VMware Cloud Community
khughes
Virtuoso
Virtuoso

Iomega StorCenter ix4-200d NAS Server performance?

I was curious if anyone is running the Iomega StorCenter ix4-200d NAS Server? I'm planning on maybe picking one up for our R/D and DR testing but was wondering how the performance was. Granted I'm not going to be running production on it, and using it mainly for proof of concept and testing but how much can 4 spindles really run?

  • Kyle

-- Kyle "RParker wrote: I guess I was wrong, everything CAN be virtualized "
Reply
0 Kudos
101 Replies
qmacker
Contributor
Contributor

Hmmm...for the flexibility of running VMware on a NAS, would I not be better off making my own NAS box with a few 10K or 15K SAS drives?

I'll soldier on and try the VMs on this device. i was just hoping for something "out of the box" that would be outstanding. Pity they aren't SAS drives inside the ix4-200d.

Reply
0 Kudos
tganzer
Contributor
Contributor

Sorry for the late response. I was looking for an article that helps explain storage performance..and I found it and provided the link below. Its an article from Anandtech that is a little dated because newer faster SSDs are out, but that doesn't change the picture much. As a storage and VMware admin (NetApp, HP MSA, and local SAS disk) for a small business (100+ mailboxes on Exchange 2003) and a fairly large enterprise (10,000+ mailboxes, 200+ TB) I have come to understand how often storage is the performance bottleneck. Also that sequential read/write performance means very little when trying to run applications that are very random read/write based. I think the article from Anandtech demonstratres that well. Sequential read/write is a best case scenario for SATA, and for this device. I mostly use it for file sharing and streaming videos to my xbox which it does nicely.

For a storage sensitive application like Exchange, although Exchange 2007 is about 50-60+% better than 2003, it still requires a decent storage system to run well. If you have it running on a server already you can use perfmon to find out how many IOPS you are currently using and then run loadgen against this device to see if it can handle your workload. Also from the Anandtech article it should get you asking the question, what about putting SSD in this device, if you don't need a lot of storage that might be a solution that would work for a small business.

I am also evaluating this device to be used as a repository for VMware Data Recovery product for the small business. Again I would recommend testing as much as possible and I would love to see those results with SSD because performance is so much higher on a smaller number of spindles. Wish I could get my hands on some SSDs to find that out... Smiley Happy If anyone is doing that let me know.

http://www.anandtech.com/IT/showdoc.aspx?i=3532

Reply
0 Kudos
Cstructure
Contributor
Contributor

The drive is not the limitation on the ix4-200d. It's actually the CPU that can't handle the load.

If you did design your own NAS it would out perform almost all out of the box nas and will be alot cheaper. (Just not as nice)

Reply
0 Kudos
tganzer
Contributor
Contributor

Admittly these devices are not powerhouses.. Smiley Happy A good site for building your own NAS and seeing benchmarks for these type of devices is www.smallnetbuilder.com.

Reply
0 Kudos
qmacker
Contributor
Contributor

OK. Some more results.

I setup an NFS share on the NAS and attached it to my ESXi server, a Dell Vostro 400 Intel Core 2 Quad Q6600 @ 2.40GHz with 8GB RAM. I added a dedicated NIC for the NFS (Intel PRO/1000 GT PCI). In fact, no switch, just a Cat6 cable directly from the Intel PRO 1000 to the 2nd port on the NAS. I put it on a different IP scheme and subnet from the other NIC (which I'm using to manage the NAS and do backups, etc.) just to be sure.

I straight COPIED an XP VM that I had handy. It has an 8GB "HDD" with about 1.40GB unused. I installed HD Tune and got the following results (see attached screenshot):

HD Tune: VMware Virtual disk Benchmark

Transfer Rate Minimum : 21.1 MB/sec

Transfer Rate Maximum : 41.2 MB/sec

Transfer Rate Average : 36.9 MB/sec

Access Time : 7.2 ms

Burst Rate : 41.4 MB/sec

CPU Usage : 7.6%

Next, I think I'll copy my own SBS 2008 server onto the NAS datastore. Is that the way to do it, by the way? Just "copy" from datastore to datastore? It's about 160GB. It's gonna take a while. Then do I just "add" it as a VM? What about when it askes me "Did you move or copy?"Does it matter, as long as I leave the old one off. This is a domain controller (DC) after all! Sorry if this last bit is off topic.

