Scaling

Scaling Guide

Performance Sizing

By Network Capacity

Component	1 Gbps	10 Gbps	25 Gbps	100 Gbps
OpenSearch	1 node, 8GB	3 nodes, 16GB	3 nodes, 32GB	5+ nodes, 64GB+
ntopng	4GB RAM	16GB RAM	32GB RAM	128GB RAM
Prometheus	2GB, 200GB	4GB, 1TB	8GB, 2TB	32GB, 10TB
Loki	2GB, 500GB	8GB, 2TB	16GB, 5TB	64GB, 20TB
Total Memory	16GB	48GB	96GB	384GB+

By Number of Devices

Metric	<50	50-500	500-5000	5000+
LibreNMS Pollers	1	2-4	4-8	8-16
Prometheus Retention	90d	60d	30d	14d
OpenSearch Nodes	1	3	5	7+

Tuning

Prometheus

--storage.tsdb.min-block-duration=2h
--storage.tsdb.max-block-duration=4h
--query.max-concurrency=40

OpenSearch

thread_pool.write.queue_size: 10000
index.translog.durability: async
index.refresh_interval: 60s

ntopng

--max-num-flows=5000000
--pkt-max-size=9216

Loki

limits_config:
  ingestion_rate_mb: 50
  max_streams_per_user: 50000

Storage

Component	Minimum	Recommended
Prometheus (WAL)	SSD	NVMe RAID 10
Loki (Chunks)	SSD	NVMe RAID 10
OpenSearch	SSD	NVMe RAID 10
MariaDB	SSD	NVMe RAID 1

Capacity Planning

• Prometheus: ~1GB per 100k time series / day
• Loki: ~1GB per 10GB log input / day
• OpenSearch: ~3GB per TB indexed / day

---

Full reference: docs/scaling/