Messages in this thread | | | Date | Fri, 7 May 1999 12:15:52 +0200 | From | Thierry Danis <> | Subject | Re: Test program: DU good Linux bad |
| |
On Thu, May 06, 1999 at 09:50:10AM -0400, Jim Nance wrote: > Hello All, > I was playing around with an idea for avoiding the "thundering herd" > problem on select/accept, and I think I found a similar problem on read(). > I thought it was worth posting about because neigher Digital Unix nor > Solaris has this problem so that probably means its fixable. > The program creates a pipe and forks off a number of children. Each of > these children tries to read an int from the pipe, decrement it and write it > back into the pipe. When the int goes to zero the program stops and prints > its results. The times I get for 5000000 iterations are: > > Processes Linux(sec) Dig Unix(sec) Solaris(sec) > 2 16.775 23.223 41.774 > 25 45.589 25.022 43.158 > 50 87.721 23.707 43.008 > > Linux 2.2.7 333Mhz PII 500M ram > Digital Unix 4.0D 600Mhz 21164 1.5G ram > Solaris 5.6 >
Here, on an 'old' PC running FreeBSD-2.2.6 :
Running 5000000 inerations with 2 processes took 50.930 seconds Running 5000000 inerations with 25 processes took 46.229 seconds Running 5000000 inerations with 50 processes took 46.567 seconds
> -- > ---------------------------------------------------------------------------- > Jim Nance Avant! Corporation > (919) 941-6655 Do you have sweet iced tea? jim_nance@avanticorp.com > No, but there's sugar on the table.
-- Thierry Danis Poste : 53 53 danis@spmo.sagem.fr
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |