security onion storage requirements

Security Onion Storage Requirements: The Complete Guide to Sizing, Planning, and Performance

By Melissa Smith
AD

Security Onion is a powerful open-source platform for network security monitoring, intrusion detection, endpoint visibility, and threat hunting. Its ability to centralize alerts, metadata, logs, and full packet capture makes it a favorite for SOC teams, but that visibility comes with a serious storage bill. Understanding Security Onion storage requirements is not just about buying the biggest disk you can afford; it is about matching capacity, throughput, and retention to the traffic and investigative needs of your environment.

In practice, storage is often the first bottleneck that appears after deployment. A node may have plenty of CPU and RAM, yet slow dashboards, dropped packets, and failed searches usually trace back to undersized or misconfigured storage. The following guide breaks down the core storage demands, sizing calculations, hardware choices, and operational habits that keep Security Onion responsive and reliable.

Understanding Security Onion Storage Requirements

Security Onion stores several very different classes of data. Alert records, Zeek logs, Suricata events, endpoint telemetry, and system logs flow into Elasticsearch, where they are indexed for fast search and dashboards. At the same time, full packet capture or selective PCAP may be written to local disk for deep investigation. These workloads have distinct storage profiles: indexing and search demand high random I/O, while packet capture is a continuous sequential write stream.

AD
Cybersecurity is Like An Onion
Cybersecurity is Like An Onion
Security Onion Installation
Security Onion Installation
an info sheet with the words, what is a onion site? and other information
an info sheet with the words, what is a onion site? and other information

Because of that mix, a single generic disk recommendation rarely fits every deployment. A small lab monitoring a few devices has modest requirements, but a production SOC evaluating thousands of events per second can generate terabytes per day. The right storage design starts with a clear picture of what data you must retain, how quickly you need to search it, and how much traffic you intend to inspect.

Core Data Types and Retention Policies

The search backend is usually the largest storage consumer. Elasticsearch indices hold normalized events, alerts, dashboards, and metadata. Hot data, which is actively searched, needs fast SSD or NVMe storage. Warm data, which is older but still needed for compliance or periodic review, can live on slower, high-capacity HDDs. Retention policies determine how long each index remains in hot, warm, or cold tiers before deletion or archival.

Onion Storage Field Guide
Onion Storage Field Guide
the title for how to create a onion site, with an image of a person wearing a
the title for how to create a onion site, with an image of a person wearing a
the onion service info sheet is designed to look like it could be on fire or ice
the onion service info sheet is designed to look like it could be on fire or ice

PCAP is another major category. Full packet capture preserves the actual network traffic, which is invaluable for confirming an alert, extracting files, or performing retrospective analysis. However, full capture on a busy link can consume enormous space. Many teams use selective capture, metadata-only logging, or shorter PCAP retention windows. A clear retention policy should define how long alerts, logs, metadata, and PCAP are kept, and whether any data must be archived for regulatory reasons.

Performance Metrics That Drive Storage Design

Capacity alone does not define a successful storage system. IOPS, throughput, and latency matter just as much. Elasticsearch performs many small reads and writes during indexing, searching, and segment merging, so it benefits from low-latency storage. If the storage cannot keep up, indexing queues grow, searches time out, and the cluster may enter a yellow or red state. Packet capture, by contrast, needs sustained write throughput so that packets are not dropped when traffic spikes.

AD

When planning, measure events per second, average event size, peak network bandwidth, and the number of concurrent analysts. A deployment that ingests 2,000 events per second with occasional hunting queries has a very different profile from one that ingests 20,000 events per second and runs continuous dashboards. Use those numbers to estimate both daily data growth and the I/O load that the disks must handle during peak hours.

