dnsbench - DNSSEC & UPDATE test case

Rationale

Think of and implement performance tests for DDNS and DDNS+DNSSEC. It is essential for deployment in some environments (e.g. co.uk).

The best test case is probably to let it load N (feature) differently sized zone files and measure:

1. Zone resign time
2. (small, medium, large) UPDATE on a signed zone

A bar chart would be probably the best with 1 + 3 bars per zone file, first showing a resign time and the rest showing the times for the UPDATEs processing.

The dnsbench would probably be okay platform for that, but it would require a few modifications.

Steps

1. Create a test case folder with three different (unsigned) zones
2. Create a somewhat valid UPDATE scenarios
3. dnsbench should be modified in #192 (closed) to support different test types
4. Create a DDNS test type with measured times for each test as described
5. Update data2www.sh to support multicolumn graphs (it should be done in #192 (closed) already)
6. Publish!

@jkadlec, @jvcelak, @dsalzman - ideas are welcome

The most important information is probably how many updates per second (minute) can we process with / without DNSSEC. Say, for 3 different sizes of updates.

Type of the update operations (add/remove/remove all/etc.) is IMO irrelevant, as the process is more-or-less the same. But in the larger update there may be different kind of operations, just for diversity.

The size of the update won't probably matter right now, because NSEC(3) chain is recreated after each update (will improve until final 1.4, hopefully). But the benchmark with different update sizes will still be useful later on.

Done.

Status changed to closed