• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

ESXi 4.1 SCSI DAS array performance issues.

S

simm0

n00b
Joined
Nov 8, 2009
Messages
32
I asked about this over in the storage section and got nothing, hope to have better luck here.

The array is an Eonstor A16U-G2421 SCSI-SATA DAS system. I never expected stellar performace with this config but performance is unacceptable (to be polite). IOmeter results summarised below.

IOps: 6094.557039

Read IOps:3050.704829

Write IOps:3043.852211

MBps:23.806863

Read MBps:11.916816

Write MBps:11.890048


IOmeter tests performed from VM hosting the storage using lopoetve's "vmware-iometer-testconfig.icf"
1TB file copy (writing to array) across Gigabit network takes > 4 days. Yes days.

Configuration

Array Side
  • 12 x Hitachi Ultrastar A7K1000 HUA721010KLA330 SATA 3GHz 1TB disks.
  • RAID 5 + Hot Spare, 128 KB stripe size.
  • 5 x 1907218 MB partitions on the Logical Volume.
  • Each partition = 1 LUN.
  • Channel configuration at the link below. The problem appears to be that the "Current Transfer Clock = 40MHz" vs "Default Transfer Clock = 160 MHz". So the array is performing at only 25% capacity.
A16uG2421_ChannelConfig.jpg



Server Side
  • Server running ESXi 4.1.
  • Intel SE7520BD2 Motherboard.
  • Adaptec 39320A-R 64-bit 133 MHz PCI-X, dual-channel Ultra320 SCSI HBA.
  • Using embedded VMware AACRAID driver.
  • SCSI HBA Config as per Eonstor instructions:
    ○ Disabled "QAS" (function not supported by RAID controller).
    ○ Enabled "BIOS Multiple LUN Support".
    ○ Disabled "Domain Validation".
    ○ Disabled "Host RAID".
ESXi
  • 5 x 19.7218 MB LUNs = 5 x 1.82TB datastores.
  • 8 MB block size.
VM
  • Buslogic Parallel SCSI Controller.
  • Windows XP SP3 32 bit.
  • 1 x 1.82 TB VMDK per datastore.
  • All 5 VMDKs striped (RAID 0) under Windows.
  • 4KB cluster size.

I hope it's just a config error somewhere on my part, the array has performed well in the past but not in this setup.

Cheers
Simon
 
olay itshay. I'll look at this as soon as I"m in the office. Can you get me a vm-support from the host?
 
Thanks lopoetve.

I have gathered the support files (VIclient --> File --> Export --> Export System Logs).
It's 200MB, how can I get it to you? I can put it on our FTP if you like.
 
PM with FTP details sent.
Thanks and goodnight.
 
Last edited:
I haven't managed to get to the bottom of this yet, the array is finally out of production so I can sort this out.

Below is a clip from resxtop +u (storage (disk) device resource utilization screen).

resxtop_scsi.png


I'm guessing a DQLEN of 4 and a QUED of 121 is not so good. Would increasing the Device Queue Length help to resolve these performance issues? How would I go about doing so?
 
The firmware version is 3.47R.02 which I think is up to date for this device.

Cheers
Simon
 
Hi there,

I discovered this thread trying to get a Infortrend A16u-G1410 working. After finally getting the store to be discoverable I seem to have the same issue.

Running esx 4.1 Adaptec 29320ALP, connected scsi. I've got the same clock speed 40mhz and have the same settings as you in raidwatch.
Latest firmware so am watching this thread.
 
I have tried replacing the scsi cable with a new one, still stuck on 40mhz.
Infortrend can't seem to answer the question either.
 
Infortrend support blows, you may have better luck with your vendor.
I have attached a second SCSI cable to the second channel on the Infortrend and to our other ESXi host. The "Current Transfer Clock" on the second channel is 160 MHz but the actual performance is as bad or worse.

I really think the issue is to do with the Device Queue Length being set to 4 but cannot find information on setting the device queue length in ESXi.
 
Just an update from my end. I will be fixing the problem by getting rid of the infortrend completely. At this stage my download speed from the internet is faster than the directly attached infortrend.

Plus as mentioned their support is completely hopeless.
 
You must log in or register to reply here.
Back
Top