Telemetry is written all day in huge volumes, never changed once written, and read rarely: AI agents query it in short bursts during an investigation, not dashboards refreshing all day. VictoriaMetrics and InfluxDB register every series in an index so a dashboard can find it in a millisecond; that index, and the memory it needs, grows with every new label combination. Quiver keeps no series index, so the number of series barely changes its cost.
Every size we ran
Each tile is one run. Kept up: took in every series, held the write rate, and every sample checked came back exactly. Quiver passed every size up to 100M, the largest we tested, without reaching its limit. For VictoriaMetrics and InfluxDB we ran the last size that passed and the first that failed.
Quiver's memory doesn't grow with the number of series
Engine memory while ingesting, by unique series
QuiverVictoriaMetricsInfluxDBEngine memory while ingesting. A cross marks an engine killed at its 2 GB cap. Not included: Garage, the S3-compatible object store Quiver writes to (20–325 MiB), and Telegraf, InfluxDB's remote-write adapter. Both are in Addendum A.
Pods replaced every 2 minutes: Quiver stays flat
200,000 series active at once, every pod renamed every 2 minutes, as deploys and autoscaling do: 1.4M distinct series in 12 minutes.
Memory per quarter, incl. Garage / Telegraf
QuiverVictoriaMetricsInfluxDBBuilt for agents, not dashboards
VictoriaMetrics answers from its index in under a millisecond. Quiver scans the time range instead: 7 ms to 1.2 s.
That matters to a dashboard refreshing dozens of panels. It doesn't to an agent: each query sits between LLM calls that take seconds, so a few hundred milliseconds is lost in the wait. Skipping the index is also why Quiver doesn't mind cardinality.
| Query | Quiver | VictoriaMetrics | InfluxDB |
|---|---|---|---|
| Memory of one pod (last 15 min) | 281 | 0.2 | 2.5 |
| CPU rate of one container (last 1 h) | 215 | 0.1 | 2.7 |
| Request rate per pod of one service (last 1 h) | 1,047 | 0.4 | 231 |
| Top 5 pods by CPU in one namespace (now) | 102 | 0.4 | 15.2 |
| 5xx error ratio of one service (last 1 h) | 1,186 | 0.3 | 326 |
| Average memory of one pod (last 1 h) | 294 | 0.1 | 2.3 |
| Which pods one service has (last 1 h) | 6.7 | 0.1 | 2.2 |
Typical (p50) response in milliseconds at 100k series, the size all three hold. 20 runs each; every answer checked against the true value. Similar with ingest running.
Method
- Data: Kubernetes-style metrics, 68 series per pod with
cluster,namespace,service,pod,nodeandjoblabels. Identical for every engine. - A fixed 10,000 samples a second at every size; only the number of series changes. Each run writes the history (every series at least twice), then a live phase, then checks.
- Ingest: Quiver over Arrow Flight, VictoriaMetrics over Prometheus remote write, InfluxDB through Telegraf (InfluxDB 2 has no remote-write endpoint).
- Limits: 2 cores / 2 GB per engine, no swap, so going over the cap means being killed. Garage 1 core / 1 GB; Telegraf 2 cores / 2 GB.
- Not compared: storage size (Quiver's compaction did not run in these short tests), histogram percentile queries, clusters and paid editions.
Every run and versions
AEvery run
| Engine | Unique series | Each series written every | Result | CPU, cores (incl. Garage / Telegraf) | Engine memory | Memory incl. Garage / Telegraf | Engine memory peak | Series creation, samples/s | Sample count |
|---|---|---|---|---|---|---|---|---|---|
| Quiver | 100k | 10 s | Kept up | 0.04 | 79 MiB | 125 MiB | 581 MiB | 2.21M | exact |
| Quiver | 1M | 100 s | Kept up | 0.04 | 81 MiB | 101 MiB | 539 MiB | 2.65M | exact |
| Quiver | 10M | 1,000 s | Kept up | 0.06 | 81 MiB | 116 MiB | 204 MiB | 1.29M | exact |
| Quiver | 100M | 10,000 s | Kept up | 0.09 | 95 MiB | 420 MiB | 1.15 GB | 1.60M | exact |
| VictoriaMetrics | 100k | 10 s | Kept up | 0.02 | 106 MiB | 106 MiB | 348 MiB | 2.35M | exact |
| VictoriaMetrics | 1M | 100 s | Kept up | 0.02 | 492 MiB | 492 MiB | 1.06 GB | 487k | exact |
| VictoriaMetrics | 3M | 300 s | Kept up | 0.04 | 1.23 GB | 1.23 GB | 1.92 GB | 146k | exact |
| VictoriaMetrics | 5M | 500 s | Out of memory | – | – | – | 2.00 GB | – | – |
| InfluxDB | 100k | 10 s | Kept up | 0.13 | 309 MiB | 441 MiB | 526 MiB | 164k | exact |
| InfluxDB | 500k | 50 s | Kept up | 0.16 | 949 MiB | 1.06 GB | 1.92 GB | 100k | exact |
| InfluxDB | 1M | – | Out of memory | – | – | – | 1.91 GB | – | – |
Pods replaced every 2 minutes, per quarter of the live phase:
| Engine | Q1 CPU | Q2 CPU | Q3 CPU | Q4 CPU | Q1 memory | Q2 memory | Q3 memory | Q4 memory | Engine memory peak | Sample count |
|---|---|---|---|---|---|---|---|---|---|---|
| Quiver | 0.03 | 0.04 | 0.05 | 0.05 | 103 MiB | 96 MiB | 95 MiB | 106 MiB | 180 MiB | exact |
| VictoriaMetrics | 0.06 | 0.06 | 0.06 | 0.06 | 783 MiB | 664 MiB | 863 MiB | 979 MiB | 1.06 GB | exact |
| InfluxDB | 0.22 | 0.23 | 0.20 | 0.22 | 944 MiB | 0.99 GB | 1.18 GB | 1.42 GB | 1.82 GB | exact |
BVersions
- Machine
- AWS us-east-1, c8a.2xlarge (8 vCPU AMD EPYC 9R45, 16 GiB), 200 GB gp3 (3,000 IOPS, 250 MB/s), Ubuntu 24.04, Docker 29.9.3
- Quiver
- core0 build 35b4016, single node
- Object storage
- Garage v2.5.0, single node, same disk
- VictoriaMetrics
- v1.155.0 single-node, open source (Apache 2.0), default settings
- InfluxDB
- 2.9.1 OSS (MIT), default settings; Telegraf 1.40.3 (MIT), prometheusremotewrite parser, metric_version 1