TR Forums

Scrotos wrote:

I'm aware it'd be easier if I didn't go for the free Hyper-V version and had a Win8 or Server 2012 machine that could administer it, but I don't got the budget to buy the full Hyper-V or the desire to deploy Win8 or Server 2012. I plan to revisit it again sometime but meh, VMWare does what we need it to do and it does it for free.

Server 2012 Standard allows you to install one copy on the physical machine to run Hyper-V, as well as two virtual machines, so you may already own what you need to run a full-UI Hyper-V server. I'm not sure if OEM editions are the same way.

Scrotos wrote:

I'll give 'er a shot on my next go-around with Hyper-V, thanks. I tried using one or two utilities like this but didn't have any luck. This also looks interesting: http://blogs.technet.com/b/keithmayer/a ... -tool.aspx

Does Win7 RSAT work with Hyper-V 2012? Or version 3.0, whatever, the free one based on 2012 server core? I remember there being complaints about that.

I'm not sure if the Win7 version of RSAT will work for that. I've only done it with 8, but a quick search looks like it will only work for Powershell, and not the fun GUI tools. Feel free to send me a PM if you're doing a non-commercial install of it for testing. Even better would just be to create a forum thread about it. Ryu probably knows more than I do, and I typically just quickly scan the forums without logging in, so that would do a better job of getting my attention.

Quick performance testing update. I hooked the RE4 up to the H200 controller in whatever configuration Dell sent it from the factor, made sure I booted to my W7 install and ran HDTune. Results are as follows:
HD Tune: ATA WDC WD1003FBYX-1 Benchmark

Transfer Rate Minimum : 61.4 MB/sec
Transfer Rate Maximum : 132.7 MB/sec
Transfer Rate Average : 103.8 MB/sec
Access Time : 12.2 ms
Burst Rate : 166.3 MB/sec
CPU Usage : -1.0%

So, a fair amount better than the S100 results (may be driver related) but at least it wasn't significantly worse. I'm downloading IOMeter right now and I'll see if I can't get some read and write results from this drive on both connections. The SSDs should also be waiting for me at the office and if I can get some time tomorrow night we'll see if I can't get some results from those on both controllers as well.

Good luck!

So the SSDs arrived and I had some time to do some testing last night. There was definitely a range of results depending on the controller and RAID configuration I used. I don't know if I can create a table in a post or not, so I made a picture instead.

The first test was a control of sorts - hooking a single 840 to the 6gbps port on my Z77 workstation and running Crystaldiskmark (I haven't had a chance to set up iometer yet, and CDM just works out the box). This should be the best case performance for the 840 and it pretty much is in all categories. I also ran a single drive attached to the SAS 6gbps port on the H200 - performance was similar to the Intel port so at least the H200 makes a decent SAS port if not RAID controller.

Armed with knowledge of the what the disk could achieve I set out to find the best performance with redundancy that I could achieve with my hardware options. My first attempt was a RAID1 array on the H200 (I suspect some people must order their servers configured this way from Dell). Seq read speed was half of the single drive performance and seq write speed is an order of magnitude lower than a single drive. Random 4k writes are even worse, showing less than 1MB/s with a queue depth of 1. I figured this must be the whole disabled write-cache issue I've heard so much about. I downloaded the LSI management utility and enabled WC for the array. Performance improved significantly vs. WC disabled, but performance is almost exactly half of what a single drive could achieve (how's that for a selling point: twice the price, half the performance!). So, I can get RAID1 with what would probably be livable performance out of the H200, but it's not ideal and I lose TRIM as far as I can tell so I would depend on garbage collection on the disks to keep performance up long term.

