LSI MegaRAID 9260 new firmware 12.6.0-0017(APP-2.60.03-0829)

[Wintel64]

************************************

LSI Corp. MegaRAID Firmware Release

************************************

Release date: 05/05/10

======================
Supported Controllers:
======================

MegaRAID SAS 9260-4i
MegaRAID SAS 9260-8i
MegaRAID SAS 9260DE-8i
MegaRAID SAS 9261-8i
MegaRAID SAS 9280-4i4e
MegaRAID SAS 9280-8e
MegaRAID SAS 9280DE-8e


Component:
=========
SAS MegaRAID Firmware Release for MegaRAID Controllers
Release date: 05/05/10


Version Numbers:
===============
Current Firmware Package: 12.6.0-0017
Current Package Details: APP-2.60.03-0829_BB-2.02.00.00-0000_BIOS-3.13.00_WEBBIOS-6.0-9-e_5-Rel_PCLI-04.04-00100008_BOOT-01.250.04.219_NVDATA-2.03.0056_2010_03_23



Previous Firmware Package: 12.3.0-0022
Previous Package Details:
APP-2.30.03-0775_BB-2.02.00.00-0000_BIOS-3.09.00_WEBBIOS-5.0-8-e_0-Rel_PCLI-04.02-00400008_BOOT-01.250.04.219_NVDATA-2.03.0044



Bug Fixes and Enhancements:
===========================
Enhancements:
Add support for Advanced Software Options from LSI including:
• LSI™ MegaRAID® CacheCade™
• LSI™ MegaRAID® FastPath™
• LSI™ MegaRAID® Recovery
• LSI™ MegaRAID SafeStore™

Other firmware enhancements:
LSIP200036035 AEN for CacheCade size change
LSIP200036030 SAS Firmware to store the TTY log into Flash ROM and allow OEM tools to capture it.
LSIP200011688 CacheCade Configuration Management
LSIP200008528 Add the structure in the HWR FW to store the Tape device information.


Firmware bug fixes:
LSIP200044702 10M03 FW goes to megamon when powering on/off 2nd enclosure
LSIP200044921 10M03 FW fails to set mfc value TREAT R1E AS R10
LSIP200053959 After migrating from R6 to R5 the VD IO policy canges from DIO to CIO.
LSIP200018573 CacheCade wording
LSIP200018810 After double rebuild on R6/R60 in a single span with 254 medium errors, one of the rebuilt pds is fa
LSIP200043202 Drives missing show zero capacity during post on Wasat and FW 12.3.0-0022
LSIP200043817 FW goes into Montask when trying to import an invalid Foreign configuration
LSIP200043859 CacheCade properties can be changed using MSM
LSIP200043881 Configurations containg 3 VD's can be imported on Low cost iMR controller with out issues during auto-import.
LSIP200044127 CacheCade is cached for a Secure VD
LSIP200044178 VPD is not working with vendorId/deviceId 1000/0079, SubOem 1,subVendorId/subDeviceId 1000/9280
LSIP200044289 Running io program on 63VD and 1CacheCade caused OCR
LSIP200044342 spinup parameters are hard coded & not used from nvdata
LSIP200044467 Data miscompares reported on 256K optimal R6 VD with special IO
LSIP200044615 Too many unexpected sense events for powered down SATA drives
LSIP200044647 Fix missing changes from 09M11
...............................................


http://www.lsi.com/storage_home/prod...-8i/index.html

http://www.lsi.com/DistributionSyste...60.03-0829.zip

[mbreslin]

New firmware/driver/msm installed, no time to do any testing right now. Probably later. I guess most of the new stuff is passive? The only new option I see is to make a bigger cache I guess this is using one drive for a much bigger cache? Will be fun to play with.

[trans am]

THanks man. I was just going to post cuz I just got off the phone with our buddy Garrett K. at LSI.
I have a big surprise for us tonight as I just got 2 of the new ocz vertex 2 50GB ssds. I'm gonna get the FW and drivers going and start benching.
I want to get computurd in here so I can pick his brain a bit. on the last FW and if you were on win7 64 bit we all came to the conclusion that the best drivers were the 64 bit 2008 server drivers.
I am going to go with the new win764 bit drivers to begin and then we can compare the new 2008 drivers later. BRB.

