Marcus, Howard thank you very much for your help and comments
First of all tell you that the platform works really well and fast. Maybe everything is due to our obsession with optimizing the performance of our moodle database. Maybe performance is not the most appropriate term, and excuse me if I explained it wrong:
In our case, by using AWS we are limited to the number of IOPS we provision to the database. To really be a cost-effective solution we need to reduce the number of IOPS to the minimum possible, but maintaining the latency and response time at acceptable values.
In the past we observed that "at rest" (without activity or users logged in moodle) moodle required around 130-140 IOPS (around 95% were writing operations). The cron is a like a vampire (running every minute), but we really got the improvement using a lock_factory based on memcached, which allowed us to establish our baseline at 3-4 IOPS, something more acceptable.
Working in that sense we are trying to reduce the database load to the minimum possible, offloading to the database of writes that do not provide us with useful information. In this task we find an increase in specific moments of writing operations in the mdl_logstore_standard_log table that we want to reduce.
In any case, I explained myself badly, but the solution goes through as you propose to use a custom logstore plugin. Our development team will work in that direction.
Again, thanks for your help.
First of all tell you that the platform works really well and fast. Maybe everything is due to our obsession with optimizing the performance of our moodle database. Maybe performance is not the most appropriate term, and excuse me if I explained it wrong:
In our case, by using AWS we are limited to the number of IOPS we provision to the database. To really be a cost-effective solution we need to reduce the number of IOPS to the minimum possible, but maintaining the latency and response time at acceptable values.
In the past we observed that "at rest" (without activity or users logged in moodle) moodle required around 130-140 IOPS (around 95% were writing operations). The cron is a like a vampire (running every minute), but we really got the improvement using a lock_factory based on memcached, which allowed us to establish our baseline at 3-4 IOPS, something more acceptable.
Working in that sense we are trying to reduce the database load to the minimum possible, offloading to the database of writes that do not provide us with useful information. In this task we find an increase in specific moments of writing operations in the mdl_logstore_standard_log table that we want to reduce.
In any case, I explained myself badly, but the solution goes through as you propose to use a custom logstore plugin. Our development team will work in that direction.
Again, thanks for your help.