summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorGravatar Joey Hess <joeyh@joeyh.name>2016-01-12 13:41:11 -0400
committerGravatar Joey Hess <joeyh@joeyh.name>2016-01-12 13:41:11 -0400
commit979c05445dec82f56d22619f4bf8dbee931e8678 (patch)
tree71fccd92bf1a1a8f7661714ea8d3a5d0d595866f
parent5b8ffc3ee638d4201c2b53225e37cf473a489283 (diff)
devblog
-rw-r--r--doc/devblog/day_356__benchmarking.mdwn11
1 files changed, 11 insertions, 0 deletions
diff --git a/doc/devblog/day_356__benchmarking.mdwn b/doc/devblog/day_356__benchmarking.mdwn
new file mode 100644
index 000000000..05fbce7a5
--- /dev/null
+++ b/doc/devblog/day_356__benchmarking.mdwn
@@ -0,0 +1,11 @@
+Added `git annex benchmark` which uses the excellent Criterion to benchmark
+parts of git-annex. What I'm interested in benchmarking right now is the
+sqlite database that is used to manage v6 unlocked files, but having a
+built-in benchmark will probably have other uses later.
+
+The benchmark results were pretty good; queries from the database are
+quite fast (60 microseconds warm cache) and scale well as the size increases.
+I did find one scalability issue, which was fixed by adding another index
+to the database. The kind of schema change that it's easy to make now, but
+that would be a painful transition if it had to be done once this was in wide
+use.