Hi all,
we have two Linux server VMs where our BaseX application is running
- Integration - Production
The used database is about 5 GB including a fulltext index and is identical on both systems.
The production VM is more potent in all aspects (CPU, RAM, Disk IO, ...). Dispite all of that we do not get any significantly better performance in the production system.
On top of that we even have XQuery requests that sporadically take very long (> 1 minute) that would normally take only less than one second. Again these long requests only happen in the production environment.
My first assumption was that maybe it is due to additional load on the server from other sources. But that was not the case. We have the spikes even when there is no load at all.
Has anyone similar experiences or any ideas what could be the cause for this behaviour?
Best Regards Johannes
Hello Johannes,
On your production instance, have you experimented with the maximum heap size (-Xmx) and initial heap size (-Xms) options? The delays you're seeing could be caused by garbage collection pauses, either because too much memory is allocated so GC takes a long time, or because too little is allocated so GC runs often to free up more.
It could be happening only in production because your options are different between the two machines, or simply because production is being actively used by many people, so memory gets used up and needs to be cleaned out more often.
Disclaimer: I'm not a Java developer, so my understanding is 100% textbook instead of real-life observation. I personally just try a bunch of different numbers until the application works the way I want it to. If I'm completely wrong about this, someone correct me!
-Tamara
On Wed, Nov 23, 2022 at 2:06 AM Johannes Bauer johannes.bauer@tanner.de wrote:
Hi all,
we have two Linux server VMs where our BaseX application is running
- Integration
- Production
The used database is about 5 GB including a fulltext index and is identical on both systems.
The production VM is more potent in all aspects (CPU, RAM, Disk IO, ...). Dispite all of that we do not get any significantly better performance in the production system.
On top of that we even have XQuery requests that sporadically take very long (> 1 minute) that would normally take only less than one second. Again these long requests only happen in the production environment.
My first assumption was that maybe it is due to additional load on the server from other sources. But that was not the case. We have the spikes even when there is no load at all.
Has anyone similar experiences or any ideas what could be the cause for this behaviour?
Best Regards Johannes
Hi Tamara,
thanks for the input.
The garbage collections could indeed be an issue. The database is quite large and if there are many requests at the same time the memory could be at the limit. I will test with increased heap size and check if the situation improves.
Best Regards Johannes
Hello Johannes,
On your production instance, have you experimented with the maximum heap size (-Xmx) and initial heap size (-Xms) options? The delays you're seeing could be caused by garbage collection pauses, either because too much memory is allocated so GC takes a long time, or because too little is allocated so GC runs often to free up more.
It could be happening only in production because your options are different between the two machines, or simply because production is being actively used by many people, so memory gets used up and needs to be cleaned out more often.
Disclaimer: I'm not a Java developer, so my understanding is 100% textbook instead of real-life observation. I personally just try a bunch of different numbers until the application works the way I want it to. If I'm completely wrong about this, someone correct me!
-Tamara
On Wed, Nov 23, 2022 at 2:06 AM Johannes Bauer johannes.bauer@tanner.de wrote:
Hi all, we have two Linux server VMs where our BaseX application is running - Integration - Production The used database is about 5 GB including a fulltext index and is identical on both systems. The production VM is more potent in all aspects (CPU, RAM, Disk IO, ...). Dispite all of that we do not get any significantly better performance in the production system. On top of that we even have XQuery requests that sporadically take very long (> 1 minute) that would normally take only less than one second. Again these long requests only happen in the production environment. My first assumption was that maybe it is due to additional load on the server from other sources. But that was not the case. We have the spikes even when there is no load at all. Has anyone similar experiences or any ideas what could be the cause for this behaviour? Best Regards Johannes--
Tamara Marnell Program Manager, Systems Orbis Cascade Alliance (orbiscascade.org https://www.orbiscascade.org/) Pronouns: she/her/hers
Hi Johannes
you can also try different garbage collector implementations. Have a look e.g. here https://www.baeldung.com/jvm-garbage-collectors
They behave quite differently.
And you can use VisualVM to debug the collector: https://visualvm.github.io/
regards Stefan
Sent from Outlook for iOShttps://aka.ms/o0ukef ________________________________ From: BaseX-Talk basex-talk-bounces@mailman.uni-konstanz.de on behalf of Johannes Bauer johannes.bauer@tanner.de Sent: Thursday, December 1, 2022 8:24:50 AM To: basex-talk@mailman.uni-konstanz.de basex-talk@mailman.uni-konstanz.de Subject: Re: [basex-talk] Server Performance
Hi Tamara,
thanks for the input.
The garbage collections could indeed be an issue. The database is quite large and if there are many requests at the same time the memory could be at the limit. I will test with increased heap size and check if the situation improves.
Best Regards Johannes
Hello Johannes,
On your production instance, have you experimented with the maximum heap size (-Xmx) and initial heap size (-Xms) options? The delays you're seeing could be caused by garbage collection pauses, either because too much memory is allocated so GC takes a long time, or because too little is allocated so GC runs often to free up more.
It could be happening only in production because your options are different between the two machines, or simply because production is being actively used by many people, so memory gets used up and needs to be cleaned out more often.
Disclaimer: I'm not a Java developer, so my understanding is 100% textbook instead of real-life observation. I personally just try a bunch of different numbers until the application works the way I want it to. If I'm completely wrong about this, someone correct me!
-Tamara
On Wed, Nov 23, 2022 at 2:06 AM Johannes Bauer <johannes.bauer@tanner.demailto:johannes.bauer@tanner.de> wrote: Hi all,
we have two Linux server VMs where our BaseX application is running
- Integration - Production
The used database is about 5 GB including a fulltext index and is identical on both systems.
The production VM is more potent in all aspects (CPU, RAM, Disk IO, ...). Dispite all of that we do not get any significantly better performance in the production system.
On top of that we even have XQuery requests that sporadically take very long (> 1 minute) that would normally take only less than one second. Again these long requests only happen in the production environment.
My first assumption was that maybe it is due to additional load on the server from other sources. But that was not the case. We have the spikes even when there is no load at all.
Has anyone similar experiences or any ideas what could be the cause for this behaviour?
Best Regards Johannes
--
Tamara Marnell Program Manager, Systems Orbis Cascade Alliance (orbiscascade.orghttps://www.orbiscascade.org/) Pronouns: she/her/hers
basex-talk@mailman.uni-konstanz.de
-
Johannes Bauer -
Stefan Ziegler -
Tamara Marnell