Mandalika's scratchpad [ Work blog @Oracle | My Music Compositions ]

Old Posts: 09.04  10.04  11.04  12.04  01.05  02.05  03.05  04.05  05.05  06.05  07.05  08.05  09.05  10.05  11.05  12.05  01.06  02.06  03.06  04.06  05.06  06.06  07.06  08.06  09.06  10.06  11.06  12.06  01.07  02.07  03.07  04.07  05.07  06.07  08.07  09.07  10.07  11.07  12.07  01.08  02.08  03.08  04.08  05.08  06.08  07.08  08.08  09.08  10.08  11.08  12.08  01.09  02.09  03.09  04.09  05.09  06.09  07.09  08.09  09.09  10.09  11.09  12.09  01.10  02.10  03.10  04.10  05.10  06.10  07.10  08.10  09.10  10.10  11.10  12.10  01.11  02.11  03.11  04.11  05.11  07.11  08.11  09.11  10.11  11.11  12.11  01.12  02.12  03.12  04.12  05.12  06.12  07.12  08.12  09.12  10.12  11.12  12.12  01.13  02.13  03.13  04.13  05.13  06.13  07.13  08.13  09.13  10.13  11.13  12.13  01.14  02.14  03.14  04.14  05.14  06.14  07.14  09.14  10.14  11.14  12.14  01.15  02.15  03.15  04.15  06.15  09.15  12.15  01.16  03.16  04.16  05.16  06.16  07.16  08.16  09.16  12.16  01.17  02.17  03.17  04.17  06.17  07.17  08.17  09.17  10.17  12.17  01.18  02.18  03.18  04.18  05.18  06.18  07.18  08.18  09.18  11.18  12.18  01.19  02.19  05.19  06.19  08.19  10.19  11.19  05.20  10.20  11.20  12.20  09.21  11.21  12.22 


Thursday, December 02, 2010
 
Instructions to Turn ON/OFF Hardware Prefetch on SPARC64 Systems

The hardware prefetch is ON by default on M-series servers such as M8000/M9000, M4000/M5000, M3000

The following excerpt is from a SPARC64 document:

Hardware speculatively issues the prefetch operation based on the prediction that there is high possibility to access to the following continuous address in the future, if there have been load accesses for a consecutive address.

Although this feature is designed to improve the performance of various workloads, due to the speculative nature, not all workloads may benefit with the default behavior. For example, in our experiments, we noticed 10+% improvement in CPU utilization while running some of the PeopleSoft workloads on M-series hardware with hardware prefetch turned off. Hence irrespective of the application/workload, the recommended approach is to conduct few experiments by running representative customer workloads on target M-series hardware with and without the hardware prefetch turned on.

Instructions to Turn On/Off Hardware Prefetch:

  1. Connect to the system Service Processor (XSCF)

    % ssh -l <userid> <host>
    
  2. Check the current prefetch mode by running the following command at XSCF> prompt

    XSCF> showprefetchmode
    
  3. Find the domain id of all mounted system boards (or skip to next step)

    XSCF> showboards -a
    
  4. Power-off all configured domains

    XSCF> poweroff -d <domainid> [OR]
    XSCF> poweroff -a
    

    From my experience, on larger systems with multiple domains configured, all domains must be powered off before the SP lets changing the prefetch mode. If someone has a correction to this information or better instruction that minimizes disruption, please let me know. I'd be happy to update these instructions.

  5. Wait until the domain(s) are completely powered off. Check the status by running showlogs command

    XSCF> showlogs power
    
  6. Change the prefetch mode to the desired value

    XSCF> setprefetchmode -s [on|off]
    
  7. Verify the prefetch mode

    XSCF> showprefetchmode
    
  8. Finally power-on all configured domains

    XSCF> poweron -d <domainid> [OR]
    XSCF> poweron -a
    
  9. Disconnect from SP, and wait for the OS to boot up

Note to Sun-Oracle customers:

If the default value of hardware prefetch is changed, please make sure to mention this in any service requests, bug reports, etc., that you may file with Oracle Corporation. Unfortunately none of the standard commands on Solaris report the status of hardware prefetch - so, providing this additional piece of information beforehand will help the person who is analyzing/diagnosing the case.

+Original blog post URL



Comments:
Hi Giri,
First, sorry for this indirect form of contact.

Second, I read your profile on LinkedIn, but could not reach you by phone.

I'm conducting retained searches for a Server Side Engineer/Architect, a Front-End Architect, and an Integration expert. I was hoping we could talk, and/or network on these searches.

Please call or email me and I'll give you the complete prospectus.

Matt Loera
415-508-8361 work
415-410-5790 cell
matt@celeritystaff.com
 
Post a Comment



<< Home


2004-2019 

This page is powered by Blogger. Isn't yours?