Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

Got a mod_perl-written serverside program. Im using shared apache worker mpm memory with one process only, and ~ 60 threads. Im using locks to temporary secure the access to the shared memory, just 0.1s out of 1.7s are locked time though.

The problem is that i get the same speed on 10 sequential requests compared to 10 concurrent requests.

my stress-script fire up n number of heavy URL process-requests with 0.1s delay between each. The number in brackets is the request number.

perl stressTestServer.pl n=1
DONE  ( 0 ) > 1.778257 sek 

perl stressTestServer.pl n=5
DONE  ( 3 ) > 8.71926 sek 
DONE  ( 0 ) > 8.85207 sek 
DONE  ( 1 ) > 8.976061 sek 
DONE  ( 4 ) > 9.119936 sek 
DONE  ( 2 ) > 9.218552 sek 

perl stressTestServer.pl n=10
DONE  ( 9 ) > 8.840539 sek 
DONE  ( 7 ) > 9.130133 sek 
DONE  ( 8 ) > 9.271598 sek 
DONE  ( 6 ) > 9.873936 sek 
DONE  ( 5 ) > 10.02668 sek 
DONE  ( 4 ) > 18.169001 sek 
DONE  ( 3 ) > 18.432776 sek 
DONE  ( 2 ) > 18.617186 sek 
DONE  ( 1 ) > 18.775277 sek 
DONE  ( 0 ) > 18.821972 sek 

is this normal for this setup? this is tested on a dual-core. moving to a quad core, what performance difference can i expect with this setup? I have a hard time believing more cores will help here.. feels like serialized execution could be more efficient than concurrent threads.

tx for input!

share|improve this question
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.