Did v11 do anything to improve MariaDB performance? (or os that in v12?)

[bkm]

I remember reading a thread (that I can no longer seem to find) where there was a discussion about tuning MariaDB to reduce the CPU time that it uses and improve it’s efficiency.

I also thought I remembered that part of that thread talked about integrating the improvements into the next version. This has been in the past 5 or 6 months that I remember reading all of this on the forum here.

Can anyone from the development team tell me if the improvements made it into v11?

The reason I ask is that I have a client using v10 and their database is quickly approaching 2gb in size and I would be looking to move them to a new version once I know there is going to be a performance benefit for the database.

Thanks,

BKM

[clarkej]

Perhaps you refer to this [Performance] 1 Sales Invoice Save makes ~700+ DB calls. ~1000+ DB calls on Submit - #50 by szufisher

What PR corresponds to this, to trace and confirm the above refinements to caching and/or reduced calls to the database, is the question?

I don’t think database size affects performance. MySQL can handle terabytes! However individual tables can grow large that may require indexing, but I am no expert.

[bkm]

@clarkej

Perhaps you refer to this [Performance] 1 Sales Invoice Save makes ~700+ DB calls. ~1000+ DB calls on Submit

You are partially correct. There was in fact another thread after this as rmetha indicated should happen before he closed the thread in your post. It is the follow up thread that I was trying to find. In it a developer had indicated he was going to make a PF for it.

BKM

[Tropicalrambler]

I think I might have stumbled upon it:

[Performance] Reducing number of DB calls

Continuing the discussion from [Performance] 1 Sales Invoice Save makes ~700+ DB calls. ~1000+ DB calls on Submit: @szufisher db_insert_multi is a good idea. We just need to see if it fits well in the meta data driven framework and make it generic enough to work with all doctypes having child tables. Worth giving a try. Some child tables like Quotation Item might need tweaking as currently they follow a naming series which will be a challenge for multi insert. @rmehta, @nabinhait what do you …

[bkm]

@Tropicalrambler

I think I might have stumbled upon it:

You are correct! that was it.

And searching the PRs, it appears it was never done. So I guess we may never know if the core devs took it upon themselves to do something similar. I didn’t see anything in the release descriptions.

Thanks for remembering and posting back.

BKM

[Tropicalrambler]

@bkm

omething similar. I didn’t see anything in the release descriptions.

Thanks for remembering and posting back.

BKM

You’re welcome. I think it’s good I’m slightly behind in some things. However it is becoming more relevant as more enterprise customers are desiring performance improvements with ERPNext, and especially given the topic of the conference this year!

I really hope I can scale my startup quickly so we can begin contributing back and taking up pending development topics with my team.

[Sachin_Mane]

Changes related to this were done in a combined code sprint between Frappe and ElasticRun. We introduced two key apis - get_cached_doc() and get_cached_value().

With these changes, there was ~4 times performance improvement in v11.

If you are attending the coming ErpNext conference, we would be presenting the changes we did for gaining up to 200 times !!! improvement (number of concurrent requests processed) over v11 version.

[ashish-greycube]

well done!!