Performance evaluation report after cache-related improvements in develop (in March 2016)

Continuing the discussion from raw_transactions_cache_size in counterparty-lib:

I’ve done some basic testing with the two commits to develop @rubensayshi made at the linked topic and results are quite exciting:

Just in case for those new to this:

  • X-axis: bitcoin block height
  • Y-axis: seconds counterparty-server took to process a block

I wouldn’t say they’re exactly comparable (system on which 9.54 was running had less RAM; but on the other hand it used SSD, while develop used HDD and more RAM), but even if I judge by “feeling” the changes make counterparty-server feel a lot less resource-intensive compared to v9.54 and before.


  • More uniform block processing times (i.e. no more “database is behind back-end” 2-3 times every hour)
  • Better efficiency (faster and less IO than before)
  • With this one can (probably) run Bitcoin Core and Counterparty on a fairly small system (almost guaranteed if external SSD is used). We’ll know more when the next stress-test / attack hits.