Details
-
Improvement
-
Resolution: Fixed
-
Major
-
7.0.0, Cheshire-Cat
-
1
-
KV 2023-4
Description
When running a 100% SET_WITH_META workload (e.g. XDCR initial replication & backup restore), 14% of the worker thread's runtime is spent performing JSON validation:
We should investigate is the current implementation (checkUTF8JSON() based on json_checker.c from JSON.org) is performant compared to other contemporary choices (e.g. rapidJSON, simdjson...)
Attachments
Issue Links
For Gerrit Dashboard: MB-43182 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
183347,5 | MB-43182: Add simdutf as dependency | master | manifest | Status: MERGED | +2 | +1 |
183348,31 | MB-43182: Use SIMD intrinsics to optimise JSON checker | master | platform | Status: ABANDONED | 0 | -1 |
183349,15 | MB-43182: Use SIMD-enabled JSON checker | master | kv_engine | Status: ABANDONED | +2 | -1 |
187444,11 | MB-43182: Build simdutf | master | tlm | Status: ABANDONED | -1 | -1 |
194263,4 | MB-43182: Build simdutf | master | tlm | Status: MERGED | +2 | +1 |
195447,2 | Revert "MB-43182: Use SIMD intrinsics to optimise JSON checker" | master | platform | Status: MERGED | -1 | +1 |
195575,2 | MB-43182: Use SIMD intrinsics to optimise JSON checker | master | platform | Status: MERGED | +2 | +1 |
195639,3 | MB-43182: SIMDUTF_CXX_STANDARD should be a CACHE var | master | tlm | Status: MERGED | +2 | +1 |
195657,2 | MB-43182: Use SIMD-enabled JSON checker | master | kv_engine | Status: MERGED | +2 | +1 |
195658,3 | MB-43182: Use JSON_checker_vectorized by default | master | kv_engine | Status: MERGED | +2 | +1 |
195882,3 | MB-43182: Use SIMD intrinsics to optimise JSON checker | master | platform | Status: MERGED | +2 | +1 |
197443,2 | MB-58678: Revert "MB-43182: Use JSON_checker_vectorized by default" | master | kv_engine | Status: MERGED | +2 | +1 |
198122,6 | MB-43182: Use JSON_checker_vectorized by default | master | kv_engine | Status: MERGED | +2 | +1 |
198627,9 | MB-43182: Use a cbdep for simdutf | master | tlm | Status: MERGED | +2 | +1 |
199776,2 | MB-43182: Break up scan_any_of_128bit into smaller functions | master | platform | Status: MERGED | +2 | +1 |
199777,3 | MB-43182: Fix validation of control characters | master | platform | Status: MERGED | +2 | +1 |