Performance of stakepool

Hi there colleague SPO’s

Got a question: In a normal release strategy, there’s always a test period before releasing the software to PROD. A part of the testing could/should be performance testing. Are there any figures known how many transactions a common stakepool set-up (2 nodes, 1 PB) can handle for let’s say an hour?
The reason I want to know, is that I am selling my stakepool to delegators as “high performance” But what’s high performance? In the past days, my pool was performing about 1000 transactions per hour. Is that high performance? I assume, but I didn’t really know… If it was known, we are also able to set minimum requirements, also to increase and stabilize chain density.

i dont know if that helps you :slight_smile:
In terms of hardware, you should have the following available:
4 GB of RAM
24 GB of hard disk space
a good network connection and about 1 GB of bandwidth per hour
a public IP4 address
i will find also another info i had in the past for the specification for different setup, with more nodes and better performance

1 Like

Good question. Do you really think that the number of Tx your node processes is limited by CPU, RAM, I/O or Bandwidth? If none of those is maxed out, I’d think your node could handle more. There is of course peer connectivity which is determined by the topology updater (or whatever you use). I would think, that you are somewhat at the merci of how many Tx are given to you by your peers.

Yes, some baseline metrics would be good thing to have.

without any Block, i think 1K TXs connections are ok.
you can prepare a grafana server to check the performance of your systems.

i was in the begin with only :
1 vCPUs
1GB / 25GB Disk

and then later i upgraded them with :
2 vCPUs
4GB / 80GB Disk

all of them run without any interruption, with really good performance.
Best Regards,

1 Like

Performance is a major factor in Proof of Work, but is it also here with Proof of Stake? In other words, what advantage would a machine with more CPU/RAM and faster I/O + Network have over one that has just enough to handle the current load that comes in through peers?

I’ve seen numbers in relation to Typed Chronicles i.e. how slot leader selection is done. In order for machine performance to have an impact, there would have to be some notion of competition (i.e. your machine looses against another), is there?

Performance is also important for PoS. It is not that we are all trying to mint that block and GPU power and a fast connection wins as in PoW. That’s is all pre-determined before a new epoch starts. But if you transfer an amount of ADA between 2 or more wallet’s, I can imagine you don’t want that to take 3 days. :wink: In fact, when the people start buying or selling, or voting, or updating pool parameters, the amount of transactions increase, so you want be certain a transaction be handled asap within a specific max time in the network. But there are no specs what the average time for a transaction should take.

1 Like