3 ms·
«a lot of random reads» This is not true at all. On the last 10 blocks (495808-495817, ~6000 seconds) there were 22,277 txns and 49,746 txn inputs. That means
by mrb 9y ago
«a lot of random reads»
This is not true at all. On the last 10 blocks (495808-495817, ~6000 seconds) there were 22,277 txns and 49,746 txn inputs. That means a full node needs to look up on average only 8.3 inputs per second in the UTXO database. Assuming 1 I/O per lookup, that's 8 IOPS, which is a trivial workload, even on a non-SSD setup.
I am the author of this blog.zorinaq.com post and I can assure you the real-world I/O workload I see on this $5/month node is negligible.
- makomk 9y agoYou're assuming that you're content to always be one block behind on average, that the node never goes down and has to catch up, and that the block size never increases. The block size is already increasing, and you're suggesting an even bigger increase. Also, cheap VPSes are often pretty terrible in the IOPS department because storage is shared, worse than a normal non-SSD setup.
- mrb 9y agoThe node is never one block behind. Block verification is instantaneous on my VPS. You seem to be under the false impression that when a block is announced, a node has to suddenly verify 2-3k txns in a brief moment. This is not how things work. Txns are verified incrementally as they are added one at a time to the mempool.