Evaluate dynamic vs. static cache invariant node number #54

New issue

Open

opened 2020-06-25 20:55:54 +02:00 by schmittlauch · 0 comments

schmittlauch commented

2020-06-25 20:55:54 +02:00

Owner

According to the Epichord tech report paper (page 4), the minimum number of nodes per cache slice is calculated depending on the total number of node probes and timed-out node probes.

For the sake of simplicity and avoiding another global counter needing to be synchronised, the first implementation will just use a static number of nodes for that cache invariance.

Once the software runs under real-world conditions, it might be worth evaluating the role of node time-outs for cache lookups and reconsider that decision.

According to the [Epichord tech report](http://publications.csail.mit.edu/lcs/pubs/pdf/MIT-LCS-TR-963.pdf) paper (page 4), the minimum number of nodes per cache slice is calculated depending on the total number of node probes and timed-out node probes. For the sake of simplicity and avoiding another global counter needing to be synchronised, the first implementation will just use a static number of nodes for that cache invariance. Once the software runs under real-world conditions, it might be worth evaluating the role of node time-outs for cache lookups and reconsider that decision.