Bug 604046

This request was evaluated by Red Hat Product Management for inclusion in a Red
Hat Enterprise Linux major release.  Product Management has requested further
review of this request by Red Hat Engineering, for potential inclusion in a Red
Hat Enterprise Linux Major release.  This request is not yet committed for
inclusion.

There are 2 performance related profiles in tuned/ktune at the moment:

       throughput-performance
              Sever profile for typical throughput performance tuning. Disables tuned and ktune power saving mechanisms and enables sysctl settings that improve the throughput performance of your disk and  network
              IO and switches to the deadline scheduler.

       latency-performance
              Sever profile for typical latency performance tuning. Disables tuned and ktune power saving mechanisms and enables sysctl settings that improve the latency performance of your network IO.

(see man tuned-adm).

If that is not sufficient please contact the performance team regarding more area specific profiles.

Thanks & regards, Phil

Hi Phil,

Those 2 profiles are in my testing matrix, but I missing one profile described with perf team (large-memory systems running disk and network-intensive applications)

Have a Great day!


Thanks 

Kamil

My understanding is that we agreed to setup ktune consistant with RHEL5;

1) set cpuspeed governors to "performance"
2) set io elevator=deadline 
3) reduce scheduler quantum to match RHEL5
echo 10000000 > /proc/sys/kernel/sched_min_granularity_ns 
echo 15000000 > /proc/sys/kernel/sched_wakeup_granularity_ns
4) enable transparent hugepages - if not set.
echo always >/sys/kernel/mm/redhat_transparent_hugepage/enabled
5) adjust dirty ratio back to 40
echo 40 > /proc/sys/vm/dirty_ratio
 
Not needed from RHEL5:

remove the network size adjustments (kernel does this automatically in 5.4+>
*** not needed ***

Others?

Increase the ridulously small values for;
(ie for 64GB system)
sysctl -w kernel.shmmax=6871947673600
sysctl -w kernel.shmall=429496729600
sysctl -w kernel.sem="25000      3200000   32000      12800"
sysctl -w kernel.pid_max=1280000

I retested this bug on RHEL6.0-20100722.0 with tuned-0.2.11-6.el6.

It can be closed. RESOLVED!

*** Bug 590022 has been marked as a duplicate of this bug. ***

*** Bug 590022 has been marked as a duplicate of this bug. ***

Event posted on 08-05-2010 11:40am EDT by Glen Johnson

------- Comment From mgrf.com 2010-08-05 11:28 EDT-------
(In reply to comment #52)
> These changes made by lwang.
> Bugzilla comment added:
> we have discussed the tunable parameters recommended by
> IBM internally:
>
> kernel.sched_min_granularity_ns = 10000000
> kernel.sched_wakeup_granularity_ns = 15000000
> kernel.sched_latency_ns = 80000000
> kernel.sched_features = 15834234
> kernel.sched_tunable_scaling = 0
>
> and have agreed on to set the following tunning parameters
> in tunded throughput profile: [see bug 604046 comment#10]
>
> echo 10000000 > /proc/sys/kernel/sched_min_granularity_ns
> echo 15000000 > /proc/sys/kernel/sched_wakeup_granularity_ns
>
> since sched_feature parameter has been removed from upstream,
> so we didn't put that in the profile, and tunable_scaling isn't
> in RHEL6. hence this bug is CLOSED DUP to bug 604046.
>

While I don't agree I can understand your opinion about FS here.

But what about:
kernel.sched_latency_ns = 80000000
kernel.sched_tunable_scaling = 0

Those two should definitely also get set in the same profile.

As I can't open the DUP RH Bug and neither want to leave it in
FIXEDAWAITINGTEST nor being able to set it as TESTED/ACCEPTED I'll reopen
this
bug for now.


This event sent from IssueTracker by jkachuck 
 issue 860543

IBM, please open a new bug to track your request and have it considered for RHEL 6.1.

Red Hat Enterprise Linux 6.0 is now available and should resolve
the problem described in this bug report. This report is therefore being closed
with a resolution of CURRENTRELEASE. You may reopen this bug report if the
solution does not work for you.