Corrected and expanded Ruby protocol benchmarks

[kpumuk]

This pull requests fixes and expands Ruby protocol benchmark, with the main goal is to be able to optimize protocols and measure performance regressions. Before this change (completely broken):

$ ruby protocol_benchmark.rb
Unable to load thrift_native extension. Defaulting to pure Ruby libraries.
protocol_benchmark.rb:35:in '<main>': uninitialized constant Thrift::MemoryBuffer (NameError)

transport1 = Thrift::MemoryBuffer.new
                   ^^^^^^^^^^^^^^
Did you mean?  Thrift::MemoryBufferTransport

Major changes

• Fixed incompatibilities in the structs
• Added pure-ruby option to run the harness
• Added compact and JSON protocols support
• CLI configuration for number of runs for both small and large payloads, as well as scenario selector for targeted runs
• Machine readable --json output.

Last two items allow to hook the benchmark script into autoresearch-style loops to fine-tune the protocol implementation.

I am currently using the benchmark in https://issues.apache.org/jira/browse/THRIFT-5938, and was able to catch a regression in the recursion depth enforcement.

Examples

$ ruby test/rb/benchmarks/protocol_benchmark.rb
Rehearsal ---------------------------------------------------------------------------------
ruby binary write large (1MB) structure once    0.253739   0.000987   0.254726 (  0.254751)
ruby binary read large (1MB) structure once     0.207167   0.005079   0.212246 (  0.212267)
c binary write large (1MB) structure once       0.159660   0.000990   0.160650 (  0.160692)
c binary read large (1MB) structure once        0.122496   0.000000   0.122496 (  0.122512)
ruby compact write large (1MB) structure once   0.156516   0.000990   0.157506 (  0.157522)
ruby compact read large (1MB) structure once    0.115820   0.000995   0.116815 (  0.116818)
ruby json write large (1MB) structure once      1.354133   0.011064   1.365197 (  1.366413)
ruby json read large (1MB) structure once       1.823078   0.007966   1.831044 (  1.831636)
ruby binary write 10000 small structures        0.126928   0.000000   0.126928 (  0.126932)
ruby binary read 10000 small structures         0.099345   0.001098   0.100443 (  0.100460)
c binary write 10000 small structures           0.076388   0.001005   0.077393 (  0.077431)
c binary read 10000 small structures            0.047427   0.000996   0.048423 (  0.048426)
ruby compact write 10000 small structures       0.078293   0.000000   0.078293 (  0.078310)
ruby compact read 10000 small structures        0.051696   0.000000   0.051696 (  0.051717)
ruby json write 10000 small structures          0.607577   0.000970   0.608547 (  0.608567)
ruby json read 10000 small structures           0.870446   0.001981   0.872427 (  0.872490)
header binary write 10000 small structures      0.177455   0.000000   0.177455 (  0.177460)
header binary read 10000 small structures       0.255720   0.001996   0.257716 (  0.257773)
header compact write 10000 small structures     0.137515   0.001004   0.138519 (  0.138550)
header compact read 10000 small structures      0.180783   0.000996   0.181779 (  0.181795)
header zlib write 10000 small structures        0.235636   0.000000   0.235636 (  0.235638)
header zlib read 10000 small structures         0.265005   0.000986   0.265991 (  0.265994)
------------------------------------------------------------------------ total: 7.441926sec

                                                    user     system      total        real
ruby binary write large (1MB) structure once    0.265310   0.002002   0.267312 (  0.267490)
ruby binary read large (1MB) structure once     0.199602   0.003000   0.202602 (  0.204004)
c binary write large (1MB) structure once       0.153209   0.002004   0.155213 (  0.155212)
c binary read large (1MB) structure once        0.112835   0.000000   0.112835 (  0.112835)
ruby compact write large (1MB) structure once   0.155836   0.000000   0.155836 (  0.155836)
ruby compact read large (1MB) structure once    0.120193   0.002997   0.123190 (  0.123204)
ruby json write large (1MB) structure once      1.308506   0.009005   1.317511 (  1.317579)
ruby json read large (1MB) structure once       1.914197   0.006003   1.920200 (  1.920275)
ruby binary write 10000 small structures        0.122585   0.000002   0.122587 (  0.122588)
ruby binary read 10000 small structures         0.084853   0.000000   0.084853 (  0.084850)
c binary write 10000 small structures           0.076188   0.000000   0.076188 (  0.076191)
c binary read 10000 small structures            0.046841   0.000000   0.046841 (  0.046832)
ruby compact write 10000 small structures       0.073694   0.000998   0.074692 (  0.074693)
ruby compact read 10000 small structures        0.048706   0.000000   0.048706 (  0.048716)
ruby json write 10000 small structures          0.619831   0.000000   0.619831 (  0.619882)
ruby json read 10000 small structures           0.864219   0.000989   0.865208 (  0.865246)
header binary write 10000 small structures      0.175500   0.000000   0.175500 (  0.175500)
header binary read 10000 small structures       0.242335   0.000000   0.242335 (  0.242333)
header compact write 10000 small structures     0.136052   0.000000   0.136052 (  0.136074)
header compact read 10000 small structures      0.191703   0.000000   0.191703 (  0.191704)
header zlib write 10000 small structures        0.235157   0.000987   0.236144 (  0.236168)
header zlib read 10000 small structures         0.260917   0.000000   0.260917 (  0.260932)

JSON:

ruby test/rb/benchmarks/protocol_benchmark.rb --large-runs 1 --small-runs 10000 --json --scenarios rb-bin-write-small 2&>/dev/null | jq

Output:

{
  "config": {
    "large_runs": 1,
    "small_runs": 10000,
    "scenarios": [
      "rb-bin-write-small"
    ],
    "skip_native": false,
    "native_available": false
  },
  "results": [
    {
      "id": "rb-bin-write-small",
      "label": "ruby binary write 10000 small structures",
      "benchmark": {
        "user": 0.13548200000000002,
        "system": 0.0007170000000000024,
        "total": 0.13619900000000001,
        "real": 0.13739799987524748
      }
    }
  ]
}

• Did you create an Apache Jira ticket? (Request account here, not required for trivial changes)
• If a ticket exists: Does your pull request title follow the pattern "THRIFT-NNNN: describe my issue"?
• Did you squash your changes to a single commit? (not required, but preferred)
• Did you do your best to avoid breaking changes? If one was needed, did you label the Jira ticket with "Breaking-Change"?
• If your change does not involve any code, include [skip ci] anywhere in the commit message to free up build resources.