Jump to content
RESET Forums (homeservershow.com)

Gen8 SMB performance


wondermike
 Share

Recommended Posts

Hi, I have a Gen8 and I have a network performance problem. The Gigabit Switch is fine, the cables are Cat6, but I only get 30MB/s reading from a VM whereas I get 110MB/s writing to a VM.

HW-Setup:

  •     HP ProLiant Microserver Gen8 F9A40A
  •     LSI 9260-4i + BBU, configured with 1 Virtual Disk as RAID6
  •     Cruidcal M550 with vSphere 5.1 U1 Sept 2013 (ESXi) as boot disk
  •     8 GB RAM (HP)
  •     Intel® Xeon® CPU E3-1220L V2 @ 2.30GHz
  •     4 x 3000GB Seagate Enterprise Value HDD / Terascale HDD ST3000NC002 64MB 3.5" (8.9cm) SATA 6Gb/s
  •     1-Gigabit Switch, CAT6 cables

Test-Setup:

  •     Client: Windows 7
  •     CrystalDiskMark 3.0.3a
  •     SMB-connection to VM (CentOS)

Do you have a clue why this could be and how I can improve the reading performance?
Thanks in advance
Regards

Link to comment
Share on other sites

While I don't know about improving, remember that RAID-6 is dual-parity.  Every time you write, there's a double-parity calculation, and that's going to cut performance.  If your hard drives aren't high-rpm

 

Is the RAID controller up to date on firmware?

 

P.S.  I think Tinkertry.com has had some notes about drivers on that LSI controller under ESXi.  You might want to check there.

Edited by LoneWolf
Link to comment
Share on other sites

The thing I find particularly weird is that the write performance is better than the read. Usually it's the other way round.

Link to comment
Share on other sites

Hi, meanwhile I've done lot's of fine tuning and more tests ... amongst them an installation of win7 in a VM on the same machine. And THAT really showed me, that something is wrong with my client-win7-smb, because with that scenario I'm at 200 MB/s (read) and 294 MB/s (write). The controller has the lates firmware and drivers from LSI.

 

277609d1397427105-hp-proliant-g8-g1610t-

 

Moreover, ftp read speed from my win7-client is around 100 MB/s, so I'm sorry to have brought you on a false trace, that wasn't on purpose. It is really an smb issue on the client. I will reinstall it some time in near future and see.

 

I've also tried to set severeal options in smb.conf on the centos server side, like e.g. TCP_NODELAY, but that only added a few MB/s to the performance (like 1 or 2).

 

Kind regards

Michael

Link to comment
Share on other sites

The thing I find particularly weird is that the write performance is better than the read. Usually it's the other way round.

My raid 10 gen8 performance is usually slightly ahead on the write performance also. I've got the max configured cache memory. I do notice that in blocks smaller than 250mb the write performance is in the 2,500 mb/s range though. I get 155-165 read and 171 write across nearly all synthetic benches seq.

Link to comment
Share on other sites

The thing I find particularly weird is that the write performance is better than the read. Usually it's the other way round.

My raid 10 gen8 performance is usually slightly ahead on the write performance also. I've got the max configured cache memory. I do notice that in blocks smaller than 250mb the write performance is in the 2,500 mb/s range though. I get 155-165 read and 171 write across nearly all synthetic benches seq.

Link to comment
Share on other sites

EDIT: I originally read it wrong, and thought the writes were slower than the reads.

 

I'd still talk to the guy who runs TinkerTry.com ; he's got some experience with ESXi and the LSI controllers.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

×
×
  • Create New...