Start a Conversation

Unsolved

A

1 Rookie

 • 

3 Posts

127

November 27th, 2024 16:39

ME5024 ControllerA slower IOPS than ControllerB

​Dear Team.​
Help me understand the strange behavior of the new storage system ME5024!

The situation is as follows, IOPS for random write and read on controllerA is much lower than on controllerB, with the same configurations.
Moreover, on any type of RAID arrays (RAID1, 10, 5, 6).
For testing, there is the following test
Array type virtual
PoolA x2 SSD disks: SAMSUNG MZILG7T6HBLAAD3 in RAID1 Volume 200GB
PoolB x2 SSD disks: SAMSUNG MZILG7T6HBLAAD3 in RAID1 Volume 200GB

Maping Volume SAS Connector on one Windows Server 2019 DELL HBA 355e SAS 2 cable: port0 ControllerA and port0 ControllerB

Windows Volume 64k format 

ControllerA

[Read]
  SEQ    1MiB (Q=  8, T= 1):  3136.548 MB/s [   2991.2 IOPS] <  2672.27 us>
  SEQ  128KiB (Q= 32, T= 1):  2109.784 MB/s [  16096.4 IOPS] <  1984.08 us>
  RND    4KiB (Q= 32, T=16):    98.544 MB/s [  24058.6 IOPS] < 21222.46 us>
  RND    4KiB (Q=  1, T= 1):    14.290 MB/s [   3488.8 IOPS] <   286.51 us>

[Write]
  SEQ    1MiB (Q=  8, T= 1):  1956.307 MB/s [   1865.7 IOPS] <  4277.23 us>
  SEQ  128KiB (Q= 32, T= 1):  1571.438 MB/s [  11989.1 IOPS] <  2665.46 us>
  RND    4KiB (Q= 32, T=16):    60.171 MB/s [  14690.2 IOPS] < 34711.56 us>
  RND    4KiB (Q=  1, T= 1):     9.099 MB/s [   2221.4 IOPS] <   449.90 us>

Profile: Default
   Test: 16 GiB (x5) [E: 0% (0/186GiB)]
   Mode: [Admin]
   Time: Measure 5 sec / Interval 5 sec 
   Date: 2024/11/27 19:23:23
     OS: Windows Server 2019 Standard 1809 [10.0 Build 17763] (x64)

ControllerB

[Read]
  SEQ    1MiB (Q=  8, T= 1):  3018.420 MB/s [   2878.6 IOPS] <  2776.17 us>
  SEQ  128KiB (Q= 32, T= 1):  2642.712 MB/s [  20162.3 IOPS] <  1586.41 us>
  RND    4KiB (Q= 32, T=16):   642.105 MB/s [ 156763.9 IOPS] <  3263.58 us>
  RND    4KiB (Q=  1, T= 1):   108.439 MB/s [  26474.4 IOPS] <    37.69 us>

[Write]
  SEQ    1MiB (Q=  8, T= 1):  1956.071 MB/s [   1865.5 IOPS] <  4270.24 us>
  SEQ  128KiB (Q= 32, T= 1):  2005.109 MB/s [  15297.8 IOPS] <  2087.38 us>
  RND    4KiB (Q= 32, T=16):   402.703 MB/s [  98316.2 IOPS] <  5201.16 us>
  RND    4KiB (Q=  1, T= 1):    62.178 MB/s [  15180.2 IOPS] <    65.77 us>

Profile: Default
   Test: 16 GiB (x5) [D: 0% (0/186GiB)]
   Mode: [Admin]
   Time: Measure 5 sec / Interval 5 sec 
   Date: 2024/11/27 19:13:29
     OS: Windows Server 2019 Standard 1809 [10.0 Build 17763] (x64)

Moderator

 • 

4.4K Posts

November 27th, 2024 21:31

Hello,

 

Can you confirm for us if the ME5024 controller firmware and disk firmware are up to date?

 

You may try swap controller slot and see if the issue follows.

 

The images you tried to upload didn't make it. Could you try again?

 

1 Rookie

 • 

3 Posts

December 5th, 2024 13:07

Hello, firmware and disk firmware latest firmware

I found the reason for the decrease in productivity
The thing was that when mapping the LUN, the array was on a controller that was not the one it belonged to.
Thus, it is important to maintain the correspondence when connecting, so that if the array belongs to controllerA and the mapping is done according to controllerA 

ControllerA->DG1->Vol1<->Mapping on server ControllerA

ControllerB->DG2->Vol2<->Mapping on server ControllerB

Only max IOPS 

DG2<-> Mapping ControllerA = SLOW IOPS

No Events found!

Top