[bisq-network/bisq] High CPU Consumed (#4370)

[chimp1984]

> > The DAO sync is running on userthread but the BSQ tx verification itself is super fast and the the problem. I think the problme is some client code reacting on state changes and doing expensive operations (e.g. lookups). There have been done some improvements already but I think there is more needed, but not a trivial task...
> 
> I suspect its the concurrent seednode connections via tor, that happen during "Synchronizing DAO". Last time I checked, tor connections go via some sort of io-blocking bottleneck. I noticed similar effects with the connections to the bitcoin nodes - connecting to them via clearnet caused less CPU overhead and was generally faster. Similar effect if I chose to use 1 tor bitcoin node, vs multiple (1 node was faster, smaller CPU spike). This was a few months ago, so not sure if it's still the case. But it might indicate that highly-active concurrent tor connections may cause/contribute to the CPU spike due to how multi-threading/IO is managed there.

This is expected as Tor is slower and as the connections get clogged by multiple requests. 
We have those connections streams:

- BitcoinJ block download (starts early and can finish any time after startup and app init)
- Initial data requests from multiple seeds (usually 2 if one does not respond more). After first response we start DAO block request, but other responses might come then parallel to the DAO block response
- Get BSQ blocks

So in the worst case we do 3 data requests at the same time: Btc blocks, Bisq data, BSQ blocks. Additional there are price feed requests (I think those are quite lite compared to the rest).

The initial data requests have been improved by 50% in 1.4.2 and in 1.5.2. we get anoute 25% less data (AccountAgeWitness data will be handled via HistoricalDataStore). So that will help overall to reduce bandwith requirements.


-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/bisq-network/bisq/issues/4370#issuecomment-744052009
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.bisq.network/pipermail/bisq-github/attachments/20201213/d1b14e87/attachment.htm>