Fix CachingPostSerializer defining associations twice; add FragmentCaching benchmaking

[bf4]

Summary

You can see that with caching on the caching serializer is slightly slower than the non-caching, and the fragment caching serializer is much slower

serializer_type	ips	objects
caching	190.252	8259
non-caching	211.199	7295
fragment	53.773	22761

with caching off (i.e. disabled in the controller) the serializer behavior is more similar

serializer_type	ips	objects
caching	163.787	8290
non-caching	163.678	7295
fragment	45.753	22761

(these are run on my mbp while it's mostly idle, so compare the ips rather than look at absolutes, and then with a little fuzziness. I should include sterr in the future)

Also note that the objects being serialized are all in-memory. There are no queries being made.

Related

• Caching doesn't improve performance #1586
• Demonstrate presence of bug mentioned in #1686 #1690
  • Memoize resource relationships #1686
  • Investigate why #1686 was such a big performance boot, that there are dup'ed relationships being yielded #1688

Source

Running

SUMMARIZE=true bin/bench_regression 7485c84 335869e --repeat-count=5 --env='BENCH_STRESS=true' --pattern='bm_caching'

Details

var this-pr = {
  "commit_hash": "335869e",
  "version": "0.10.0.rc5",
  "rails_version": "4.2.6",
  "benchmark_run[environment]": "2.2.2p95",
  "runs": [
    {
      "benchmark_type[category]": "caching on: caching serializers: gc off",
      "benchmark_run[result][iterations_per_second]": 190.252,
      "benchmark_run[result][total_allocated_objects_per_iteration]": 8259
    },
    {
      "benchmark_type[category]": "caching on: fragment caching serializers: gc off",
      "benchmark_run[result][iterations_per_second]": 53.773,
      "benchmark_run[result][total_allocated_objects_per_iteration]": 22761
    },
    {
      "benchmark_type[category]": "caching on: non-caching serializers: gc off",
      "benchmark_run[result][iterations_per_second]": 211.199,
      "benchmark_run[result][total_allocated_objects_per_iteration]": 7295
    },
    {
      "benchmark_type[category]": "caching off: caching serializers: gc off",
      "benchmark_run[result][iterations_per_second]": 163.787,
      "benchmark_run[result][total_allocated_objects_per_iteration]": 8290
    },
    {
      "benchmark_type[category]": "caching off: fragment caching serializers: gc off",
      "benchmark_run[result][iterations_per_second]": 45.753,
      "benchmark_run[result][total_allocated_objects_per_iteration]": 22761
    },
    {
      "benchmark_type[category]": "caching off: non-caching serializers: gc off",
      "benchmark_run[result][iterations_per_second]": 163.678,
      "benchmark_run[result][total_allocated_objects_per_iteration]": 7295
    }
  ]
}

var master = {
  "commit_hash": "7485c84",
  "version": "0.10.0.rc5",
  "rails_version": "4.2.6",
  "benchmark_run[environment]": "2.2.2p95",
  "runs": [
    {
      "benchmark_type[category]": "caching on: caching serializers: gc off",
      "benchmark_run[result][iterations_per_second]": 274.5,
      "benchmark_run[result][total_allocated_objects_per_iteration]": 6005
    },
    {
      "benchmark_type[category]": "caching on: non-caching serializers: gc off",
      "benchmark_run[result][iterations_per_second]": 294.831,
      "benchmark_run[result][total_allocated_objects_per_iteration]": 5357
    },
    {
      "benchmark_type[category]": "caching off: caching serializers: gc off",
      "benchmark_run[result][iterations_per_second]": 233.751,
      "benchmark_run[result][total_allocated_objects_per_iteration]": 6005
    },
    {
      "benchmark_type[category]": "caching off: non-caching serializers: gc off",
      "benchmark_run[result][iterations_per_second]": 251.077,
      "benchmark_run[result][total_allocated_objects_per_iteration]": 5357
    }
  ]
}

[beauby]

So nothing has changed?

[bf4]

@beauby no, I accidentally ran it against the same hash twice.. one moment

[bf4]

@beauby so, the benchmark still shows that, for this test case, caching doesn't help, but at least it doesn't hurt quite as much.

[bf4]

@beauby updated description

[beauby]

Cool, which adapter is used in the benchmark?

[bf4]

Json since it's the most historically consistent one (for regression testing)

B mobile phone

On Apr 20, 2016, at 12:03 PM, Lucas Hosseini notifications@github.com wrote:

Cool, which adapter is used in the benchmark?

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub