Query-path optimizations: benchmark old vs. new query-path backends on benchmark config

What

Following the optimizations this epic provides for, it's worthwhile to step back and view our progress.

The following suffices to this end:
Benchmarking the "before":

create a temporary branch that disincorporates the changes made within this epic (e.g. revert everything from XDR views)
- ensure this branch has the changes made by Integrate changes made to stellar-rpc-blaster #933
create a PR for that branch
add that branch to the load test coordinator and push to trigger a perf-eval run

Then, to benchmark the "after," simply repeat those steps on a branch that includes all of the changes above.

These results should be shared in a broader context (e.g. with the performance pod)

Acceptance criteria

results gathered for RPC performance on both query-path backends
results shared

Here is an artifact summarizing the results of these comparisons: https://claude.ai/artifact/NN71FNxufqGAFznva9TEN4?sk=DO4ocegh_TDV_9cmUm8LFg
Note that, while this artifact was created before all runs were merged to main, the variations between this data and the data below are attributable to noise.

Here are the results of the tests, copied raw from the most recent pre-optimization runs and the most recent post-optimization runs. Again, there were some minor fluctuation between runs, but the trends shown in the data provide a clear picture of the speedups.

All data gathered on an AWS c5.2xlarge with 500Gb gp3 EBS storage, on an RPC node serving a 7-day retention window.

Pre-Optimization:

✅ Endpoint load test — verdict: ok

🎯 Endpoint load test — `14de10961916`

Serial blast per endpoint (ramp-up 1m, duration 3m, error kill switch 75%, blaster aadc1a17595f) against the backfilled RPC (ledgers [64476654, 64597613], handoff wait 1337s).

Endpoint	Target RPS	Requests	Errors	p50 (ms)	p95 (ms)	p99 (ms)	p99.9 (ms)
getEvents	75	11225	9 (0.1%)	1.2	14.1	141.3	1586.2
getFeeStats	250	37491	0 (0.0%)	0.8	1.0	127.4	193.3
getHealth	250	37492	0 (0.0%)	0.8	1.0	118.8	204.0
getLatestLedger	15	2243	0 (0.0%)	78.3	152.7	179.7	236.5
getLedgers (limit=5)	3	440	0 (0.0%)	646.7	2094.1	5562.4	8007.7
getNetwork	100	14967	0 (0.0%)	19.6	57.9	114.2	175.5
getTransaction	75	11223	0 (0.0%)	35.3	67.2	128.3	221.1
getTransactions (limit=200)	10	1473	0 (0.0%)	126.9	387.1	460.8	500.7
getVersionInfo	100	14911	0 (0.0%)	18.3	69.8	155.0	211.2

getEvents results extended

Endpoint	Target RPS	Requests	Errors	p50 (ms)	p95 (ms)	p99 (ms)	p99.9 (ms)
getEvents/catch-up	3	449	0 (0.0%)	6.8	22.5	94.6	279.8
getEvents/deep-pager	17.25	2603	1 (0.0%)	1.1	54.3	186.0	1157.1
getEvents/deep-scan	2.25	344	2 (0.6%)	1.1	12.7	103.0	10002.4
getEvents/firehose	1.5	203	0 (0.0%)	1.2	6.7	75.5	89.7
getEvents/head-poll	36	5418	0 (0.0%)	1.2	6.7	117.2	253.3
getEvents/tail-poll	9	1301	6 (0.5%)	1.1	2.4	749.1	10002.4
getEvents/transfer-watcher	6	907	0 (0.0%)	6.9	14.0	104.3	267.5

✅ Go endpoint benchmarks — verdict: ok

Baseline v28.0.1 (273f19e4fcb1) vs candidate 14de10961916 — -benchmem -count=10, both refs sequentially on one box; rpcv2 excluded.

benchstat: baseline vs candidate

goos: linux
goarch: amd64
pkg: github.com/stellar/stellar-rpc/cmd/stellar-rpc/internal/db
cpu: Intel(R) Xeon(R) Platinum 8124M CPU @ 3.00GHz
                  │ baseline.txt │
                  │    sec/op    │
GetLedgerRange-8     953.1n ± 1%
BatchGetLedgers-8    755.9µ ± 3%
geomean              26.84µ

