Benchmark 32mb/1.5s blocks

[rach-id]

As part of the effort to reduce block times, we should benchmark whether the current stack supports 32mb/1.5s and which changes are needed to get there.

Acceptance criteria:

• Talis benchmark 32mb/1.5s and see whether we can consistently produce 1.5s blocks
• Experiment with overloading the network with 2x target throughput, i.e., 40mb/s of transactions, and see whether the block times degrade. The experiment should be with small blobs like 500kb, and also large blobs 8mb.
• Experiment how 32mb/1.5s impacts the fibre throughput and whether there are any degradations.

After finishing the experiments, we can create new issues to investigate the parts that need improvements.

[mindstyle85]

hey, is this planned testing on an internal devnet? i assume at some point the same might be needed on Mocha or even Mainnet, so when we get there we should announce since i presume disks might start filling up quickly. At 1.5s im also unsure how the geolocation diversity on mainnet will go

[rach-id]

Good points 👍

is this planned testing on an internal devnet?

Yes, using Talis where we spin up a network in different regions to simulate mainnet geolocation distribution.

when we get there we should announce since i presume disks might start filling up quickly.

Yes, but if it's not utilised, then it shouldn't be that much. So it depends on the usage

For now, this is a scouting issue to see the feasability. Then, we will definitly test it on Arabica and mocha before it lands on mainnet.