Fairlock and query monitoring
Monday, 22 October 2018 · 2 messages
MG
Maximilian Gärber22 Oct 2018 13:46 #
Hi,
for some time now, I've switched a production system (running Basex 8.6) to
FAIRLOCK = true. With PARALLEL = 16.
While this helped speed up write operations, there are situations where queries go from sub-second to minute(s) if a few more users are reading/writing at the same time.
The DBs in question are incrementally indexed.
Since single queries are fast, I don't know what the best way of handling these situations would be.
As a last resort I was thinking about adding a cache solution (write-behind) that allows for saving and delays (batches) writes.
But before introducing another component to the system, I'd like to hear what other ideas might exist.
Br,
MaxCG
Christian Grün22 Oct 2018 18:06 #
Hi Max,
Are read-only queries similarly slow if
FAIRLOCK is set to false? If yes, it might help to optimize the incremental database in regular time intervals. If you do that anyway, we could set up a little telko and discuss further possibilities.
Cheers
ChristianQuoted text · Maximilian Gärber · 24 lines
On Mon, Oct 22, 2018 at 1:46 PM Maximilian Gärber <mgaerber at arcor.de> wrote: > > Hi, > for some time now, I've switched a production system (running Basex > 8.6) to FAIRLOCK = true. With PARALLEL = 16. > > While this helped speed up write operations, there are situations > where queries go from sub-second to minute(s) if a few more users are > reading/writing at the same time. > > The DBs in question are incrementally indexed. > > Since single queries are fast, I don't know what the best way of > handling these situations would be. > > As a last resort I was thinking about adding a cache solution > (write-behind) that allows for saving and delays (batches) writes. > > But before introducing another component to the system, I'd like to > hear what other ideas might exist. > > > Br, > Max
Written for BaseX 9.0. BaseX may have changed since: please check the current documentation.
⚡Generated with XQuery