I tried the S100 controller in RAID1 and enabled WC from the very beginning. Read performance is the worst of all configurations tested but write performance is actually pretty good. Finally, on a lark, I decided to try making a software RAID1 mirror set using the built in W7 disk manager while connected to the SAS 6gbps ports on the H200. Performance nearly matched or exceeded single drive performance, this option should have TRIM capability and using the 6gbps SAS ports seems to enable the drive to perform at its full potential. At this point in time I might need someone to convince me why I shouldn't use the software RAID1 mirror option? Dynamic disks make me a little nervous and I need to see if I can boot from a Dynamic mirror set (I think I can) but otherwise it seems to be the winner where performance and redundancy are concerned.

I'm not done yet and I want to make sure I give both the S100 and H200 a fair shake under Server 2012 that I will be installing on the SSDs in the near future. Current testing is done under W7 Professional and I'm not sure the drivers I was using were fully baked or offered the best the controllers have to offer. I can say with confidence that Write Cache makes a huge difference and makes me wonder what Dell and LSI were thinking with the H200 - I can only imagine how bad the performance would be with mechanical drives and I wonder how many folks are wondering why their new servers are so very slow.

Truth be told, though I've only had limited exposure to Dell server offerings, I've been very unimpressed with their management and RAID offerings compared to the HPs I've used. Reading up on the H200 it seems like it's absolutely horrific. You might do better to get a $250 RAID card aftermarket. My own testing with a PERC 5/i with 256 MB of BBWC showed it sucked with mirror performance (stayed at single drive speed) while windows mirroring and HP mirroring both scaled under load.

The SSDs will love the 6Gbps ports, though. If you get another RAID card, try to stick with that.

Something like these: http://www.storagereview.com/lsi_megara ... w_part_one

And oops, shows what I know, they are like $450.

So the comment about the Dell PERC 5/i with 256 MB of BBWC sucking got me curious about how our current server performed. I tried to wait for a quiet time without a lot of traffic to skew the results. It's a RAID 5 of 3x 160GB 7200RPM WD mechanical drives. Apparently I have had the incorrect assumption that a hardware RAID controller with battery backed cache memory would be decently quick. Boy was I wrong. Performance pretty well sucks and I'm thinking just about any configuration I tested with the SSDs would be an improvement. See below - yikes.


Sequential Read : 82.293 MB/s
Sequential Write : 16.881 MB/s
Random Read 512KB : 39.614 MB/s
Random Write 512KB : 8.582 MB/s
Random Read 4KB (QD=1) : 0.498 MB/s [ 121.6 IOPS]
Random Write 4KB (QD=1) : 0.133 MB/s [ 32.4 IOPS]
Random Read 4KB (QD=32) : 1.257 MB/s [ 306.8 IOPS]
Random Write 4KB (QD=32) : 0.142 MB/s [ 34.7 IOPS]

EDIT - I re-ran the test when I was the only one here and it's a little better but still pretty dismal. This new server will be like greased lightning compared to the current performance. I can't believe I've let it go on like this for so long...

Sequential Read : 94.186 MB/s
Sequential Write : 22.628 MB/s
Random Read 512KB : 30.755 MB/s
Random Write 512KB : 14.340 MB/s
Random Read 4KB (QD=1) : 0.430 MB/s [ 105.0 IOPS]
Random Write 4KB (QD=1) : 0.265 MB/s [ 64.8 IOPS]
Random Read 4KB (QD=32) : 1.091 MB/s [ 266.3 IOPS]
Random Write 4KB (QD=32) : 0.589 MB/s [ 143.8 IOPS]

Well, the PERC 5/i "sucked" for mirrors but from what I read it'd probably do fine with RAID 5. I wasn't able to test it in that configuration.

Honestly what's probably killing you on the old server are the 7200 RPM drives. We're moving from a 3Ware-based mirror and, well, you can see that in my iometer results under the "current" tab. Those were 7200 RPM drives as well. With the "SATA Database" graph you see the SmartArray P800 at top, then in order it's the Windows-based mirror with 2 7200 RPM drives running on the PERC 5/i, then the PERC-based mirror, then a 3Ware 9500 mirror, then a regular SATA controller with a single drive, and finally a regular SATA controller with Windows RAID. Yeah, I dunno why the single drive scaled better than SATA windows mirror, but there ya have it.

