16 replies to this topic

[Greg Welch]

Thought it was time i stop lurking in the Podcast and post a little in the forums LOL.


O.K. Guys whats your thoughts.


First the System then the problem

X8DTH-6F,E5620,12GB-ecc reg Ram,LSI MegaRaid 9260-4i raid card.

Ok just bought the LSI MegaRaid 9260-4i and not being able to wait for 4-2TB hatachi Drives

i put in 3-2TB WD EARS green drives in Raid 5 using LSI mega Raid Storage Manager then tried using ATTO to bench it ,

So sad, what happend?



O.K. after i was done wiping away the tears

i thought why not just stick a single drive on the 9260 and give it a bench with ATTO



what ? thats I i said ! (i did the test multiple times with the same results)

so i tried CMD



O.K. little more like Expected.

This is what my Raid Manager was set up like, remember this is just 1 drive till i order the 4 Hatachi the other below is the onboard SAS raid working as my WHS 2011 Raid




now when i do a drive to drive transfer.

from onboard SAS to the single Drive i get 250 MB to 300 MB Transfer speeds witha 4GB file

from single drive back to onboard Raid 100 to 110 MB with same 4GB file


So i Surmise

1)My single drive is working great
2) LSI MegaRaid 9260-4i dosn't like the 2TB WD EARS green drives (or its something i missed


side note :

I cant wait to get the 4 -2TB Hatachi Drives then a pinch of LSI Cachecade to finish it off


Well thanks for letting me ramble on p.s. still like lurking in the stream of the podcast LOL


Greg Welch aka WELCHWERKS

[Greg Welch]

O.k. let me start out by saying it wasnt the 9260 that didnt like the atto bench it was 1 of the 3 ears drives

i have sinse just raided the 2 good drives in raid 0 and look at these scores holy crap.



im starting to like this MB and Raid Card wonder what cachecade will do to bench scores

[Greg Welch]

o.k. this is for 2 ears drives in raid 0 512 sector and 512 format

so this tells a differnt story that above maybe i need a third test

[timekills]

You're CDM shows a fairly linear growth - exactly what you'd want to see when you add an additional drive. Obviously there will be a point of diminishing return. The smaller file transfers will of course show less speed increase with additional hard drives because the RAID overhead is closer to the size ofthe file itself.

[ikon]

O.k. let me start out by saying it wasnt the 9260 that didnt like the atto bench it was 1 of the 3 ears drives

i have sinse just raided the 2 good drives in raid 0 and look at these scores holy crap.



im starting to like this MB and Raid Card wonder what cachecade will do to bench scores

On the face of it this has the look of cache reading/writing, rather than actual drive performance.

[Greg Welch]

O.K. this is what im getting my raid 10 to the raid 0 above

now a couple of things i know will help.

1. the ears drives arent 6.0 Drives

2. the raid 10 is 4x2TB 6.0 but they are connected to the onboard SAS with no cache (but there own 64Mb each)

from onboard raid 10 to Raid 0 above a 4GB Folder (tangled movie if you must ask)



from Raid 0 to onboard Raid 10



these speeds are about the average for these file transfers

[ikon]

O.K. this is what im getting my raid 10 to the raid 0 above

now a couple of things i know will help.

1. the ears drives arent 6.0 Drives

2. the raid 10 is 4x2TB 6.0 but they are connected to the onboard SAS with no cache (but there own 64Mb each)

from onboard raid 10 to Raid 0 above a 4GB Folder (tangled movie if you must ask)



from Raid 0 to onboard Raid 10



these speeds are about the average for these file transfers

silly me. I completely forgot about the existence of SATA III drives. I don't have any such drives so I have never run ATTO on one. I still find the discrepancy between read/write speeds a bit odd.

[Greg Welch]

i know not ideal drives for testing, hoping to add 4-2tb hatachi 6.0 to the card in the next month once i get wife/cfo approval

Edited by welchwerks, 15 August 2011 - 09:33 PM.

[c0op3r]

I have this card also, I was using the Intel on Board SATA RAID Controller for a RAID Ø, (2 Intel X-25 SSD's) but really wanted to get way from the quasi-software RAID. I was looking for a card that had on-board RAM and processor and was also more then PCIe 1x. So the LSI 9260-4i was basically the best price point for those specs without using the HighPoint (which I am not convinced are not using the processor of the workstation itself).

