Merge branch 'benchmarks' into 'main'

add page for benchmark results

Closes #11

See merge request obnam/obnam.org!58

2 files changed, 35 insertions, 0 deletions

diff --git a/benchmarks.mdwn b/benchmarks.mdwn
new file mode 100644
index 0000000..0438b9b
--- /dev/null
+++ b/benchmarks.mdwn
@@ -0,0 +1,34 @@
+[[!meta title="Benchmark results"]]
+
+This page collects results of benchmarks of Obnam. It's not very
+systematic yet.
+
+# Large file backup
+
+## Goal
+
+> Given a freshly created Obnam server with an empty chunk store, how
+> long does it take for the Obnam client to back up a 1 GiB sparse
+> file?
+
+## Steps
+
+* create a 1 GiB sparse file
+  - in shell: `truncate --size 1G filename`
+* create an empty chunk store and start the Obnam server
+* run the Obnam client, backing up the sparse file
+
+Measure the duration of the backup command. That number is the
+benchmark result.
+
+## 2021-10-10 run
+
+Lars ran this on a Thinkpad T480 laptop with a SATA SSD.
+
+Ran with and without `cargo flamegraph`. Time without is 7.8 seconds.
+The flamegraph shows that almost all the time is spent on sha256
+checksum computation.
+
+This is the baseline result, because it's the first result.
+The steps should be automated; the source tree has a `bench.sh` shell
+script that is a start.
diff --git a/index.mdwn b/index.mdwn
index 9d8306b..95c18a7 100644
--- a/index.mdwn
+++ b/index.mdwn
@@ -39,6 +39,7 @@ to smart server via HTTP; server never sees cleartext.
 * Acceptance criteria are documented using [Subplot][], and are on
   [doc.obnam.org](https://doc.obnam.org/).
 * [[Contributing]] to Obnam
+* [[Benchmark results|benchmarks]]
 * [Obnam blog posts by Lars](https://blog.liw.fi/tag/obnam/)
 * Other backup software:
   * [Borg](https://www.borgbackup.org/)