If I had any spare 15k drives, I could run some more tests on that PERC 5/i, but that system is set up to take 3.5" SAS drives and all my 15k ones are 2.5" SAS so it's kinda ghetto trying to prop them up.

You still may want to investigate getting a proper RAID controller, though, be it a PERC H700 or older PERC 6/i or something like that MegaRAID as even though the H200 is faster than your old system, you'll probably be leaving performance on the table with the SSDs.

At the very least, once you have a benchmark routine down, you could buy a controller from some place with a decent return policy and see if a proper battery-backed cached controller gives you any real-world performance. If it don't, return it. That's part of the reason I wanted to get iometer working for me so I could figure out if I were heading in the right direction with hardware and tuning for the expected workloads on the servers I was buildin'.

More PERC blah blah blah: http://www.dell.com/downloads/global/pr ... debook.pdf

BTW, what was the RAID controller in the PE840 you tested? And did it have cache and a working battery?

Scrotos wrote:

Well, the PERC 5/i "sucked" for mirrors but from what I read it'd probably do fine with RAID 5. I wasn't able to test it in that configuration.

BTW, what was the RAID controller in the PE840 you tested? And did it have cache and a working battery?

It's a PERC 5/i with 256 cache and a functional battery. Yes, they're old 160GB drives so their performance isn't going to be amazing even on their own, but 40MB/s write speed is pretty sad regardless. When I get the new server in place I'll see if I can't scratch together three more recent drives and see what the ole' perc can do, but it's pretty horrible.

If I get some time tonight I'll see about installing server 2012 on a single SSD connected to the H200 as well as a RAID1 array. I might end up with a combination that leaves some of the SSD performance on the table, but I need something reliable, fault tolerant, and faster than our current. Even my worst performance with the SSDs was better than current so it can only get better. As far as buying another controller - I'm only doing RAID1 so unless the H200 performs significantly better under Server2012 than Windows 7, I'll probably being doing software RAID inside windows. Performance nearly matched single drive performance, I get TRIM support... I have been researching and i can't find a huge downside to dynamic disks. I'll test all my backup and restore functionality as that's the thing I'm most worried about Dynamic disks messing up.

Yeah, if I think about it enough, I might give the ol' PERC 5/i a fair shot at some 10k or 15k drives. I'm just not sure I want to mess with it too much as it's a non-business system.

I don't think you'll find the OS makes much of a difference. As long as the write cache is enabled in both operating systems, it should be pretty similar. I'll be interested to see what you find, though! My OS of choice has been Win2K8 R2 for my testing.

Did you go with the Samsung 840 non-Pro drives for cost reasons? I sprung for the Pro ones due to longer rated life for writes.

To the OP (and all who have responded thus far):

Thank you, thank you, thank you for this thread. I am thinking of setting up a small server for personal use/experimentation later this year, and this thread has already been a wealth of information for me.

Bookmarked for future reference.

Gratitude!

!#@$*(&!$@ PERC 5/i benchmarks running today. RAID 5 w/128k stripe and 7 10k 72GB SAS drives, then RAID 10 with 6 and then RAID 50 with 6 just for funsies. TR patterns but going to a queue depth of 256. I'll post the results in the iometer thread and here after I put the drives back in their caddies and put them back in the HP server I ganked them from and run similar tests on that for a comparison.

Heh, my windows is not genuine. I may be a victim of counterfeiting! Blah blah blah etc. Guess it's been a while since I installed Win2K8 for temporary testing.

Scrotos wrote:

Yeah, if I think about it enough, I might give the ol' PERC 5/i a fair shot at some 10k or 15k drives. [SNIP]
Did you go with the Samsung 840 non-Pro drives for cost reasons? I sprung for the Pro ones due to longer rated life for writes.

Thanks for checking into that - if you find that performance is better than I'm seeing with the H200 I might devise some kind of plan to use the 5i in the new server, but probably not. I would be interested in what you see performance wise for a RAID1 of two 15k drives (or SSDs if you can spare them).