                  │ baseline.txt │
                  │     B/op     │
GetLedgerRange-8      16.00 ± 0%
BatchGetLedgers-8   520.8Ki ± 0%
geomean             2.852Ki

                  │ baseline.txt │
                  │  allocs/op   │
GetLedgerRange-8      4.000 ± 0%
BatchGetLedgers-8    2.314k ± 0%
geomean               96.21

pkg: github.com/stellar/stellar-rpc/cmd/stellar-rpc/internal/feewindow
                                                           │ baseline.txt │
                                                           │    sec/op    │
ComputeFeeDistribution/computeFeeDistribution-8               9.130µ ± 0%
ComputeFeeDistribution/alternativeComputeFeeDistribution-8    225.6µ ± 3%
geomean                                                       45.38µ

                                                           │  baseline.txt  │
                                                           │      B/op      │
ComputeFeeDistribution/computeFeeDistribution-8                0.000 ± 0%
ComputeFeeDistribution/alternativeComputeFeeDistribution-8   565.4Ki ± 0%
geomean                                                                   ¹
¹ summaries must be >0 to compute geomean

                                                           │ baseline.txt │
                                                           │  allocs/op   │
ComputeFeeDistribution/computeFeeDistribution-8              0.000 ± 0%
ComputeFeeDistribution/alternativeComputeFeeDistribution-8   30.00 ± 0%
geomean                                                                 ¹
¹ summaries must be >0 to compute geomean

pkg: github.com/stellar/stellar-rpc/cmd/stellar-rpc/internal/methods
                               │ baseline.txt  │             candidate.txt              │
                               │    sec/op     │    sec/op      vs base                 │
GetEventsTopicFilters-8          2.582m ± 2%     2.608m ± 1%         ~ (p=0.105 n=10)
GetEvents-8                      162.0µ ±  ∞ ¹   163.5µ ±  ∞ ¹       ~ (p=1.000 n=1)  ²
GetLedgers-8                     1.397m ±  ∞ ¹   1.132m ±  ∞ ¹       ~ (p=1.000 n=1)  ²
JSONTransactions/JSON_format-8   7.762µ ± 0%     7.701µ ± 0%    -0.79% (p=0.000 n=10)
JSONTransactions/XDR_format-8    2.574µ ± 0%     2.617µ ± 1%    +1.69% (p=0.000 n=10)
GetProtocolVersion-8             43.70µ ±  ∞ ¹   45.11µ ±  ∞ ¹       ~ (p=1.000 n=1)  ²
geomean                          89.38µ          87.17µ         -2.47%
¹ need >= 6 samples for confidence interval at level 0.95
² need >= 4 samples to detect a difference at alpha level 0.05

                               │  baseline.txt  │              candidate.txt              │
                               │      B/op      │      B/op       vs base                 │
GetEventsTopicFilters-8          190.8Ki ± 0%     191.2Ki ± 0%         ~ (p=0.063 n=10)
GetEvents-8                      43.16Ki ±  ∞ ¹   43.16Ki ±  ∞ ¹       ~ (p=1.000 n=1)  ²
GetLedgers-8                     1.369Mi ±  ∞ ¹   1.043Mi ±  ∞ ¹       ~ (p=1.000 n=1)  ²
JSONTransactions/JSON_format-8   2.000Ki ± 0%     2.000Ki ± 0%         ~ (p=1.000 n=10) ³
JSONTransactions/XDR_format-8    1.336Ki ± 0%     1.336Ki ± 0%         ~ (p=1.000 n=10) ³
GetProtocolVersion-8             7.908Ki ±  ∞ ¹   7.456Ki ±  ∞ ¹       ~ (p=1.000 n=1)  ²
geomean                          25.00Ki          23.66Ki         -5.33%
¹ need >= 6 samples for confidence interval at level 0.95
² need >= 4 samples to detect a difference at alpha level 0.05
³ all samples are equal

                               │ baseline.txt  │             candidate.txt              │
                               │   allocs/op   │   allocs/op    vs base                 │
GetEventsTopicFilters-8          1.524k ± 0%     1.528k ± 0%         ~ (p=0.054 n=10)
GetEvents-8                       621.0 ±  ∞ ¹    621.0 ±  ∞ ¹       ~ (p=1.000 n=1)  ²
GetLedgers-8                     4.723k ±  ∞ ¹   4.122k ±  ∞ ¹       ~ (p=1.000 n=1)  ³
JSONTransactions/JSON_format-8    25.00 ± 0%      25.00 ± 0%         ~ (p=1.000 n=10) ²
JSONTransactions/XDR_format-8     23.00 ± 0%      23.00 ± 0%         ~ (p=1.000 n=10) ²
GetProtocolVersion-8              122.0 ±  ∞ ¹    120.0 ±  ∞ ¹       ~ (p=1.000 n=1)  ³
geomean                           260.7           254.2         -2.47%
¹ need >= 6 samples for confidence interval at level 0.95
² all samples are equal
³ need >= 4 samples to detect a difference at alpha level 0.05

pkg: github.com/stellar/stellar-rpc/cmd/stellar-rpc/internal/preflight
               │ baseline.txt │         candidate.txt         │
               │    sec/op    │   sec/op     vs base          │
GetPreflight-8    812.2µ ± 1%   811.5µ ± 0%  ~ (p=0.481 n=10)

               │ baseline.txt │         candidate.txt          │
               │     B/op     │     B/op      vs base          │
GetPreflight-8   52.10Ki ± 0%   52.10Ki ± 0%  ~ (p=0.470 n=10)

               │ baseline.txt │        candidate.txt         │
               │  allocs/op   │ allocs/op   vs base          │
GetPreflight-8     290.0 ± 0%   290.0 ± 0%  ~ (p=1.000 n=10)

pkg: github.com/stellar/stellar-rpc/cmd/stellar-rpc/internal/rpcv1/sqlitedb
                  │ candidate.txt │
                  │    sec/op     │
GetLedgerRange-8      941.4n ± 1%
BatchGetLedgers-8     721.4µ ± 3%
geomean               26.06µ

                  │ candidate.txt │
                  │     B/op      │
GetLedgerRange-8       16.00 ± 0%
BatchGetLedgers-8    476.8Ki ± 0%
geomean              2.730Ki

                  │ candidate.txt │
                  │   allocs/op   │
GetLedgerRange-8       4.000 ± 0%
BatchGetLedgers-8     3.120k ± 0%
geomean                111.7

pkg: github.com/stellar/stellar-rpc/cmd/stellar-rpc/internal/store
                                                           │ candidate.txt │
                                                           │    sec/op     │
ComputeFeeDistribution/alternativeComputeFeeDistribution-8     206.3µ ± 2%
ComputeFeeDistribution/ComputeFeeDistribution-8                9.262µ ± 0%
geomean                                                        43.71µ

                                                           │ candidate.txt  │
                                                           │      B/op      │
ComputeFeeDistribution/alternativeComputeFeeDistribution-8   565.4Ki ± 0%
ComputeFeeDistribution/ComputeFeeDistribution-8                0.000 ± 0%
geomean                                                                   ¹
¹ summaries must be >0 to compute geomean

                                                           │ candidate.txt │
                                                           │   allocs/op   │
ComputeFeeDistribution/alternativeComputeFeeDistribution-8    30.00 ± 0%
ComputeFeeDistribution/ComputeFeeDistribution-8               0.000 ± 0%
geomean                                                                  ¹
¹ summaries must be >0 to compute geomean

Raw benchmark logs (s3://stellar-rpc-ci-load-test/runs/36015474013/go-bench/): baseline.txt, benchstat.txt, candidate.txt

Post-optimization (commit `28ebc4c1`)

✅ Endpoint load test — verdict: ok

🎯 Endpoint load test — `50e2c12c1b82`

Serial blast per endpoint (ramp-up 1m, duration 3m, error kill switch 75%, blaster aadc1a17595f) against the backfilled RPC (ledgers [64806153, 64927112], handoff wait 1351s).

Endpoint	Target RPS	Requests	Errors	p50 (ms)	p95 (ms)	p99 (ms)	p99.9 (ms)
getEvents	75	11225	5 (0.0%)	0.8	8.8	35.1	1427.5
getFeeStats	250	37493	0 (0.0%)	0.4	0.5	0.5	1.1
getHealth	250	37492	0 (0.0%)	0.4	0.4	0.5	1.0
getLatestLedger	15	2245	0 (0.0%)	3.3	6.4	13.9	29.2
getLedgers (limit=5)	3	443	0 (0.0%)	158.0	982.0	1638.4	1751.0
getNetwork	100	14978	0 (0.0%)	0.4	0.5	0.6	10.8
getTransaction	75	11224	0 (0.0%)	5.3	11.8	17.8	24.5
getTransactions (limit=200)	10	1473	0 (0.0%)	27.8	315.4	368.1	490.5
getVersionInfo	100	14920	0 (0.0%)	0.4	0.5	0.6	3.2

getEvents results extended

Endpoint	Target RPS	Requests	Errors	p50 (ms)	p95 (ms)	p99 (ms)	p99.9 (ms)
getEvents/catch-up	3	434	0 (0.0%)	3.6	15.8	34.0	73.5
getEvents/deep-pager	17.25	2584	2 (0.1%)	0.7	26.0	142.6	1600.5
getEvents/deep-scan	2.25	319	2 (0.6%)	0.7	2.7	10.7	10002.4
getEvents/firehose	1.5	222	0 (0.0%)	0.7	2.6	3.5	3.7
getEvents/head-poll	36	5369	0 (0.0%)	0.8	5.2	8.2	44.7
getEvents/tail-poll	9	1369	1 (0.1%)	0.7	1.0	182.1	6930.4
getEvents/transfer-watcher	6	928	0 (0.0%)	6.9	11.7	13.7	15.1

✅ Go endpoint benchmarks — verdict: ok

Baseline v29.0.1 (055dc9c7a9a9) vs candidate 50e2c12c1b82 — -benchmem -count=10, both refs sequentially on one box; rpcv2 excluded.

benchstat: baseline vs candidate

goos: linux
goarch: amd64
pkg: github.com/stellar/stellar-rpc/cmd/stellar-rpc/internal/db
cpu: Intel(R) Xeon(R) Platinum 8124M CPU @ 3.00GHz
                  │ baseline.txt │
                  │    sec/op    │
GetLedgerRange-8     948.3n ± 1%
BatchGetLedgers-8    752.5µ ± 3%
geomean              26.71µ

                  │ baseline.txt │
                  │     B/op     │
GetLedgerRange-8      16.00 ± 0%
BatchGetLedgers-8   520.8Ki ± 0%
geomean             2.852Ki

                  │ baseline.txt │
                  │  allocs/op   │
GetLedgerRange-8      4.000 ± 0%
BatchGetLedgers-8    2.314k ± 0%
geomean               96.21

pkg: github.com/stellar/stellar-rpc/cmd/stellar-rpc/internal/feewindow
                                                           │ baseline.txt │
                                                           │    sec/op    │
ComputeFeeDistribution/computeFeeDistribution-8               9.591µ ± 0%
ComputeFeeDistribution/alternativeComputeFeeDistribution-8    211.4µ ± 1%
geomean                                                       45.02µ

                                                           │  baseline.txt  │
                                                           │      B/op      │
ComputeFeeDistribution/computeFeeDistribution-8                0.000 ± 0%
ComputeFeeDistribution/alternativeComputeFeeDistribution-8   565.4Ki ± 0%
geomean                                                                   ¹
¹ summaries must be >0 to compute geomean

                                                           │ baseline.txt │
                                                           │  allocs/op   │
ComputeFeeDistribution/computeFeeDistribution-8              0.000 ± 0%
ComputeFeeDistribution/alternativeComputeFeeDistribution-8   30.00 ± 0%
geomean                                                                 ¹
¹ summaries must be >0 to compute geomean