Reply
0 Kudos
jefmes
Contributor
Contributor

I don't see why that wouldn't work off-hand, qmacker. You could also download Converter too and do it that way, but I probably would just copy the VM folder over to the NAS and see if you run into anything. Like you said, as long as the original is off, you should be fine.

Thanks for taking the time to post your results, I haven't even had time to set mine up yet. I'll try to make sure I share my experiences too when I'm able so we can see if we're all performing about the same. I intend to use this for our home backups too, so I'm still playing with a few ideas on how I want to configure it.

Reply
0 Kudos
rengler
Contributor
Contributor

iSCSI doesn't require specific drivers to work, just a target and a host. The ESX host should talk to anything that talks iSCSI as there are no "flavors" that I know of that might require specific drivers. The better question in my mind is can you get multiple ESX boxes to talk to this. I'd expect so.

-Rob

Reply
0 Kudos
qimen
Contributor
Contributor

Jimm, do you still have this issue? The ESX servers do not require any speicial drivers or configuration. And the Iomega iSCSI targets support multiple sessions. Both your ESX servers should be able to see the same LUNs. I'd suggest you to check the LUN information on the Iomega management GUI to see which initiators are logged in. If both the ESX servers are logged in, then make sure you have the correct CHAP information if the LUN is secured. Based on your description, it sounds like the ESX server could not log in to the targets. Check your CHAP settings. Are you running ESX3.5 or 4?

Reply
0 Kudos
qimen
Contributor
Contributor

Hi, have you got mutual CHAP to work yet? Attached please find a document that describes in details how to use Iomega iSCSI on ESX, Windows, Linux, and Mac. There is a section about how to set up mutual CHAP for your ESX environment.

(sorry, the pdf version of the file exceeds the upload size limit. please see the Word version below.)

Reply
0 Kudos
qimen
Contributor
Contributor

Hi, sorry about your difficulties with Mutual CHAP setup. Here is a white paper that gives detailed instructions on using Iomega iSCSI on Windows, Linux, Mac, and ESX. The paper should solve your mutula CHAP issue. We'll get the paper on the Iomega website as well.

Reply
0 Kudos
khughes
Virtuoso
Virtuoso

Good to know that mutual chap is not supported with 3.5 and the 200d product. I was trying to get it to work but was having no luck and figured it was me messing up a configuration, not the case though. I'll be posting some IO tests hopefully at the end of the day today. I'm just building out the Windows 2003 VM's on the NFS LUN and the iSCSI LUN. I think the results might be a little skewed because I'm having to use a 100MB switch, but we'll see how it goes.

  • Kyle

-- Kyle "RParker wrote: I guess I was wrong, everything CAN be virtualized "
Reply
0 Kudos
qmacker
Contributor
Contributor

Okay, this is starting to really depress me.

I just did the same test, on that little XP VM, which I moved back over to direct-attached storage on the ESXi server. Bear in mind now, that my direct-attached drives are 7,200 RPM SATA-II setup in a RAID-1 configuration, so I'm not expecting them to be that fast. Not RAID-0 or RAID-10, just plain old RAID-1. Additionally, there is an SBS 2008 server running, as well as a Windows Server 2008 TS and a Windows Server 2003 TS. There's nobody logged on, and these servers aren't doing very much if anything - but they are turned on and Exchange 2007 is handling a little email traffic. So I expect a slight performance hit.

EVEN STILL, my own SATA drives waaaaaaay outperform the ix4-200d NAS, which had nothing at all running on it, other than my little XP VM. This kind of answers my questions...negatively, I'm sad to say. I had kind of hoped that the NAS in a RAID-10 configuration would significantly outperform my direct-attached SATA drives (on an HP E200i card, by the way). Definitely NOT the case. I'm already hesitant about installing direct attached SATA-II drives at any client sites - maybe I might in a RAID-10 configuration. However, there is no way I would install this NAS at a client site, even a small one. I'm kind of thinking of returning it now. I mean, really...what use is it? What a bummer.