Yes, the 840s vs. Pro was primarily a price decision. Our workload is not so large that we'll wear them out before the useful life of the server, but the RAID1 is also some protection against that as well. I am also going to leave some additional spare area unformatted to allow for better wear leveling and write performance (although TRIM should help that as well). I have a Pro at home and love it, but for the capacity I wanted (250GB) and for the price I got them for ($170 each) vs. the cost of the 256gb pro ($230ish) the relative performance difference wasn't justifiable for the price premium (for me). We'll see if that was a poor decision down the road if the 840s have some kind of horrible hidden bug that remains to be discovered, but for now it was my best option for the price.

I did get Server essentials 2012 installed on a single drive and then made a dynamic disk mirror, it synced to the new disk and as far as I can tell everything is working as it should. I didn't get much more time than to shut it down before bed last night (redbox had the Hobbit on bluray so I wasn't as focused as I could have been) but this weekend I should be able to do so more testing before I settle on a configuration to move forward with the install.

I actually do have 2 spare 15k drives so I'll fire up the ol' system one more time and get some RAID 1 scores. Ain't got no SSDs, I probably should order a few in.

So, got a free night and started digging into Server 2012. Somewhere between just poking around and getting familiar and running the crystaldiskmark benchmark something went awry. The screen locked up and I didn't have mouse or keyboard input, but a press of the power button initiated a shutdown. After that the system rebooted and sat on the loading screen for half an hour. I shut down by holding the power button and restarting and the same thing. So, I pulled one of the drives and hooked it to my W7 machine. I showed as a corrupted Dynamic disk and my only option was to convert to a basic disk and essentially wipe it out. Interesting. Possibly a big strike against dynamic disks if they were the cause of the problem here. Who knows. I didn't exactly have a lot invested in the installation so I pulled the other disk, wiped it as well (the windows installer didn't have the ability to overwrite dynamic disks directly so I was planning ahead). Now I have to decide what road I would like to take for the reinstall - dynamic disk mirror, try the H200 with write cache enabled, etc. I think I might try the H200 mirror and see how bad it might be.

Oh, I did see the results of the bench before i shut down - it was 300ish read, 60ish write. Hmm....maybe server 2012 disables write cache for mirrors as well, but W7 doesn't? I'll eventually get around to testing that scenario again and hopefully I can dig into that some more.

I've never seen corruption due to use of dynamic disks. Usually it is failing hardware of some sort.

just brew it! wrote:

I've never seen corruption due to use of dynamic disks. Usually it is failing hardware of some sort.

Well that's reassuring - I would hope that a seemingly core feature of the disk/file system wouldn't be that flaky. Hardware failure makes me a little more nervous. More testing and break-in required before pressing into service I guess.

frumper15 wrote:

Yes, the 840s vs. Pro was primarily a price decision. Our workload is not so large that we'll wear them out before the useful life of the server, but the RAID1 is also some protection against that as well.

I am jumping in the middle of this, so if this is out of place no worries. But is Raid 1 a good option for SSD's? If you are worried about disk failure that is one thing, but if you are using Raid 1 to protect against data wear that seems counter productive. In Raid 1 the exact same information will be written to the drives in the exact same time. Theoretically, making these 2 devices 1 drive. If you had spinning disks that's great, because if one of the disks goes bad you have failover. But with SSD's the write integrity of both disks will be compromised at the exact same rate. If the disks could actually be the same that means both drives would fail at the exact same time for the exact same reason.

You would be better off with Raid 0 or JBOD and doing a bare metal backup to an external drive at regular intervals. Increase space, increased performance, and the drives will wear at the correct rate.

RAID 1 will protect against physical failure just the same with SSD as with mechanical. I don't think he's doing it primarily for protection against wearing out from too many writes. If he cared that much about it, he'd go with the Pro or with something SLC-based.

My assumption is that it's there for speed, some redundancy, and he's got a backup plan in place so it's not the end of the world in case they both die at the same time.

