balanced · SlateDB 0.15.0Version0.15.00.14.10.0.0-sha-90117dc27a560.0.0-sha-7574f293cc250.0.0-sha-5b494cc46af80.0.0-sha-4d2a2bee5015<br>Recorded7/29/2026, 8:40:48 PM UTC7/29/2026, 8:40:42 PM UTC<br>Workloadsbalancedidlepoint-read-missingpoint-read-skewedpoint-read-uniformrange-scanread-heavysustained-ingesttransaction-contentionupdate-heavy<br>Golden datagolden<br>Object storeS3 · us-east-1<br>Machinewarp-ubuntu-latest-arm64-32x ↗ · Ashburn, Virginia, US<br>PatchesNone<br>Raw resultRaw time seriesActions log ↗SlateDB source ↗Benchmark source ↗<br>Application operations
APItotalavg/sp0.1/sp1/sp50/sp99/sp99.9/smin/smax/sget›<br>5.59M6.22K4.5K4.66K6.28K7.37K7.46K4.47K7.49KLoading chart…
put›<br>5.59M6.21K4.49K4.66K6.27K7.41K7.6K4.47K7.77KLoading chart…
flush›<br>100000000Loading chart…
Application throughput
APItotal GiBavg MiB/sp0.1p1p50p99p99.9minmaxget›<br>2.192.491.81.872.522.952.991.793Loading chart…
put›<br>2.192.491.81.872.512.973.051.793.11Loading chart…
Application latency
APIavg msp0.1p1p50p99p99.9minmaxget›<br>10.280.010.010.0551.291.840793.09Loading chart…
put›<br>0.0100.010.010.040.0805.29Loading chart…
flush›<br>79.2679.379.379.379.379.379.2379.3Loading chart…
durableiTime from when a write returns until SlateDB first reports its sequence durable.›<br>116.5335.242.75115.26219.9301.5727.01493.57Loading chart…
Object-store operations
Methodtotalavg/sp0.1/sp1/sp50/sp99/sp99.9/smin/smax/s$$/monthGET›<br>2.1M2.33K2.09K2.12K2.34K2.47K2.66K2.08K2.74K$0.8384$2,414.31Loading chart…
LIST›<br>1540.170004507.99$0.000770$2.22Loading chart…
PUT›<br>9.38K10.4388.99102574775$0.0469$135.11Loading chart…
DELETE›<br>8.24K9.15000224.91235.170239.2$0.00$0.00Loading chart…
POST›<br>1940.22000810011$0.000970$2.79Loading chart…
Object-store throughput
Methodtotal GiBavg MiB/sp0.1p1p50p99p99.9minmaxGET›<br>11.0812.617.817.948.74157.83511.327.77526.5Loading chart…
LIST›<br>0.030.040001.251.2602.08Loading chart…
PUT›<br>6.597.491.791.912.66129.19642.081.78645.79Loading chart…
DELETE›<br>000000.090.0900.09Loading chart…
POST›<br>0000000.0100.01Loading chart…
Process statistics
Metricavgp0.1p1p50p99p99.9minmaxCPU utilization (cores)›<br>0.980.690.730.922.316.370.686.82Loading chart…
RSS (GiB)›<br>16.437.667.8217.3622.5222.567.622.57Loading chart…
Machine statistics
Metricavgp0.1p1p50p99p99.9minmaxCPU (%)›<br>3.072.12.262.917.219.782.0421.28Loading chart…
Memory used (GiB)›<br>18.29.439.5819.1524.2724.329.3924.32Loading chart…
Network receive (MiB/s)›<br>14.639.449.5910.61163.47519.949.35534.9Loading chart…
Network send (MiB/s)›<br>125.96.197.03136.86666.415.85667.07Loading chart…
Disk read (MiB/s)›<br>00000.120.2100.46Loading chart…
Disk write (MiB/s)›<br>0.020000.131.4101.42Loading chart…
Disk read ops/s›<br>0.15000120117Loading chart…
Disk write ops/s›<br>0.530001323034Loading chart…
Benchmark configuration
KeyValuescale1dataset / record count300,000,000dataset / key bytes20 Bdataset / value bytes400 Bdataset / value compression ratio1caches / block bytes8 GiBcaches / metadata bytes4 GiBtask / taskbalancedtask / clients64task / warmup ms300,000 mstask / measurement ms900,000 mstask / initial stategoldentask / key selectionscrambled-zipfian-0.99task / operation mix / get0.5task / operation mix / put0.5task / scan limit—task / transaction hot keys—task / transaction reads—task / transaction updates—build profilereleaseenabled featuresaws, foyer, wal_disable, zstd
SlateDB configuration
KeyValueflush interval100msWAL enabledYesmanifest poll interval1smanifest update timeout300smin filter keys1,000L0 SST size bytes64 MiBmax WAL flushes before L0 flush4,096L0 max ssts32L0 max ssts per key32L0 flush parallelism4max unflushed bytes2 GiBcompactor options / poll interval5scompactor options / manifest update timeout300scompactor options / max concurrent compactions4compactor options / enable trivial moveNocompactor options / scheduler options / max compaction sources16compactor options / worker / max concurrent compactions4compactor options / worker / compactions poll interval5scompactor options / worker / heartbeat interval10scompactor options / worker / max SST size268,435,456compactor options / worker / max fetch tasks4compactor options / worker / bytes to fetch2,097,152compactor options / worker / max subcompactions8compactor options / worker / min filter keys1,000compactor options / worker / compression codec—compactor options / worker / metric level—compactor options / metric level—compactor options / commit compacted interval1scompactor options / worker heartbeat timeout30scompactor options / object store max retries—compression codec—object store cache options / root folder—object store cache options / max cache size bytes—object store cache options / part size bytes4 MiBobject store cache options / cache on flushNoobject store cache options / cache on compactionNoobject store cache options / preload disk cache on startup—object store cache options / scan interval3600sobject store cache options / max open file handles1,000garbage collector options / manifest options / interval60sgarbage collector...