops from SATA compared to 15k SAS with the same disk count. If you switch from RAID50 to RAID10 you will double your write iops while keeping the read iops the same.

[admin]

Your right. Sorry i was thinking the ps5000e with the 16 disks.

[BenConrad]

That might be true, but it's going to cost you +$30K more to get that performance ($40K vs $70K). And if you over commit the PS5500E with IOPS you are stuck with an enormous amount of sluggish storage. Having at least 2 pools (15K , 7.2K) gives you the flexibility to place volumes on the appropriate storage.

Ben

[_VR_]

PS5000XV (16 * 450GB) would give you 5.4TB usable in RAID 50

2450 read iops / 612.5 write iops

list price ~56k

PS5500E (48 * 500GB) would give you 11.5TB usablein RAID 10

3320 read iops / 1610 write iops

list price ~79k

[BenConrad]

What test profile are you quoting? % Read, block size, % sequential, Outstanding I/Os?

[3sh]

don't forget that the 6000 series came out now as well. Faster controllers, 10GbE will be added soon & more Cache on the controllers.