Add a benchmarking wrapper script for BlobDB#9015

Closed

ltamasi wants to merge 28 commits intofacebook:mainfrom ltamasi:blob_benchmark_scripts

+196-1

Commits on Oct 12, 2021

This should do the trick
ltamasi
committed
Reduce target_file_size_base and max_bytes_for_level_base only if blob files are enabled
ltamasi
committed
Add run_blob_bench.sh
ltamasi
committed
Delete output_dir before starting
ltamasi
committed
Put DB under /data, increase DB and memtable size, test more/bigger value sizes
ltamasi
committed
Update duration for read and read/write tests
ltamasi
committed
Enable blob GC iff blob files are enabled (just in case)
ltamasi
committed
Update comment
ltamasi
committed
Make the GC cutoff for bulk loading configurable
ltamasi
committed
Add the GC cutoff for bulk loading to output_dir
ltamasi
committed
Go back to no GC during bulkload's compaction, try different cutoffs during the later tests instead
ltamasi
committed
Put back Merge operators since they are now supported with BlobDB
ltamasi
committed
Revert some parameter tuning that was done to take advantage of the test box
ltamasi
committed
Clean up/refactor run_blob_bench.sh, move some settings over to it from benchmark.sh (relying on the recently added extra parameters functionality, and reverting unnecessary changes along the way)
ltamasi
committed
Update comment and make it possible to configure some more parameters using env variables
ltamasi
committed
Whitespace fix
ltamasi
committed
Update head comment, add help
ltamasi
committed
Adjust default for target file size base, small cleanup
ltamasi
committed
Dump benchmark setup
ltamasi
committed
Ditch report.txt and report2.txt in favor of report.tsv (which is generated by benchmark.sh)
ltamasi
committed
Unset DURATION for the write-only phase
ltamasi
committed
Make OUTPUT_DIR mandatory
ltamasi
committed
Add support for JOB_ID
ltamasi
committed
No need to create output directory, benchmark.sh will take care of it
ltamasi
committed
Add comment
ltamasi
committed
Add blob_garbage_collection_force_threshold
ltamasi
committed
Address some lints
ltamasi
committed
More lint
ltamasi
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add a benchmarking wrapper script for BlobDB#9015

Commits on Oct 12, 2021

This should do the trick

Reduce target_file_size_base and max_bytes_for_level_base only if blob files are enabled

Add run_blob_bench.sh

Delete output_dir before starting

Put DB under /data, increase DB and memtable size, test more/bigger value sizes

Update duration for read and read/write tests

Enable blob GC iff blob files are enabled (just in case)

Update comment

Make the GC cutoff for bulk loading configurable

Add the GC cutoff for bulk loading to output_dir

Go back to no GC during bulkload's compaction, try different cutoffs during the later tests instead

Put back Merge operators since they are now supported with BlobDB

Revert some parameter tuning that was done to take advantage of the test box

Clean up/refactor run_blob_bench.sh, move some settings over to it from benchmark.sh (relying on the recently added extra parameters functionality, and reverting unnecessary changes along the way)

Update comment and make it possible to configure some more parameters using env variables

Whitespace fix

Update head comment, add help

Adjust default for target file size base, small cleanup

Dump benchmark setup

Ditch report.txt and report2.txt in favor of report.tsv (which is generated by benchmark.sh)

Unset DURATION for the write-only phase

Make OUTPUT_DIR mandatory

Add support for JOB_ID

No need to create output directory, benchmark.sh will take care of it

Add comment

Add blob_garbage_collection_force_threshold

Address some lints

More lint