About the baker and so on


Hi, guys. We are thinking about start baking and have a questions that would help us to understand how to optimize our work. We had some thoughts about creating an autoscale groups or similar:

  1. Can we scale baker horizontally somehow?
  2. Will baker consume all the resources of the VM where it is launched?
  3. Which metric best of all will show us that baker should be scaled?
  4. Should we scale other services (accuser, endorser, node itself) with the baker?
  5. Can we create an load balancer that will manage an incoming traffic? Does it requies any additional settings or it is as simple as TCP load balancing?

  1. Why do you need to scale the baker? It is a single-threaded process and if you double-bake then you loose all rewards for the entire cycle.
  2. No. We run a 2vCPU VM for our baker/endorser/node/accuser with 2GB RAM. The baker hardly uses any CPU. This is proof-of-stake, not proof-of-work. Horsepower is not needed; that’s the whole point behind PoS. This is why you can bake on a raspberryPi.
  3. You don’t need to scale the baker.
  4. No. You could have backups VMs on standby but you should absolutely never run more than 1 baker at a time per private key otherwise you risk double baking.
  5. You don’t need this.


Krixt’s point #2 about baking on a raspbery caught my attention.

So, why kind of equipment is optimal for baking?

I would appreciate link to discussions of hardware and software for baking.


You probably need some ram (more is better), and a large storage unit, cpu is not a factor. RPI3 is on the lower (not optimal) side of the spectrum, it requires a lot of extra care. Keep in mind that the computational resources are not fixed in time, as Tezos evolves so the hardware requirement, and not necessarily it will move upwards.

I suppose you can join the Tezos Riot.im channel and/or send a request to Obsidian system (tezos@obsidian.systems) to join the discussion on Slack