Questions and Answers :
Unix/Linux :
RHEL-3-smp, how to get multiple instances running on an smp box?
Message board moderation
Author | Message |
---|---|
Area51 Send message Joined: 3 Apr 99 Posts: 3 Credit: 98,183 RAC: 0 |
Running boinc_4.19_i686-pc-linux under RHAS-3-SMP. System is a dual P3, and I have already added boinc to my /etc/rc.d/init.d so it runs as a daemon. I want to thread one instance through each processor like I did with SETI classic, but the documentation is rather sparse. Someone has got to be doing it already - - - any help is greatly appreciated! Thanks! |
raghu Send message Joined: 5 Mar 01 Posts: 13 Credit: 172,588 RAC: 0 |
You don't need to run multiple instances of BOINC. BOINC automatically configures itself to use all processors. Look under your preferences --> your computers in the BOINC/Seti webpage to check the specs for your computer. |
Area51 Send message Joined: 3 Apr 99 Posts: 3 Credit: 98,183 RAC: 0 |
=> I have my preferences set to allow up to 8 processors, because I have a few 4-way Xeon systems which appear as 8 processors to BOINC. On my windows machines there are multiple instances of SETI_4.09 running - one per processor, as expected. On my RedHat Enterprise systems (2.4.21-27.EL), however, only one instance of the SETI_4.09 module is initializing, even though they are dual processor machines. Also, I notice that the throughput of the Linux client is only about 2/3rds of the Win32 client on identical hardware (dual-boot machine). Any thoughts? Thanks! |
raghu Send message Joined: 5 Mar 01 Posts: 13 Credit: 172,588 RAC: 0 |
Could be due to the kernel: I read somewhere (I cannot seem to find it now) that you required 2.6.10.x to run BOINC successfully. I thought that was the problem in my case: I was running 2.6.9, and I coudn't get seti to run. BOINC would download all the WU's, but I'd get error messages. Seti started crunching the numbers successfully after I upgraded to the 2.6.10.760 kernel. I guess it is worth a try. As for your Windows question, a lot of forum members have reported similar results. I think as a work around, a lot of members are running BOINC on top of Wine to get the best out of their hardware. Probably the Windows version might recognize the number of processors correctly too. |
parkut Send message Joined: 9 Aug 99 Posts: 69 Credit: 9,779,243 RAC: 0 |
> Could be due to the kernel: I read somewhere (I cannot seem to find it now) > that you required 2.6.10.x to run BOINC successfully. I run Linux 2.4.7-10smp AMD MP (dual processor) and Linux 2.4.20-28.7smp with dual P3-500's with no problems. So, maybe not the kernel? Some users have resorted to compiling and running an optimized version of BOINC and Seti |
Area51 Send message Joined: 3 Apr 99 Posts: 3 Credit: 98,183 RAC: 0 |
I'm running RH Enterprise-AS, kernel version is 2.4.21-27.ELsmp. According to the SETI web-site, BOINC is properly recognizing the processor type and quantity...see the page for this system: http://setiweb.ssl.berkeley.edu/show_host_detail.php?hostid=670781 I really doubt it has anything to do with the kernel, any defect in the core of the system would certainly show in other aspects. I may have to try compiling the entire BOINC/SETI package myself when I get a chance... |
ssafco Send message Joined: 27 Aug 04 Posts: 1 Credit: 56,266 RAC: 0 |
> => I have my preferences set to allow up to 8 processors, because I have a > few 4-way Xeon systems which appear as 8 processors to BOINC. > > On my windows machines there are multiple instances of SETI_4.09 running - one > per processor, as expected. On my RedHat Enterprise systems (2.4.21-27.EL), > however, only one instance of the SETI_4.09 module is initializing, even > though they are dual processor machines. > > Also, I notice that the throughput of the Linux client is only about 2/3rds of > the Win32 client on identical hardware (dual-boot machine). Any thoughts? > > Thanks! > > you can utilize as many processors as you like by running multiple instances of boinc in diferent directories. I have a Quad processor xeon and top shows all 4 procs running at 99.x% |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.