With the controller he has, who knows if RAID 0 would even give a performance boost!

LaChupacabra wrote:

frumper15 wrote:

Yes, the 840s vs. Pro was primarily a price decision. Our workload is not so large that we'll wear them out before the useful life of the server, but the RAID1 is also some protection against that as well.

I am jumping in the middle of this, so if this is out of place no worries. But is Raid 1 a good option for SSD's? If you are worried about disk failure that is one thing, but if you are using Raid 1 to protect against data wear that seems counter productive. In Raid 1 the exact same information will be written to the drives in the exact same time. Theoretically, making these 2 devices 1 drive. If you had spinning disks that's great, because if one of the disks goes bad you have failover. But with SSD's the write integrity of both disks will be compromised at the exact same rate. If the disks could actually be the same that means both drives would fail at the exact same time for the exact same reason.

You would be better off with Raid 0 or JBOD and doing a bare metal backup to an external drive at regular intervals. Increase space, increased performance, and the drives will wear at the correct rate.

Scrotos answered correctly, but I'll just elaborate. RAID1 is simply an uptime benefit for me. If 1 drive dies for whatever reason things will simply carry on as normal. It appears it will carry a slight performance penalty regardless of the configuration I ultimately land on, but the "insurance" is worth it to me. Regarding drive life - Of the 235ish formatted GB available on the drives I have, 100GB will be the OS/program partition and 100GB will be a data partition and 35ish will be unused. It's my understanding that this unformatted area will be used by the drive for wear leveling and should give me a cushion for garbage collection if the RAID solution I use doesn't support TRIM.

So, interesting wrinkle I've discovered. Windows Server 2012, by default, disables write cache even on the Individual drive. I installed server 2012 on a single 840 and tested the performance. I was once again down at 60ish MB/s writing. Hmm. Sure enough, write cache was disabled for the single disk. Weird, I thought, but whatever. Enabled write cache, retested and results were back over 250MB/s. I restarted in the process of preparing the server and ran the test again - back down again. Apparently being a domain controller means that Windows is going to disable write-cache automatically. I don't recall that behavior earlier in my testing, but who knows. According to this thread it's expected, but there is also a script I can run. Hmm.. http://social.technet.microsoft.com/For ... 2496d39ee/

Now I'm a bit torn. From what I'm reading I feel like I'll be compromising reliability if I enable write cache when I've invested in two drives to increase it. Hmm...more testing to gather data to make a decision - sometimes being analytic is a pain...

I'm delayed on my comparative test of 10k RPM drives. I have the PERC done but not the SmartArray.

I've enabled write cache on all machines that I can, including 3 AD machines (1 being in a VM on a machine I enabled write cache on for all the VMs). This is 2008R2 so I assume they aren't disabling it again. Just checked, still good to go via HP array config utility.

Since all our machines are backed up via UPS, I don't fear that a power outage will leave the drives in an inconsistent state due to not having the writes all completed. Now my understanding is that there are two levels of write cache, the RAID cache and the per-drive physical write cache. Most proper RAID controllers have a battery-backed write cache so that a power outage will let the controller complete the writes once power is restored. For me, not so useful because why would you have a server not protected with a UPS? But you need it to expose the best performance for most controllers. A physical drive cache, though, would just lose data in that scenario since it's past the RAID cache and not yet committed to the physical media. And even on the HP controllers they throw up gang signs warnings when you do it. Definitely helps performance, though.

So what you're dealing with is the physical drive write cache since I believe the H200 has no cache of its own. Again, danger danger danger but if the server is protected by UPS I wouldn't be too concerned. If you're paranoid, after you migrate to the new server, move the other machine into a corner and have it be a secondary AD machine as a backup in case your main machine dies. AD is pretty lightweight duty especially with as few people as I think you mentioned work there.