I am not going to make to many comments about how good the card is as I really don't have a lot of bases to make those kind of comments as I have only used a few cards.

I can show you the stats for the outcome of how the installation before and after.

This is CrystalDiskMark with the (2) Intel X-25 SSD's on the Intel Chipset RAID Controller RAID Ø configuration. This configuration yields a 7.9 on the Windows Experience Index.



Here is the CDM run on the LSI 9260-4i with the newest drivers and flash installed on the card:



The LSI was faster in everything except the 4k Read/Write QueDepth 32 I am not sure why. I am going to look at some of the option available for the LSI card, they have a battery which I see no need for as I have a Battery Back on the workstation. FastPath which is some sort of accelerator and CacheCade which looks like its used to put a SSD and a conventional Hard Drive in a strip to make it faster. Personally I think its a little crappy that they nickel and dime you for the extras, but I guess that is there business model.

[ikon]

So the LSI 9260-4i was basically the best price point for those specs without using the HighPoint (which I am not convinced are not using the processor of the workstation itself).

I'm with you re: the HighPoint. Nothing against them, but they're what I call Host Bus Adapters rather than full RAID cards. For me, a true RAID card has to do all the processing on board, and have it's own memory for all data transactions. LSI is, AFAIK, still a good brand of these.

[c0op3r]

I'm with you re: the HighPoint. Nothing against them, but they're what I call Host Bus Adapters rather than full RAID cards. For me, a true RAID card has to do all the processing on board, and have it's own memory for all data transactions. LSI is, AFAIK, still a good brand of these.

Oh the one BIG set back to the LSI card (and this mght be for all Processor on board cards) is the my POST times are very long now, it add at least 20 to 30 seconds to startup.

Cooper

[ikon]

Oh the one BIG set back to the LSI card (and this mght be for all Processor on board cards) is the my POST times are very long now, it add at least 20 to 30 seconds to startup.

Cooper

True, depending on the make/model of card the start times will increase by xx seconds, but unavoidable since they do have to load their own 'OS' as it were. Still, the benefit of increased performance from offloading the rest of the system makes it worth IMHO. The biggest downside I feel is the $$$. Full RAID cards ain't cheap baby!

[timekills]

You can offset that by disabling the card's ability to be bootable. You might (unfortunately) also have to set the onboard ports to RAID, even if you don't use them that way, but the onboard RAID BIOS checks are much, much faster than the add-on card's checks. If you disable the card's ability to boot, it won't go through the check during POST.

I could tell you how to do it with a lowly HBA-type adapter like the Highpoints but you're on your own with the LSI's.

[ikon]

You can offset that by disabling the card's ability to be bootable. You might (unfortunately) also have to set the onboard ports to RAID, even if you don't use them that way, but the onboard RAID BIOS checks are much, much faster than the add-on card's checks. If you disable the card's ability to boot, it won't go through the check during POST.

I could tell you how to do it with a lowly HBA-type adapter like the Highpoints but you're on your own with the LSI's.

Would be interesting to see how much time that can shave off.

[c0op3r]

You can offset that by disabling the card's ability to be bootable. You might (unfortunately) also have to set the onboard ports to RAID, even if you don't use them that way, but the onboard RAID BIOS checks are much, much faster than the add-on card's checks. If you disable the card's ability to boot, it won't go through the check during POST.

I could tell you how to do it with a lowly HBA-type adapter like the Highpoints but you're on your own with the LSI's.

Can't do that, as I am booting from the card, the RAID Ø is my OS drive I have a 2TB WD Black for the data drive

[tinkererguy]

Fond of LSI 9260 and 9265 RAID controllers myself. Consider setting ATTO Disk Benchmark to help minimize the effects of the cache on the RAID controllers.

Looking forward to CacheCade 2.0 myself, due 1Q2012 for certain LSI cards:
http://tinkertry.com...gandesxsupport/

[fredp1]

The Dell PERC H700 controller has a feature called “Cut-through IO (CTIO)” that should supposedly increase the throughput for SSD drive arrays.
I believe the Dell is similar to the LSI. (I have a Dell 700 but no SSD)

As per the documentation it says its enabled on an LD (logical drive) by disabling read ahead and enabling write through cache (WT + NORA).
You should see a perfromance increase on a SSD drive.

here are a couple of links
http://sqlblog.com/b...ugh-giggle.aspx
http://marcitland.bl...t-and-ssds.html