We have a performance issue with Openfire/smack, we have a customIQ returning a payload (16 KB ~) and we are trying to test this, so we sent 5 of them and the round-trip took an average of 4 seconds per IQ ! and this translate to 4 KB/s ! We tried to use PocketCollector OR PocketListener and they both suffer from poor performance. Below is a break down of the issue.
- We are using multithreads to send pocket requests.
- Smack debug shows sent pockets go all out at once.
- Smack debug shows received pockets are returning at sequential rate (once every 5 seconds) !!!
Even if Smack is single threaded, the returning results is almost as if smacks is waiting for the first pocket to be (sent and received) and then sending the second one, this because the liner time of the received pockets ! (please look at the attached img and the times)
We really appreciated any help in this matter, we are at the point of abandoning openfire if this issue can’t be resolved.