pkg: github.com/stellar/stellar-rpc/cmd/stellar-rpc/internal/methods
                               │  baseline.txt  │               candidate.txt               │
                               │     sec/op     │     sec/op      vs base                   │
GetEventsTopicFilters-8           2.590m ± 1%      2.481m ± 0%      -4.19% (p=0.000 n=10)
GetEvents-8                       152.6µ ±  ∞ ¹    152.6µ ±  ∞ ¹         ~ (p=1.000 n=1)  ²
GetLedgers-8                     1314.0µ ±  ∞ ¹    868.4µ ±  ∞ ¹         ~ (p=1.000 n=1)  ²
JSONTransactions/JSON_format-8    7.624µ ± 1%     14.216µ ± 0%     +86.47% (p=0.000 n=10)
JSONTransactions/XDR_format-8     2.599µ ± 0%      7.931µ ± 1%    +205.21% (p=0.000 n=10)
GetProtocolVersion-8              42.41µ ±  ∞ ¹    50.29µ ±  ∞ ¹         ~ (p=1.000 n=1)  ²
GetLatestLedger/uncached-8                         9.427m ± 3%
GetLatestLedger/cached-8                           1.101µ ± 0%
geomean                           87.09µ           108.6µ          +27.39%
¹ need >= 6 samples for confidence interval at level 0.95
² need >= 4 samples to detect a difference at alpha level 0.05

                               │  baseline.txt   │               candidate.txt               │
                               │      B/op       │      B/op       vs base                   │
GetEventsTopicFilters-8          190.75Ki ± 0%     47.79Ki ± 0%     -74.95% (p=0.000 n=10)
GetEvents-8                       43.16Ki ±  ∞ ¹   23.67Ki ±  ∞ ¹         ~ (p=1.000 n=1)  ²
GetLedgers-8                     1402.2Ki ±  ∞ ¹   750.8Ki ±  ∞ ¹         ~ (p=1.000 n=1)  ²
JSONTransactions/JSON_format-8    2.000Ki ± 0%     4.922Ki ± 0%    +146.09% (p=0.000 n=10)
JSONTransactions/XDR_format-8     1.336Ki ± 0%     4.781Ki ± 0%    +257.89% (p=0.000 n=10)
GetProtocolVersion-8              7.905Ki ±  ∞ ¹   8.498Ki ±  ∞ ¹         ~ (p=1.000 n=1)  ²
GetLatestLedger/uncached-8                         10.68Mi ± 3%
GetLatestLedger/cached-8                             152.0 ± 0%
geomean                           24.99Ki          26.92Ki           -5.85%
¹ need >= 6 samples for confidence interval at level 0.95
² need >= 4 samples to detect a difference at alpha level 0.05

                               │ baseline.txt  │              candidate.txt               │
                               │   allocs/op   │   allocs/op    vs base                   │
GetEventsTopicFilters-8          1524.0 ± 0%      865.0 ± 0%     -43.24% (p=0.000 n=10)
GetEvents-8                       621.0 ±  ∞ ¹    487.0 ±  ∞ ¹         ~ (p=1.000 n=1)  ²
GetLedgers-8                     4.723k ±  ∞ ¹   2.689k ±  ∞ ¹         ~ (p=1.000 n=1)  ²
JSONTransactions/JSON_format-8    25.00 ± 0%      46.00 ± 0%     +84.00% (p=0.000 n=10)
JSONTransactions/XDR_format-8     23.00 ± 0%      47.00 ± 0%    +104.35% (p=0.000 n=10)
GetProtocolVersion-8              122.0 ±  ∞ ¹    158.0 ±  ∞ ¹         ~ (p=1.000 n=1)  ²
GetLatestLedger/uncached-8                        29.00 ± 3%
GetLatestLedger/cached-8                          5.000 ± 0%
geomean                           260.6           124.1           +3.57%
¹ need >= 6 samples for confidence interval at level 0.95
² need >= 4 samples to detect a difference at alpha level 0.05

                           │ candidate.txt │
                           │      B/s      │
GetLatestLedger/uncached-8    281.8Mi ± 3%
GetLatestLedger/cached-8      2.301Ti ± 0%
geomean                       25.47Gi

pkg: github.com/stellar/stellar-rpc/cmd/stellar-rpc/internal/preflight
               │ baseline.txt │           candidate.txt            │
               │    sec/op    │   sec/op     vs base               │
GetPreflight-8    707.4µ ± 1%   726.7µ ± 0%  +2.73% (p=0.000 n=10)

               │ baseline.txt │         candidate.txt          │
               │     B/op     │     B/op      vs base          │
GetPreflight-8   52.10Ki ± 0%   52.10Ki ± 0%  ~ (p=0.907 n=10)

