[Haskell-cafe] Re: Odd parallel haskell observations (some more numbers)

[Alexander Kotelnikov]

Hello once again.

So I continue to issue thousands of HTTP GET requests to a local apache
an got some ThreadScope pictures and numbers (BTW, I all this happens on
a 4-core machine).

I would point out the following as deserving an explanation:
1. It looks like that none of tests used resources of more than 2 cores.
2. Sometimes there is no activity in any thread of a running program
(get.N4qg_withgaps.eventlog.png, does this mean that process is in a OS
queue for scheduling or something else?)
3. Without RTS's -c or -qg multithreaded run suffers from excessive GC
actions.
4. Even with -c/-qg thread's run looks to be iterrupted too frequent.
5. Provided that 10000 requests in a row can be completed in ~3.4s I
would expect that 4 threads might come close or even under 1s, but 1.9s
was the best result.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: get.N4c.eventlog.png
Type: image/png
Size: 34628 bytes
Desc: not available
Url : http://www.haskell.org/pipermail/haskell-cafe/attachments/20100808/a34bbf44/get.N4c.eventlog-0001.png
-------------- next part --------------
A non-text attachment was scrubbed...
Name: get.N4qg_withgaps.eventlog.png
Type: image/png
Size: 32754 bytes
Desc: not available
Url : http://www.haskell.org/pipermail/haskell-cafe/attachments/20100808/a34bbf44/get.N4qg_withgaps.eventlog-0001.png
-------------- next part --------------
A non-text attachment was scrubbed...
Name: get.N4.eventlog.png
Type: image/png
Size: 38709 bytes
Desc: not available
Url : http://www.haskell.org/pipermail/haskell-cafe/attachments/20100808/a34bbf44/get.N4.eventlog-0001.png
-------------- next part --------------


10000 requests sequentially:
18:02 sacha at loft4633:/tmp 81> for i in `seq 1 5 `;do ./get 1 10000;done
1 3.320623s
1 3.381713s
1 3.394865s
1 3.470693s
1 3.336757s

Using 3 thread and no RTS support
18:20 sacha at loft4633:/tmp 101> for i in `seq 1 5 `;do ./get 3 10000 ;done
3 3.055638s
3 3.226093s
3 3.079507s
3 3.087109s
3 3.279543s

Using 3 threads with RTS support 
17:58 sacha at loft4633:/tmp 80> for i in `seq 1 5 `;do ./get 3 10000 +RTS -N3;done
3 10.971925s
3 9.805483s
3 10.644644s
3 10.734496s
3 9.008394s

Using 3 threads with RTS -c
17:58 sacha at loft4633:/tmp 79> for i in `seq 1 5 `;do ./get 3 10000 +RTS -N3 -c;done
3 2.198638s
3 2.203567s
3 2.190789s
3 2.258964s
3 2.21575s

4 threads with RTS -c
18:22 sacha at loft4633:/tmp 103> for i in `seq 1 5 `;do ./get 4 10000 +RTS -N4 -c;done
4 1.998448s
4 1.970375s
4 2.016234s
4 1.946955s
4 1.983241s

4 thread disabled parallel GC
18:35 sacha at loft4633:/tmp 147> for i in `seq 1 5 `;do ./get 4 10000 +RTS -N4 -qg ;done
4 2.511088s
4 1.939382s
4 1.888141s
4 1.916858s
4 1.913141s

The worst results are achieved for plain -N4
18:43 sacha at loft4633:/tmp 157> for i in `seq 1 5 `;do ./get 4 10000 +RTS -N4;done
4 20.043056s
4 17.02037s
4 16.00271s
4 17.061253s
4 15.914139s