5 Replies to “BSD Now 281: EPYC Server Battle”

  1. Justin

    Since you mentioned the benchmarks linked above aren’t useful – what benchmarks have you seen that you would consider useful?

  2. Everybody likes benchmarks, but few actually do the work required to replicate real world workloads or explain what the benchmarks mean once they are done. Most of the Phoronix benchmarks don’t make much sense. The typical response is “but they are reproducible”. No effort is given to explain the underlying system differences and setups. Its quite sad because I’ve come across sysadmins who’ve made decisions based on what they read at Phoronix.

  3. A thoughtful Hammer2 versus ZFS versus BTRFS comparison that was run on reasonable server hardware and simulated a years worth snapshots and other file activity would be useful. Similarly useful would be a careful analysis of OpenMP multiprocessor scaling for a variety of parallel numeric computations. Finally, it would be nice to test something to do with network latency and throughput and how those two are affected by firewall rules and simultaneous connections. The question is who is qualified and has the time to do such tests.

  4. I would love to see that in-depth of a study… that’s a solid week’s worth of work, though, and few people have the combination of time, skill, and interest to do that.

Comments are closed.