)]}'
{"/PATCHSET_LEVEL":[{"author":{"_account_id":33451,"name":"Yushiro Furukawa","display_name":"Yushiro Furukawa","email":"yushiro.furukawa@lycorp.co.jp","username":"yushiro2"},"change_message_id":"768230415f1bb8948b48fc7a9719b7435032dad4","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"05f56148_1c7715db","updated":"2026-09-07 04:23:51.000000000","message":"LGTM","commit_id":"a421e5c8a70d4847743cd8c8f3c65041e72ade95"},{"author":{"_account_id":39414,"name":"Chris Pickel","email":"christopher.pickel@lycorp.co.jp","username":"cpickel"},"change_message_id":"24a6affcb38ce3f8da5ec61b92a6761e7baf678e","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":2,"id":"235338a5_4af29e45","updated":"2026-09-09 05:21:24.000000000","message":"For a quick test, I brought up a devstack instance and created 2000 zones. Then, I tried listing all zones 10 times:\n\n* Before, it averaged 117ms (102-133ms)\n* After, it averaged 75ms (68-80ms).\n\nThat’s not a dramatic improvement, but of course devstack is the best-case scenario for latency: the API server and DB are on the same machine. In our production deployment, the DB is normally in one region, so queries from another region see a more pronounced difference. It’s on the order of 6-8s to 1.5-2s.\n```\n# before\n$ for i in 0 1 2 3 4 5 6 7 8 9; time curl -H \"X-Auth-Token: $OS_TOKEN\" http://127.0.0.1:60053/v2/zones \u003e/dev/null; done\nreal\t0m0.120s\nreal\t0m0.117s\nreal\t0m0.133s\nreal\t0m0.122s\nreal\t0m0.127s\nreal\t0m0.125s\nreal\t0m0.107s\nreal\t0m0.105s\nreal\t0m0.102s\nreal\t0m0.110s\n\n# after\n$ for i in 0 1 2 3 4 5 6 7 8 9; do time curl -H \"X-Auth-Token: $OS_TOKEN\" http://127.0.0.1:60053/v2/zones \u003e/dev/null; done\nreal\t0m0.076s\nreal\t0m0.074s\nreal\t0m0.079s\nreal\t0m0.072s\nreal\t0m0.068s\nreal\t0m0.080s\nreal\t0m0.078s\nreal\t0m0.077s\nreal\t0m0.070s\nreal\t0m0.071s\n```","commit_id":"aba094a207c733c6c557474747525dde0070c7bf"}]}