Field Notes

Self-Hosted Observability Capacity Calculator

Size your Kubernetes observability stack before you provision: storage, compute, nodes, and AWS cost broken down to individual Helm chart components.

Data Layer
LGTM Stack · Victoria Stack · ClickHouse OSS
Visualization
Grafana · HyperDX (ClickHouse option)
Global Parameters
×
Multiplier applied to the average ingest rate.
%
Reserved capacity for growth and uneven load.
%
Reserved for Kubernetes and DaemonSets.
%
Reserved for Kubernetes and DaemonSets.
%
Target utilization after node reservations.
%
Additional capacity for metadata and uneven distribution.
Logs
10 100 500 2K
GB/day
days
×
Default: 5×. Published range: 5–10×.
Replica minimum; workload sizing may add pods.
hours
Recovery window used to size each ingester PVC.
Advanced Loki parameters
%
Upstream default: 90%.
MB/s
Value of Loki's ingestion_rate_mb setting.
%
Usable fraction of the configured limit after operational reserve.
GB
Scratch PVC for the Compactor StatefulSet.
GB
PVC for each Index Gateway StatefulSet pod.
Chart default: 2 replicas.
Chart default: 1 replica.
MB
Upstream default: 1.5 MB.
×
Query-capacity multiplier; 1× is the default load.
MB
Container limit is configured memory × 1.2.
MB
Container limit is configured memory × 1.2.
0 disables the cache; add replicas for higher read volume.
0 disables the cache; 1 is the default replica count.
0 disables the Ruler; enable it for recording or alert rules.
Traces
0 10 100 2K
GB/day
days
×
Default: 3×. Published observed range: 2–4×.
Advanced Tempo parameters
bytes
Average serialized span size; replace with measured data.
hours
MB/s
Default partition throughput used for capacity sizing.
MB/s/replica
Published sizing range: 6–10 MB/s per replica.
MB/s/replica
Published sizing range: 1–2 MB/s per replica.
GB/pod
Recommended per-pod memory; Minimum uses the starter request.
GB/pod
Recommended per-pod memory; Minimum uses the starter request.
GB/pod
Recommended per-pod memory; Minimum uses the starter request.
Increase when compaction cannot keep pace with ingestion.
MB/pod
0 disables Memcached. The chart does not set resource requests.
0 disables trace-derived RED metrics.
0 disables the NGINX routing proxy.
Metrics
50K 500K 2.5M 10M
series
days
Advanced Mimir parameters
bytes
Default: 2 B/sample for compacted chunks and index.
bytes
Default serialized Kafka record size, including series labels.
hours
Recovery window retained for an unavailable ingester.
Default: 3 copies for production availability.
Kafka mode keeps two ingester copies per partition; partitions scale automatically.
Minimum: 3 pods across the zone-aware StatefulSets.
0 disables Alertmanager; enabled pods use a 1 GiB state PVC.
0 disables the Ruler; enable it when Mimir evaluates rules.
MB
0 disables the cache; enabled plans use 1 Minimum and 3 Recommended pods.
MB
0 disables the cache; enabled plans use 1 Minimum and 3 Recommended pods.
MB
0 disables the cache; enabled plans use 1 Minimum and 3 Recommended pods.
MB
0 disables the cache; enabled plans use 1 Minimum and 3 Recommended pods.
Grafana Medium
1 15 50 200
users
Small ≤50 · Medium ≤200 · Large ≤500 · XL >500
Small ≤2 · Medium ≤8 · Large ≤15
Small ≤30 · Medium ≤150 · Large ≤500
Adds the selected managed PostgreSQL deployment to monthly cost.
0 uses tier sizing: 1 replica for Large and 2 for XL.
Infrastructure
General-purpose shapes provide 4 GB RAM per vCPU · EC2 prices ↗ · RDS prices ↗
Recommended worker minimum for a three-zone deployment.
Advanced network and storage performance
Gbps
0 disables NIC validation. Enter the sustained bandwidth for the selected instance.
MiB/s
Default: gp3 baseline of 125 MiB/s. Override for provisioned gp3 or io2.
Default: gp3 baseline of 3,000 IOPS.
KiB
0 leaves IOPS unevaluated. Use a measured read/write average.
MiB/s
Adds observed query reads, compaction, and merge traffic to each modeled data PVC.
%
0 excludes cross-AZ cost. Replace with measured or topology-derived traffic.
$/GB
AWS us-east-1 default. Both transfer directions are charged.
Total Storage
—
CPU (min)
—
RAM (min)
—
Nodes
—
Storage
SignalSizeTypeBasis
Network and disk performance
Network
BackendAverage ingestPeak ingestPeak internalN−1
Data PVCs
Storage pathPodsPeak writes/podPlanned peak I/O/podStatus
Compute
Minimum = low-load starter requests and required topology. Recommended = workload sizing, HA, and headroom.
Node plan
StepMinimumRecommended
Monthly cost AWS us-east-1 on-demand rates, Sep 2026 · excludes unmodeled data transfer, API requests, backups, provisioned IOPS charges, and external Kafka compute
Line item Minimum Recommended Share

Scenario comparison

Minimum plan for the selected backend only. Costs exclude the shared EKS control plane.

Stack comparison

Same workload and infrastructure settings. Each card shows Minimum → Recommended.