Here are the results from the same test run on the XM VM (NOTE: THIS IS NOT THE NAS):

HD Tune: VMware Virtual disk Benchmark:

Transfer Rate Minimum : 17.3 MB/sec

Transfer Rate Maximum : 71.2 MB/sec

Transfer Rate Average : 51.7 MB/sec

Access Time : 6.6 ms

Burst Rate : 79.4 MB/sec

CPU Usage : 8.3%

See attached, and also, again note it is NOT the NAS.

Reply
0 Kudos
qimen
Contributor
Contributor

ix4-200r being a nework shared storage device, you can definitely take advantage of VMware HA, DRS, and VMotion with it. Now the question is, how do you define "enterprise"? And what is the performance requirements associated with that "enterprise"? I'd argue that the most important factor in sizing up your solution is to determine what is "good enough" performance. Take iSCSI for example, ix4-200r can deliver about 100MB/s in sequential read and about 85MB/s in sequential write, that's a performance close to the practical bandwidth of a 1Gbps network. Also, in theory, each SATA drive can deliver 71 IOPS, with a 3+1 RAID 5, the aggregate performance is no more than 280 IOPS. ix4-200r can deliver 242 IOPS with a 80% read 100% random profile in Iometer. Although ix4-200r is not officially in the Microsoft ESRP listing, testing has shown that ix4-200r can support up to 100 Exchange 2007 users using Jetstress. The only reason the product is not in ESRP is because it cannot physically separate Exchange data and log on separate spindles, which is an ESRP requirement.

So, ix4-200r may very well be the fit for your enterprise solution, it all depends on your application, the workload profile, etc. Some posts mentioned the Netgear ReadyNAS products and their high performance numbers. We need to be very careful with performance numbers, because they are determined by many factors such as I/O size, file size, I/O profile, cache size, etc. For instance, this is a performance report from SmallNetBuilder on ReadyNAS: . Take a look at the first chart, when the file sizes are small, most I/Os are cached in memory, so you get ridiculously high throughput. But when file sizes are increased to take out the cache effect, you get real performance numbers around 40MB/s. So there are marketing numbers that are flashy, and then there are real performance nubmers that matter.

Reply
0 Kudos
qimen
Contributor
Contributor

Just wanted to clarify, both ix4-200r and ix4-200d suppot mutual CHAP. It's ESX 3.5 that makes it difficult to configure mutual CHAP, I believe you will have to go to the CLI and manually modify the initiator config file to enable mutual CHAP support on ESX 3.5.

Reply
0 Kudos
khughes
Virtuoso
Virtuoso

Yes thanks, I did understand that part. I was reading through the white paper and noticed the difference between the 3.5 iSCSI settings and the 4.0 settings in regards to mutual CHAP. I had it all configured on the 200d side of things but it wasn't linking up correctly, because of the extra work you had to do on the 3.5 host. I just disabled it for the time being because I don't need it for my R/D and DR testing. I come from a FC shop so learning iSCSI was something I wanted to do. Will wait until I upgrade my lap to 4.0 for that.

  • Kyle

-- Kyle "RParker wrote: I guess I was wrong, everything CAN be virtualized "
Reply
0 Kudos
qmacker
Contributor
Contributor

Before I totally dis' the box, I should setup iSCSI on it first, and see how that goes. Currently creating a LUN...

A word on backups: One of the reasons I'm "sad" (rather than angry) about my results so far, is that the device is so user-friendly and nice to work with. The Backup feature is really cool. I was able to plug in an external USB drive which the device picked up right away. It actually adds it as a sharepoint within the NAS itself, kind of interesting. I then set a scheduled backup to backup everything in the NFS folder. I left that XP VM running and it backed up everything. You'll have to excuse my ignorance on this matter - I don't know if that backup is usable but it got everything in the folder. I suppose you could set a snapshot to run before the backup runs?

