In the previous quarter, the main goal was to migrate the testnet from pulsar-2 to pulsar-3, and that was challenging but went really well in the end.
Thanks again to the Secret Labs team and other testnet participants.
In terms of budget, we ran for an extra month again, cutting support hours, since that demand was minimal, and reduced the number of funded validators to 8.
We can do with 1 or 2 less now and then, as folks drop off for various reasons, but it’s important to maintain diversity, we pick up more issues with different hardware and configuration.
I’ve also reduced my hourly rate by $20, and I’ve reduced the extra nodes we run to support testnet relayers from 3 to 1 node. I’m happy to reassess if there’s demand.
The charter was updated to cleanup some completed tasks, and to add stats to the upcoming quarter. Besides usual maintenance and upgrades, I want to move from Nginx to HAProxy for load balancing.
This will provide better stats out of the box, as well as features like dynamically adding and removing nodes, which should provide an even more robust service.
Can you provide a little bit more of explanation on the expenses, calculation, and motivation?
For example, why should we fund 8 validators and which ones (why not others)?
Why do backup nodes cost so much? Why API servers this much? How many requests are they handling so that we can estimate if its appropriate? Even better would be to have evidence of the real world expenses ourselves!
This would be in line with the recently passed prop#281: we demand that only open-source project may receive funding from the community pool, so why shouldn’t we apply the same to hardware projects like this one?
For example, why should we fund 8 validators and which ones (why not others)?
8 is an arbitrary number reduced from 10, and may be reduced further in future.
There are several reasons for 8 compared to say 3 or 4, such as diversity and stability. Diversity can lead to bugs or edge cases being discovered in testnet rather than mainnet.
And stability from being able to sustain nodes missing blocks under heavy load.
The only criteria is that the validators are also mainnet validators, anyone can join if there’s room, else join the waitlist.
The costs are fairly transparent to anyone who is a validator, typically their expenses are about half of this, for hardware rental, some validators may spend more and others less. We don’t ask for receipts, some people bought hardware and others rent them.
Hardware costs are one part of it, there are operational costs, time to setup, ongoing maintenance, network upgrades, etc
Why do backup nodes cost so much?
Why API servers this much?
API servers and the equivalent backup nodes are archive nodes, and myself and Secret Saturn each run a cluster of 3. This is so if either of us has some kind of problem, eg domain, networking, service provider maintenance etc, users have a choice to use a different endpoint, explorers and wallets etc still work
We have discussed 2 each, which cuts it close but we may try it in future.
How many requests are they handling so that we can estimate if its appropriate?
There’ll be better stats about this coming this quarter.