[m.oreilly]

any word on the fastpath software? i've been running the new FW since this morning, would be nice to really see what it does.

[trans am]

is this any good? I just did the 9260 fw rebooted and set up the raid using the lsi webbios and used the auto config. index, dfrag and superfetch are disabled. I didnt touch anything else.








system is:

core i3 540 @ 3.374gb [ 160x21) qpi 3534.7]
maximus III formula 1156 (1607 bios)
5850(8x) / (lsi 9260-4i 8x) split x16 8x8 (2 pcie x16 populated)
2 x 50 gb ocz vertex raid0 / 64kb stripe using lsi 9260-4i sas/raid card [Driver = Windows 7 (64-bit MegaRAID Release 4.4) 4.24.0.64 06-MAY-10 / FW MegaRAID Firmware
(MegaRAID Release 4.4) 12.6.0-0017(APP-2.60.03-0829) 06-MAY-10

4x2gb elpida hyper @ 1600 / 7-7-7-24-72 1t

---------------------------------------------------
heres the lsi config:

[Computurd]

ummm fellas look at yer random 4k

[trans am]

ummm fellas look at yer random 4k

[trans am]

ummm fellas look at yer random 4k

paul thats what I need to know. I never really learned the proper way to bench this kind of setup. I dont want to just bench cache. can you give me a quick rundown or point me to a thread that has the right way to do this? Also what is fastpath and how do we config?

[Computurd]

i dont like as ssd as it forces you to write too damn much. i shy away from write testing, period.
especially 4k random as it does. as ssd rapes yer drives.

run some iometer, etc. we need a sticky on how to configure it.
had a bit of a scare while playing with new firmware. while changing configuration in the MSM i had a breaker flip and lost my array. first time ever on a LSI. however, acronis FTW!!!
took a while of troubleshooting etc thought i had borked my card for a while. so my testing is seriously sidetracked.
my initial impression is a increase of around 30 percent in small file random performance. in comparison to previous numbers the random is leaps and bounds ahead. it approaches 9211 territory at 4k, and my array was seriously degraded when i did that test.

got it reimaged, controller working fine, gonna idle overnight to let her garbage collection clean it up, and when i get off work tommorow, its benchin time! more results then!

[Tiltevros]

Enhancements:
Add support for Advanced Software Options from LSI including:
• LSI™ MegaRAID® CacheCade™
• LSI™ MegaRAID® FastPath™
• LSI™ MegaRAID® Recovery
• LSI™ MegaRAID SafeStore™


to set those settings enable u need a key... later today i will post some results from fast path.
i will have the chance to get some keys unfortunately i wont be able to get to many.
Now its time for all of us to think differend.

[GullLars]

i dont like as ssd as it forces you to write too damn much. i shy away from write testing, period.
especially 4k random as it does. as ssd rapes yer drives.
*snip*
my initial impression is a increase of around 30 percent in small file random performance. in comparison to previous numbers the random is leaps and bounds ahead. it approaches 9211 territory at 4k, and my array was seriously degraded when i did that test.

Double entendres ftw (the alternative being 4KB random write assrapes your SSDs, in case anyone didn't get it)
AS SSD writes 1GB of 4KB random QD 1, and 1GB of 4KB random with 64 parallel threads QD1 each, for a sum of 2GB/4KB=500.000 IO's.
This is extremely and unrealisticly heavy, but fairly benign if only run once in a while and with enough sequential writes in between. Some of the RAID arrays here on the forum can finish such a workload in single digit seoconds.
I wish there were an option in IOmeter to configure test lenght more extensively, like "up to 30 seconds, or up to 100.000 IO's" so setups with insane IOPS performance don't end up doing several millions of IO's in each run. And a test config can contain many runs, my more extensive ones are around 60-80 runs (though it's random read).
It may seem a bit over the top, but i wish it was possible to configure IOmeter much more than it is today :P
I know there are tools out there, but then we're talking proprietary enterprise apps costing 5 figures, and/or restricted who may use them with possibilities of NDAs.


Hopefully, this new firmware lowers latency at the lower queue depths, and softens the spike in latency normally showing around 64-128QD, but with the performance of RAIDs on the 92xx controllers, that high QDs likely won't be common.

[Computurd]

the high QD spiking with is mainly with teh 9211, as most HBA solutions have the same problem at higher QD as well. lack of cache being the primary culprit. also, rigid stripe size and no read ahead options.

[Computurd]

.............

[eva2000]

fastpath sounds sweet

[m.oreilly]

"you need a key..." sounds expensive...i wonder if the upcoming arecas will need a 'key'...

[Tiltevros]

0,08 ms 9260i4

[mbreslin]

I never checked my latency in everest with the old fw, crossed flashed with intel's fw I was at .15, new lsi fw .13, getting better.

[Computurd]

[Computurd]

TILT---very nice numbers there, care to show more

[alfaunits]

Is FastPath essentially force-enabled on 9211 seeing as it can provide higher IOPs with softraid or no?
Wondering how to set up a new build with 9211/C300s.

Enhancements:
Add support for Advanced Software Options from LSI including:
• LSI™ MegaRAID® CacheCade™
• LSI™ MegaRAID® FastPath™
• LSI™ MegaRAID® Recovery
• LSI™ MegaRAID SafeStore™


to set those settings enable u need a key... later today i will post some results from fast path.
i will have the chance to get some keys unfortunately i wont be able to get to many.
Now its time for all of us to think differend.

[mbreslin]

I have to be honest I'm not interested in this at all anymore, buy our hardware then months later we'll charge you again for software that actually lets you reach rated specs. No thanks.

[Computurd]

yes, the more i think about it the more i feel like i am being raped here. they advertised these cards at specs that fastpath enables. that is wrong and possibly false advertising. are they are intentionally downclocking our cards so they can pull more money out of us?

[trans am]

paul. I think the lsi guys will listen if we tell them the truth. I have been in touch with the engineers and I highly recommended you on several phone calls as one of the best in the US even the world as far as complicated storage understanding goes.
I think we should make a thread started by you. (cause they will listen if it comes from someone smarter than their own techs) we make the petition and get the damn soft for everyone. They are foolish to do this. its only going to hurt LSI in the end once the word gets out. I can see areca or someone saying, "hey LSI is charging extra for software they should have available in the 1st place. Lets do the same crap and charge the same for their technology and include the software." what are you going to end up buying?

you guys lets get the petition up and get what we paid for. I think lsi can make a cutdown ver. of this server crap we dont need and as long as we get the cachecade and fastpath I think we will be happy. Those guys at LSI seem very cool as far as recommendations go. I think they will listen to us if we make a big stink about this. whos down? !!

[mbreslin]

I don't care enough to bother, I'm done with pcmv now, it would have to be an insane improvement that would blast me to 40k+. I'm considering going to 3-4 drives on ich so that I can free up the 8 extra pcie lanes for my second graphics card. Sell my 9260 and 9211 and maybe a few c300s or maybe just secure erase and stick them away and wait for intel's next big chipset with a 6gb/sec controller. I had a lot of fun but am also glad to just be using my pc 24/7 not having to mess with anything.

[Computurd]

lemme know if you sell anything

@trans am- they have made no official announcements yet about availability and who exactly has to pay for it, etc. once more details are available we should evaluate the situation then...i think that cachecade will have to be paid for because really that is not a advertised capability of the card. also, it is quite ingenious and im sure that took tons of development on their part.
however, fastpath sounds like another word for firmware update to me. it is just making it faster, correct? but there are unknowns, on the package i have ordered there is some sort of 'software' and what that constitutes i have no idea, so upon further reflection, we need to be sure of the exact contents and what exactly this is before we rally the troops (if we even should)