Tuan-Dat Tran 09b943e41d fix(simulation): correct exponential sampling and cache behavior logic
- Fixed incorrect use of exponential sampling for `next_request` and `next_refresh` timings.
- Adjusted cache eviction logic to properly handle capacity checks.
- Improved simulation output and logging for least accessed objects.
- Corrected CSV export to include index labels for better data clarity.
- Reran experiments

Signed-off-by: Tuan-Dat Tran <tuan-dat.tran@tudattr.dev>
2024-11-28 11:19:04 +01:00

18 lines
881 B
Markdown

# Experiments: No Refresh with variable TTL
Explanation for files in each experiment:
- `details.csv`: Access Count, Hit, Miss, Mu, Lambda and Hit Rate for each object
- `hit_age.csv`: Shows hit rate/average age at time of request for each object.
- `lambda_distribution.pdf`: Lambda Distribution across all objects/discrete
values of the Zipf distribution
- `lambda_vs_access_count.pdf`: Displays the access count against lambda,
expecting a higher lambda to result in a higher access count.
- `objects_in_cache_over_time.pdf`: Amount of cache entries at given time.
- `overall_hit_age.csv`: Cumulative description of `hit_age.csv`
Length of simulation doesn't change much since we're not touching the request
rate across the simulations.
Break condition for the simulation is when the each database object has been
requested at least `ACCESS_COUNT_LIMIT` (i.e. 10) times.