Impact of disabling compacting...

[clebert.suconic]

Say you have a producer sending on message / second on destinationA


And another producer / consumer on destinationB, working as fast as it can.



In the above scenario,

- I was able to process 30K messages / second with compacting configured

- and 25K messages / second with compacting disabled.

Having the compacting disabled, made the journal to create many more files what affected performance.

Also, if I disabled compacting, reloading would take forever. I could fix the OME by cleaning up the list every time the list grew too large, but still would take a lot of time.


I will be in vacation from tomorrow for 1 week, and I will take a look at the cleanup when I'm back.

I believe we need it, but I will provide some input before implementing when I'm back. (It should be an easy one though).

[clebert.suconic]

Just for reference, this is the fast consumer/producer config I used:

num-messages=5000000
num-warmup-messages=20000
message-size=1024
durable=true
transacted=false
batch-size=1000
drain-queue=false
throttle-rate=-1
address=perfAddress
queue-name=perfQueue
host=<My server IP>
port=5445
tcp-buffer=1048576
tcp-no-delay=false
send-window=1048576
pre-ack=true
block-ack=false
block-persistent=false

And this is the slow producer one:

num-messages=200
num-warmup-messages=0
message-size=1024
durable=true
transacted=true
batch-size=10
drain-queue=false
throttle-rate=1
address=perfAddress2
queue-name=perfQueue2
host=<My server IP>
port=5445
tcp-buffer=1048576
tcp-no-delay=false
send-window=1048576
pre-ack=true
block-ack=false
block-persistent=false