Today, Alexsander talks about what we’ve done to improve MAAS performance, and how we’re making performance improvements a part our regular work.
https://drive.google.com/file/d/1RdDcDPOhmgNNbLp0jFUmRrDi27O3iiXx/view?usp=sharing
Please be sure to include your comments and questions below.
Your dataset is too small…
Also, I had a very difficult time understanding Alexsander.
Where’s the caching? it’s time to start leveraging memcache (or similar) for volatile data caching so you’re not limited by DB speed. i.e. my 4658 machines take several MINUTES to load (granted I’m still on 2.9.2), that should be able to be cached so it loads in milliseconds (4+ orders of magnitude faster) on followup requests, I can ponder that many areas are ripe for cache-ability, which benefits both the UI and the API.
Have each region controller run memcached, and have the config leverage every region controller in the config (for HA scenarios), easy, peasy, bing bang done (cache side), then the API would just need a DB class overload that checks cache FIRST on reads, if a miss, fetch DB, push into cache, then return to called. a write can either be sloppy and just invalidate the cache entry, or wise and update it when updating the DB.