Maybe I'm throwing caution to the wind too much but the physical drive write cache doesn't make me break a sweat. I guess ask me after I get burned by it. ;D

So after much testing and back-and-forth between different RAID1 configurations on my available controllers, I have settled on using a dynamic disk mirror after benchmarking, testing failure modes, backup solutions, etc. I feel it is the best compromise of speed, reliability and usability. I hope I won't be kicking myself a few months down the road because of what I consider to be a complication of using dynamic vs. basic disks, but I don't suppose it's any more compromised than a proprietary RAID5 controller, etc. as far as ability to troubleshoot drives, etc.

I wish I could say I was more enamored with Server 2012 than I have been so far. It's fought me more than I think it should for what I would think are simple things like installing the DHCP and getting the alert that I haven't enabled Microsoft Update to clear (I have) to name the two on the top of my mind. I guess I don't feel like I should have to go hunt down obscure error codes to figure out how to work around errors on a clean server install on brand new hardware. But I digress...

Next up is installing and testing all our applications, configure and test all backup schemes, etc. I can say the hardware/software passed the "disk failure" test without any intervention. The machine kept running without issue and restarted without issue as well. I'll have to see about configuring an alert if the dynamic mirror fails, otherwise it seems like it simply just breaks and carries on with nary a word. I would have thought for built in functionality (that has existed for a good long time) something like that would pop up in the alert viewer at the very least...

If everything passes the smell test and appears to be ready for prime time I'll have a cutover party where I change the client and network configurations to support the new server and the old server can go into hibernation. If all goes well for a while I'll figure out what useful purpose the old server can have in its second life.

There's probably some event log you can trigger on. I just set this up for another server:

http://www.whiz-tech.com/blog/2013/01/0 ... windows-7/

Though I ended up making it call a .bat and used something called bmail on the command line as our Exchange server uses a specific port for internal anonymous email and you can't specify that in schedule task triggers.

So it's been over a month on the new server and (knocks on artificial wood desk) no problems to speak of. Well, I take that back - I have a problem with Crashplan becoming unresponsive, but I'm working through that with support right now. The server itself has been quick, reliable, and generally hasn't drawn any attention to itself. It's even significantly quieter than the machine it replaced. So, overall, very happy with how things turned out.

That allowed me to bring the old server home to do some experimenting. We have a number of dell desktops here that serve as client machines for a number of folks. Dell, in their wisdom, has an OEM and recovery partition that don't/won't backup with the client backup built into Server 2012 essentials (works just like WHS does/did) or at least not correctly. While one of the users was out last week I pulled the 500GB 7,200 RPM drive that came in the machine and reinstalled W7 pro, drivers, programs, etc. on an 80GB drive I had as a spare. Now, I'm not cruel, that was merely step 1 to get a clean install image that would back up like it should have without all the Dell partition craziness. Step 2 is to replace the largish mechanical drives with smallish SSDs.

To that end I just had delivered 4 x 120GB Samsung 840 drives. Before I press the drives into service I needed (NEEDED I tell you) to test their speed and reliability and I chose to see just how good/bad the PERC 5/i was when paired with much faster drives than the old 160GB drives it had been running. First up, 4 drive RAID0 to see the full potential of things

Sequential Read : 495.781 MB/s
Sequential Write : 419.766 MB/s
Random Read 512KB : 353.530 MB/s
Random Write 512KB : 419.430 MB/s
Random Read 4KB (QD=1) : 18.606 MB/s [ 4542.6 IOPS]
Random Write 4KB (QD=1) : 41.205 MB/s [ 10059.9 IOPS]
Random Read 4KB (QD=32) : 92.573 MB/s [ 22600.7 IOPS]
Random Write 4KB (QD=32) : 45.218 MB/s [ 11039.6 IOPS]

Pretty good - i don't think the 8x pcie slot would limit the bandwidth so I wonder if the individual drive interface is holding us back - or maybe that's all the better the card/drives can do. I've got another 8 port PCIX card that I want to try and luckily the PE840 has PCIX slots - now to figure out how to power the 4 drives - I'm not sure I have enough sata splitters so I may have to hotwire another PSU to get the drives running.

