Performance test: 24 SSD 150 GB Intel on MegaRAID SAS 9361-24i

No RAID, simply jbod mode dstat results:

Reading:
lsscsi | grep SSDSC2BB15| awk '{printf"dd if=%s of=/dev/null bs=16K & \n",$7}'| sh

  0  18  21  60   0   0|6421M    0 | 120B  876B|   0     0 |  36k   53k
  0  18  17  64   0   1|6414M    0 | 120B  346B|   0     0 |  36k   53k
  0  18  14  67   0   1|6413M    0 | 120B  346B|   0     0 |  36k   53k

Writing:
lsscsi | grep SSDSC2BB15| awk '{printf"dd of=%s if=/dev/zero bs=16K & \n",$7}'| sh 

  0   9  11  77   0   3|   0  4503M| 120B  346B|   0     0 |  24k 8055
  0   9  11  77   0   3|   0  4525M| 120B  346B|   0     0 |  24k 8046
  0   9  13  75   0   3|   0  4505M| 120B  346B|   0     0 |  25k 7968

adding more stress:
lsscsi | grep SSDSC2BB15| awk '{printf"dd if=%s of=/dev/null bs=16K & \n",$7}'| sh
lsscsi | grep SSDSC2BB15| awk '{printf"dd of=%s if=/dev/zero bs=16K & \n",$7}'| sh 

  0  19   1  75   0   4|3420M 3401M| 120B  362B|   0     0 |  37k   36k
  0  18   1  77   0   3|3282M 3261M| 180B  362B|   0     0 |  36k   34k
  0  19   2  76   0   3|3297M 3215M| 180B  362B|   0     0 |  35k   33k

Not Bad:)

 

Lustrefs: a big performance hit on lfs find after patch of: CVE-2017-5754 CVE-2017-5753 CVE-2017-5715

Disable: sh set-protection.sh 0

time lfs find /lustre/arm2arm/| wc -l
1454706

real    0m14.941s
user    0m1.633s
sys    0m10.770s

 Enable: sh set-protection.sh 0

time lfs find /lustre/arm2arm/| wc -l
1454706

real    0m10.468s
user    0m0.959s
sys    0m5.521s

Let us hope that the situation will change in the near future....

 And the script set-protection.sh content is:

 #!/bin/bash

[ ! -d  /sys/kernel/debug/x86 ]&& mount -t debugfs debugfs /sys/kernel/debug
echo $1 > /sys/kernel/debug/x86/pti_enabled
echo $1 > /sys/kernel/debug/x86/ibrs_enabled
echo $1 > /sys/kernel/debug/x86/ibpb_enabled

Enable or disable Meltdown and Spectre attack protection on CentOS

After kernel and microcode update one can enable or disable the protection:
Enable:
sh ./set-protection.sh 1
or
Disable:
sh ./set-protection.sh 0 

And the script content is:

 #!/bin/bash

[ ! -d  /sys/kernel/debug/x86 ]&& mount -t debugfs debugfs /sys/kernel/debug
echo $1 > /sys/kernel/debug/x86/pti_enabled
echo $1 > /sys/kernel/debug/x86/ibrs_enabled
echo $1 > /sys/kernel/debug/x86/ibpb_enabled

Q:Why should I disable the protection?
A: If you enable the protection you might accounter the performance degradation.
For some tasks one can enable or disable it:
Synthetic test iperf3 shows network performance over IB degradation about x2:
node01:iperf3 -s
node02:iperf3 -c node01.ib

Enabled:

[ ID] Interval           Transfer     Bandwidth       Retr
[  4]   0.00-10.00  sec  8.70 GBytes  7.48 Gbits/sec    0             sender
[  4]   0.00-10.00  sec  8.70 GBytes  7.47 Gbits/sec                  receiver

Disabled:
[ ID] Interval           Transfer     Bandwidth       Retr
[  4]   0.00-10.00  sec  17.6 GBytes  15.1 Gbits/sec    0             sender
[  4]   0.00-10.00  sec  17.6 GBytes  15.1 Gbits/sec                  receiver






[node01~]#ibstat
CA 'mlx4_0'
    CA type: MT4099
    Number of ports: 1
    Firmware version: 2.40.7000
    Hardware version: 0
    Node GUID: 0xXXXXXXX
    System image GUID: 0xXXXXXXX
    Port 1:
        State: Active
        Physical state: LinkUp
        Rate: 56
        Base lid: 338
        LMC: 0
        SM lid: 3
        Capability mask: 0xXXXXXX
        Port GUID: 0xXXXXXXX
        Link layer: InfiniBand

Fixing X11 forwarding on CentOS7.4

I have an client Centos 6.5 trying to connect to Centos 7.4 with X11 forwarding, but no luck:

Finally I found that somehow the important option is missing in the server side in /etc/ssh/sshd_config :

X11UseLocalhost no

After adding everything is ok now.