> I don't see the difference between warm-up and cold checkpoints if I don't have cache trace files.
>
> I am guessing warm-up is used to stabilize caches. After warm-up when I checkpoint using write-configuration I don't understand how the cache tags states are saved.
>
We create cache warmup files. This entails running the simulation for
sometime with Ruby configured with large caches. Then we save the Ruby
caches into a file.
To use the cache warmup files, the contents are replayed as single
requests to Ruby. This allows us to use the same large cache warmup files
for multiple cache sizes and configurations. The downside is that it
typically takes 10-40 minutes just to replay our cache warmup files.
Then we continue the simulation.
> Do I have to generate seperate cache-warmup files if I want to use warm-up checkpoints. If so , how ?
>
To dump the cache data into a warmup file, use the following command:
ruby0.dump-cache-data 0 my_file.cache
--Mike
|