Next up was RAID5 - the mode I had been running in for 5 years with 3 x 160GB sata drives. 4 drive RAID5, default options:

Sequential Read : 393.708 MB/s
Sequential Write : 76.188 MB/s
Random Read 512KB : 359.592 MB/s
Random Write 512KB : 63.312 MB/s
Random Read 4KB (QD=1) : 19.240 MB/s [ 4697.3 IOPS]
Random Write 4KB (QD=1) : 6.579 MB/s [ 1606.3 IOPS]
Random Read 4KB (QD=32) : 109.138 MB/s [ 26645.0 IOPS]
Random Write 4KB (QD=32) : 16.766 MB/s [ 4093.3 IOPS]

Wow - I think it's safe to say the RAID5 calculations are seriously holding back the performance of these drives. If I get some time I'll run RAID10 to see what's possible, but it's pretty clear in my mind that I don't want to use the 5/i for much of anything.

I haven't had a chance to finish my P400 comparison to the PERC 5/i under the same RAID 5/10/50 configurations. However, I think EVERYTHING is holding the SSDs back. The SAS/SATA links. The backplane itself. The controller. The PCI-X slot. Same with the HP controller I've been using.

I'd give comparisons but I'm using the TR test pattern, database only. The closest I have is a RAID 10 with 8 840 Pros at 4K data random write pattern, IOps range from 5000 to 15000 going from QD 1 to 256. 20 MBps to 60 MBps.

Put one of those drives in a modern machine with SATA3 and it will blow our servers out of the water as far as performance goes. For me the price for entry into a more modern HP server was high enough that it made sense to throw lipstick on a pig and shove SSDs into a system that wouldn't fully exploit them. My other alternative was a bunch of 15k SAS drives that would end up costing more than the SSDs.

Interestingly - it seems like this guy is getting perfectly acceptable performance out of his 5/i - http://www.overclockers.com/dell-perc-5 ... enchmarks/

Hmm...it's worth looking into a little bit for me, but I'm not sure where I would even start... Maybe I'll see what I can do to get the controller into my gaming rig to see if maybe the whole system might be the issue.

About to run out. I don't know anything about ATTO, but I see 256 MB as the total length? Is that the entire size of his test? My 5/i has a cache of 256 MB standard and I think it can be expanded to 512 MB? I dunno, is it possible that his entire test is being run from the cache? When I did my benchmarking I used a size of multiple GB to try and eliminate the RAID cache as a factor.

Allotta good stuff in here. A couple of things, minor points. VM hosts should generally NOT be joined to a domain; it's the chicken. Having a DC in a VM also can play havoc with time sync, especially if you have a lot of VM's. If you have a lot of VM's, make at least one physical DC so that it can sync domain time to the outside world. Mo betta.

Exchange 2012 is a not a finished product. If you do upgrade, stick to Exchange 2010, at least it installs correctly (eventually) and isn't as ugly. However, the trend these days is to migrate your Exchange to a hosted solution, which is why it isn't bundled with SBS2012 (sorry, it's now called WS2012Essentials). If I remember correctly, Exchange 2012 won't install on SBS2012 at all. Personally I like in-house, but it's hard to convince admin types of its goodness over trendy.

I do like Server 2012's Server Manager, but I'm so totally not sold on the rest of the Win8 UI. Server 2008R2 is pretty damn fine for just about everything. Get off my lawn.

Also, am I the only one who likes Dell Outlet?

Here!

Scrotos wrote:

About to run out. I don't know anything about ATTO, but I see 256 MB as the total length? Is that the entire size of his test? My 5/i has a cache of 256 MB standard and I think it can be expanded to 512 MB? I dunno, is it possible that his entire test is being run from the cache? When I did my benchmarking I used a size of multiple GB to try and eliminate the RAID cache as a factor.

Hmm...interesting thought - I have been running the 1GB test in CrystalDiskMark so I'll try running it with the smaller test sizes to see if it makes a difference.