[DNM] Password auth optimization

[SangJunBak]

Release binary password auth non optimized vs optimized:

Release binary password auth non optimized vs optimized vs. no password auth:

Release binary password auth optimized QPS benchmark

Notes:

• Query is SELECT * FROM indexed_view LIMIT 1
• The average QPS looks quite low, but mainly because the initial connection startup is quite slow:

============================================================
Run 5/9: Testing with 16 workers
============================================================

Benchmarking Workload: simple select
------------------------------------------------------------
Queries in workload: 1
Concurrency:         16 workers
Duration:            10 seconds
------------------------------------------------------------

[   3s] Workload: simple select        | Rate:  1102.1 q/s | Total:    3429 | Failed:     0 | Latency (ms): avg=   4.6 min=   2.3 max=  10.6
[   4s] Workload: simple select        | Rate:  3215.5 q/s | Total:    6645 | Failed:     0 | Latency (ms): avg=   4.9 min=   2.3 max=  19.4
[   5s] Workload: simple select        | Rate:  3234.4 q/s | Total:    9880 | Failed:     0 | Latency (ms): avg=   4.9 min=   2.6 max=  11.5
[   6s] Workload: simple select        | Rate:  3329.6 q/s | Total:   13210 | Failed:     0 | Latency (ms): avg=   4.7 min=   2.2 max=   9.7
[   7s] Workload: simple select        | Rate:  3295.4 q/s | Total:   16506 | Failed:     0 | Latency (ms): avg=   5.0 min=   2.6 max=  12.4
[   8s] Workload: simple select        | Rate:  3354.2 q/s | Total:   19861 | Failed:     0 | Latency (ms): avg=   4.7 min=   2.3 max=  11.3
[   9s] Workload: simple select        | Rate:  3133.2 q/s | Total:   22995 | Failed:     0 | Latency (ms): avg=   5.4 min=   2.8 max=  22.3
[  10s] Workload: simple select        | Rate:  3024.4 q/s | Total:   26020 | Failed:     0 | Latency (ms): avg=   5.0 min=   2.5 max=  10.8
[  11s] Workload: simple select        | Rate:  3192.4 q/s | Total:   29213 | Failed:     0 | Latency (ms): avg=   4.9 min=   2.7 max=  10.7

============================================================
Workload Summary: simple select
============================================================
Duration:             12.2 seconds
Concurrency:          16 workers
Total Queries:   32438
Failed Queries:  0
Success Rate:         100.0%
Average Rate:         2654.5 queries/second

Latency Statistics:
  Average:            4.93 ms
  Minimum:            2.05 ms
  Maximum:            27.87 ms

Latency Percentiles (ms):
----------------------------------------
  p50   :     4.39 ms
  p75   :     5.36 ms
  p90   :     7.04 ms
  p95   :     7.91 ms
  p99   :    11.88 ms
  p99.9 :    20.45 ms

Latency Distribution:
------------------------------------------------------------
      2.1 -     2.3 ms:                                              13 (  0.0%)
      2.3 -     2.7 ms:                                              70 (  0.2%)
      2.7 -     3.0 ms: █                                           442 (  1.4%)
      3.0 -     3.5 ms: ██████                                     1650 (  5.1%)
      3.5 -     3.9 ms: ██████████████████                         5063 ( 15.6%)
      3.9 -     4.5 ms: ████████████████████████████████████████  10700 ( 33.0%)
      4.5 -     5.1 ms: ████████████████████                       5485 ( 16.9%)
      5.1 -     5.8 ms: █████████                                  2643 (  8.1%)
      5.8 -     6.6 ms: ████████                                   2254 (  6.9%)
      6.6 -     7.6 ms: ███████                                    1874 (  5.8%)
      7.6 -     8.6 ms: █████                                      1361 (  4.2%)
      8.6 -     9.8 ms: █                                           318 (  1.0%)
      9.8 -    11.2 ms:                                             155 (  0.5%)
     11.2 -    12.7 ms:                                             173 (  0.5%)
     12.7 -    14.5 ms:                                              84 (  0.3%)
     14.5 -    16.5 ms:                                              69 (  0.2%)
     16.5 -    18.8 ms:                                              30 (  0.1%)
     18.8 -    21.5 ms:                                              33 (  0.1%)
     21.5 -    24.5 ms:                                              20 (  0.1%)
     24.5 -    27.9 ms:                                               0 (  0.0%)

as we increase the duration per test run, QPS eventually amortizes into a better one.

We see QPS and latency for the optimized / not is roughly the same:

Motivation

Tips for reviewer

Checklist

• This PR has adequate test coverage / QA involvement has been duly considered. (trigger-ci for additional test/nightly runs)
• This PR has an associated up-to-date design doc, is a design doc (template), or is sufficiently small to not require a design.
• If this PR evolves an existing $T ⇔ Proto$T mapping (possibly in a backwards-incompatible way), then it is tagged with a T-proto label.
• If this PR will require changes to cloud orchestration or tests, there is a companion cloud PR to account for those changes that is tagged with the release-blocker label (example).
• If this PR includes major user-facing behavior changes, I have pinged the relevant PM to schedule a changelog post.