wiki:DnsBenchmarkIdeas

DNS Benchmarking Ideas

This is a just a catch-all page of ideas for benchmarking DNS. The official performance and benchmarking tests are listed on the DnsBenchmarks wikipage. Please feel free to add to this list and also link to track ticket numbers for tasks.

  • tracing a file with...
  • DO bit set
  • b10-resolver, as baseline: query for builtin NS and its address (since we don't use real root hints yet)
  • b10-xfrout, AXFR of big zone
  • b10-xfrout: IXFR
  • Startup Speed of in-memory
  • startup speed of loading lots of zones
  • startup speed of being slave for lots of zones
  • graph time from startup to ready to serve. time to start, time to first query (for different amounts of zone data)
  • graph TCP versus UDP performance (need patched queryperf?); also for non-DNS baseline use tcpbench
  • Transfer Efficiency
  • performance with different logging configurations
  • compare memory usage for different periods of time (like queries for 1 minute versus 10 minutes versus 30 minutes)
  • how fast to load zone
  • DDNS Performance
  • query performance while DDNS updates are in progress
  • DNSSSEC Performance, NSEC vs NSEC3 vs normal NXDOMAIN
  • NSEC3 using a larger number of iterations
  • How long does it take to sign a zone?
  • How long to re-sign a zone?
  • Include memory usage in performance measurements.
  • measure memory use from loading different zones sizes
  • benchmark memory usage on size of zones (after loading)
  • tests on systems with limited resources (either physical limitations or simulated/tuned limits). How do the different components behave and perform when hitting limits?, effects of low memory
  • benchmark memory, cpu, disk access, context switches, performance under load
  • ipv6 versus ipv4
  • multiple b10-auth
  • performance while stats are tuned to different gathering periods (1 second, 5, 30, 60, 300 seconds)
  • provide examples of performance for some different NIC cards
  • deep zones (resolver and for auth), long names like: a.b.c.d.e.f.g.h.i.j.k.l.m.n.o.p.q.r.s.t.u.v.w.x.y.z., both sparsely and densely populated zones would make a test case, and also when name spans multiple zones, all hosted on the same server
  • resolver with and without validation
  • zone size memory scaling
  • response latency for different zone sizes
  • reload versus restart times
  • bandwidth usage between master and slaves
  • simple recursive server benchmark tests
  • count bytes on interface (or using capture)
  • larger response sizes (large TXT for example)
  • EDNS tests
  • broken queries resulting in FORMERR / SERVFAIL (need to hack queryperf?)
  • different lengths of labels within single zone
  • wildcard matching cases
  • provide examples of largest sets practical for different hardware and operating systems
  • performance hit due to hot spot cache flush
  • DNAME to CNAME synthesis
Last modified 6 years ago Last modified on Mar 14, 2012, 6:06:46 PM