laforge has posted comments on this change. ( https://gerrit.osmocom.org/c/osmo-bsc/+/27081 )

Change subject: stats: new trackers for lchan life duration ......................................................................

Patch Set 22:

(1 comment)

Patchset:

PS22: So due to severe performance regressions, I think we will have to re-open the discussion about this code.

It adds one osmo_timer for each lchan, and that timer expires every 100ms. So thinking of a system with 80 BTS and each BTS having a single trx with ~ 16 lchan, we end up with 12800 (!) timers expiring each second! This is certainly the single biggest performance hog in osmo-bsc for any setup larger than the most trivial one.

Whatever we do, we have to do better than that. Like one timer expiring every 100ms iteratinng over the lchans?