Testnet flood, and how we manage it

Today testnet is receiving a huge performance benchmark. 🥵
Load has peaked at ~500 jetton transfer per second, about 1500 raw TPS.

This is about 30x of current mainnet baseline.

How TONAPI indexer was feeling?

We have two independent setup for testnet indexing. And with this flood we were lucky to see the difference between two setups under the performance load.
Because one of them just got a huge update ✨

The update is how we interact public liteservers and internal archive node to index data.

1) ⚖️ each liteserver performance and lag is measured to sustain nice and stable flow of blocks from the network
2) 🧮 account state is derived from block's merke update to minimize interaction with liteserver (they are usually unable to give account states during such load)

So how did it help. Well old indexer basically stuck for half an hour though it catches up quickly once liteservers are back alive. New indexer was behind the head of the blockchain at most 35 seconds - this only after sustained 30x mainnet load, and it recovers immediately once load gets down a little.

This shows continuing improvement and development of TONAPI to deliver our users most recent data as fast as possible even with future-proof performance.