               │ baseline.txt │        candidate.txt         │
               │  allocs/op   │ allocs/op   vs base          │
GetPreflight-8     290.0 ± 0%   290.0 ± 0%  ~ (p=1.000 n=10)

pkg: github.com/stellar/stellar-rpc/cmd/stellar-rpc/internal/rpcv1/sqlitedb
                                 │ candidate.txt │
                                 │    sec/op     │
GetLedgerRange-8                     933.1n ± 1%
ScanLedgers-8                        712.9µ ± 4%
OldestLedgerRangeLookup/512KiB-8     105.5µ ± 2%
OldestLedgerRangeLookup/2MiB-8       758.4µ ± 1%
OldestLedgerRangeLookup/4MiB-8       1.484m ± 1%
geomean                              151.2µ

                                 │ candidate.txt │
                                 │     B/op      │
GetLedgerRange-8                      16.00 ± 0%
ScanLedgers-8                       176.8Ki ± 0%
OldestLedgerRangeLookup/512KiB-8    7.604Ki ± 0%
OldestLedgerRangeLookup/2MiB-8      7.604Ki ± 0%
OldestLedgerRangeLookup/4MiB-8      7.605Ki ± 0%
geomean                             4.139Ki

                                 │ candidate.txt │
                                 │   allocs/op   │
GetLedgerRange-8                      4.000 ± 0%
ScanLedgers-8                        1.737k ± 0%
OldestLedgerRangeLookup/512KiB-8      94.00 ± 0%
OldestLedgerRangeLookup/2MiB-8        94.00 ± 0%
OldestLedgerRangeLookup/4MiB-8        94.00 ± 0%
geomean                               89.59

pkg: github.com/stellar/stellar-rpc/cmd/stellar-rpc/internal/store
                                                           │ candidate.txt │
                                                           │    sec/op     │
ComputeFeeDistribution/alternativeComputeFeeDistribution-8     208.3µ ± 1%
ComputeFeeDistribution/ComputeFeeDistribution-8                8.542µ ± 0%
geomean                                                        42.18µ

                                                           │ candidate.txt  │
                                                           │      B/op      │
ComputeFeeDistribution/alternativeComputeFeeDistribution-8   565.4Ki ± 0%
ComputeFeeDistribution/ComputeFeeDistribution-8                0.000 ± 0%
geomean                                                                   ¹
¹ summaries must be >0 to compute geomean

                                                           │ candidate.txt │
                                                           │   allocs/op   │
ComputeFeeDistribution/alternativeComputeFeeDistribution-8    30.00 ± 0%
ComputeFeeDistribution/ComputeFeeDistribution-8               0.000 ± 0%
geomean                                                                  ¹
¹ summaries must be >0 to compute geomean

Raw benchmark logs (s3://stellar-rpc-ci-load-test/runs/38430245116/go-bench/): baseline.txt, benchstat.txt, candidate.txt

Navigation Menu

Repository navigation

Query-path optimizations: benchmark old vs. new query-path backends on benchmark config #997

What

Acceptance criteria

Pre-Optimization:

✅ Endpoint load test — verdict: ok

🎯 Endpoint load test — `14de10961916`

✅ Go endpoint benchmarks — verdict: ok

Post-optimization (commit `28ebc4c1`)

✅ Endpoint load test — verdict: ok

🎯 Endpoint load test — `50e2c12c1b82`

✅ Go endpoint benchmarks — verdict: ok

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Query-path optimizations: benchmark old vs. new query-path backends on benchmark config #997

Description

What

Acceptance criteria

Activity

cjonas9 commented on Oct 13, 2026

Pre-Optimization:

✅ Endpoint load test — verdict: ok

🎯 Endpoint load test — 14de10961916

✅ Go endpoint benchmarks — verdict: ok

Post-optimization (commit 28ebc4c1)

✅ Endpoint load test — verdict: ok

🎯 Endpoint load test — 50e2c12c1b82

✅ Go endpoint benchmarks — verdict: ok

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

🎯 Endpoint load test — `14de10961916`

Post-optimization (commit `28ebc4c1`)

🎯 Endpoint load test — `50e2c12c1b82`