How And When To Harvest, Cure And Store Homegrown Onions
How And When To Harvest, Cure And Store Homegrown Onions
how onion routers work on computers and laptops in the same area, with instructions for
how onion routers work on computers and laptops in the same area, with instructions for
DIY Onion Storage Solutions for Long-Term Freshness
DIY Onion Storage Solutions for Long-Term Freshness
Onion storage idea
Onion storage idea
the step 1 installer screen is shown in this screenshote image, which shows how
the step 1 installer screen is shown in this screenshote image, which shows how
Best Way to Store Potatoes and Onions | Keep Them Fresh Longer
Best Way to Store Potatoes and Onions | Keep Them Fresh Longer

Hardware, Sizing, and Best Practices for Security Onion Storage

Security Onion can be deployed as an all-in-one node or as a distributed cluster with separate manager, search, and sensor nodes. The all-in-one approach is simpler and works well for labs, small businesses, or proof-of-concept environments. Distributed deployments are more complex but allow storage to scale independently from packet capture and management functions, which is essential for high-traffic networks.

Regardless of architecture, avoid treating storage as an afterthought. The operating system, Elasticsearch data, packet capture, and logs should be separated onto different volumes or mount points when possible. This separation prevents a full PCAP disk from crashing the search backend, simplifies backups, and makes it easier to tune performance for each workload. It also helps when you need to expand one data type without rebuilding the entire node.

AD

All-in-One vs. Distributed Storage Architecture

An all-in-one Security Onion node must handle management, search, and sensor duties on one server. That makes local storage critical. Use enterprise-grade SSDs or NVMe drives for the Elasticsearch volume, and consider a separate high-capacity HDD or RAID array for PCAP if you plan to retain packets. Official minimums may get a lab running, but production use requires more headroom. For small deployments, a 2 TB to 4 TB NVMe volume plus 8 TB to 16 TB of PCAP storage is a reasonable starting point, but the exact size depends on EPS and retention.

How to Store Onions And Keep Them Fresh - Eco Peanut
How to Store Onions And Keep Them Fresh - Eco Peanut
How to Hang Onions for Storage - Homestead Honey
How to Hang Onions for Storage - Homestead Honey
You’ve grown your onions, dried (cured) them, now it’s time for storage!
You’ve grown your onions, dried (cured) them, now it’s time for storage!
an image of onions hanging from the side of a fence with text overlay reading how to store onions to last months
an image of onions hanging from the side of a fence with text overlay reading how to store onions to last months

In a distributed deployment, storage responsibilities are split. Search nodes host Elasticsearch and should use fast local SSD or NVMe for hot indices, with larger HDD-backed warm nodes for older data. Sensor nodes need enough local storage for their packet capture and log buffering, especially if the network to the manager is interrupted. Manager nodes require relatively little storage, but they still need reliable disks for configuration, certificates, and administrative logs. This separation lets you add capacity where it is actually needed instead of overbuilding every node.

Capacity Planning, RAID, and Monitoring

Start capacity planning with a simple daily growth estimate. For indexed data, multiply events per second by average event size, then multiply by 86,400 seconds. Add overhead for replicas, indexing structures, and segment merges; a multiplier of 1.5x to 2x is common. For PCAP, multiply sustained bandwidth by retention time. At 100 Mbps, full capture is roughly 1.08 TB per day; at 1 Gbps, it is about 10.8 TB per day. Those numbers explain why selective capture and tiered retention are so important.

AD

For RAID, hot Elasticsearch volumes often perform best on RAID 10 because it provides both speed and redundancy. Warm or archive storage can use RAID 6 for greater capacity, though write performance will be lower. Avoid RAID 5 for write-heavy search nodes. Monitor disk usage, I/O wait, and Elasticsearch disk watermarks continuously. Keep hot storage below the high watermark, usually around 85 percent, and set up alerts before a volume fills. Regular snapshots to remote storage, tested restores, and documented retention policies turn storage from a liability into a dependable part of your security operations.

Strong Security Onion storage planning pays off every time an analyst needs a fast search during an incident. Review your EPS and retention needs quarterly, test new drive configurations in a lab, and keep an eye on the official Security Onion documentation for updated hardware guidance. With the right balance of fast hot storage, scalable warm capacity, and disciplined retention, your deployment can deliver the visibility it promises without constant firefighting.

AD