More results when I get the LUN setup.

Reply
0 Kudos
qimen
Contributor
Contributor

Hi, just wanted to offer a few comments:

1) The copy job is a single-threaded sequential write, it has not pushed the storage (either Iomega NAS or the local DAS server) enough, so IMHO, it should not matter to the comparison whether there are other applications running light load in the background.

2) As far as comparison, why would you think NAS should outperform DAS? The DAS is direct block I/O from the server to the disks, but NAS introduces some overhead such as network overhead (for instance, slicing I/Os and then re-assembling) and NFS protocol overhead at both ends of the connection.

3) From your numbers, the NAS access time is 7.2ms while the DAS is 6.6ms. That's about 10% difference, but the average transfer rate is more than 20% different. Is it possible the numbers are skewed somewhere? I wouldn't look at the burst rate, it's just a single data point.

I would say though that the NAS rate of 36.9MB/s (sequential write) is close to what I have seen using Iometer running the profile described here: . The disadvantage with DAS is that you cannot share files with other servers, certainly not in a VMware environment that requires HA, VMotion, etc.

Reply
0 Kudos
khughes
Virtuoso
Virtuoso

Along with everyone else I want to throw my two cents in as well. Unless I missed it somewhere, I don't believe that the ix4-200d or 200r are supposed to be major replacements for small/med-medium sized ESX environments. How much performance can you really get with a 4 disk RAID 5 setup? It really depends on what you are going to run within your environment and how much load is really going to be on those servers. There are some great documents to try and figure out how much resources your VM's are going to need. You can run those on the physical boxes and use those results to gauge how much power you need on your storage side.

I'll tell you first hand about now having enough power with your SAN and trying to make it work. When we first converted (granted we are slightly bigger) but we tried (and were told by the company that was helping us move virtual it was ok) to use our older SAN. Basically it didn't have enough write cache and completely bogged down the network. We had to purchase a new SAN and try to find some use for the old one, which we did down the road.

In the end, you want to be able to use the features like DRS/HA/DRM etc... it comes in handy quite a bit and is a good piece of mind. If you determine that you don't need a ton of power to move your VM's, and it is usually quite a bit less than you really think, I would say use it. It's on the HCL and it'll work to a certain capacity. It might by you enough time to save up and in a year or two down the road buy a quicker SAN that has more drives which will give you some better performance and more space if you need it. It really depends on your environment, your servers and what you need. There is no reason to buy all the bells and whistles if you're just going to get a DAS and not be able to use what you paid for.

  • Kyle

-- Kyle "RParker wrote: I guess I was wrong, everything CAN be virtualized "
Reply
0 Kudos
qmacker
Contributor
Contributor

I really appreciate everyone's help and input. It's an education for me too. I'll be the first to admint that most people on this forum know at lot more than I do about disk performance, especially regarding VMware. My "HD Tune" tests are probably a bit crude, and don't truly represent a proper load.

I'm probably being a bit unfair to the device it seems? To do the device justice, I'm going to move all my VMs onto it tonight. I'll start the SBS 2008 copy (160GB) now in a few minutes at say 1:30AM PST. Hopefully it will be done by 9 or 10AM. I'll boot it up then off the NAS, and we'll see how we're doing.

F.Y.I. - for what it's worth, I ran the same HD Tune test on the XP VM again, this time on the iSCSI interface. Below are the results:

HD Tune: VMware Virtual disk Benchmark

Transfer Rate Minimum : 18.9 MB/sec

Transfer Rate Maximum : 32.4 MB/sec

Transfer Rate Average : 29.8 MB/sec

Access Time : 6.6 ms

Burst Rate : 33.5 MB/sec

CPU Usage : 5.8%

Reply
0 Kudos
qmacker
Contributor
Contributor

One more thing:

qimen - to answer your question. All things being equal, I would NOT expect a NAS to outperform a DAS. However, I thought a NAS in a RAID-10 configuration "might" outperform a DAS in a single disk or RAID-1 configuration, even with the network overhead?

Reply
0 Kudos