Can anybody explain why running 7 users without a mongrel cluster gives
better response than with a mongrel cluster and why for 25 -50 users the
response was about the same?
These are performance figures:
without mongrel cluster (and 4 mongrels)
throughput (request process/min) was about 1175 request/min
The posted data doesn’t appear to support your assertions.
You said:
Can anybody explain why running 7 users without a mongrel cluster gives
better response than with a mongrel cluster and why for 25 -50 users the
response was about the same?
The average response time for 7 users, as you posted it is:
with mongrel cluster:
.0171 seconds
without mongrel cluster:
0.357 seconds
The rest of the data showed mongrel cluster with better response at 25
users
and identical response at 50. The most likely possibility for
identical reponse
at 50 is that the machine is baked, for 1 of 2 reasons, or a
combination of the
two:
Rails is eating the CPU
The DB is eating the CPU
In general, the performance you’re seeing is not very good, about 1/3
request/second (1200/3600) which, in my experience, is typically the
result of a very bad DB performance, potentially caused by missing
indexes.
Another possibility is an extreme number of queries per page tested.
We’ve
seen pages generating many thousands of SQL queries per page, for
instance.