Sizing and capacity planning

What you will do: estimate your daily volume and retention, then pick node sizes and storage for each tier.

Sizing is driven by three numbers: how much data arrives per day, how long you keep it hot, and how long you keep it in the archive. Measure these before you install; guesses lead to either idle hardware or full disks.

Measure your inputs

1. Daily volume. Sum the raw size of the logs, events and metrics you plan to send, per day, at peak. PortX typically cuts the volume that reaches downstream systems by 50–70 %, but size collection for the raw figure.
2. Streams. For PortX, count the distinct sources and destinations you will route. Streams are also a licensing unit; see How licensing works.
3. Hot retention. How many days must be searchable directly.
4. Archive retention. How many days the full-fidelity copy must stay in low-cost storage.

Tiers to size

• Hot data — fast disk on the nodes, sized for daily volume × hot retention plus headroom.
• Persistent queue — disk for data in flight; size for the longest outage of a destination you want to ride through.
• Archive/cold — object storage or cheap disk, sized for daily volume × archive retention.

Reference sizes

Daily volume      Nodes   CPU per node   Memory per node   Hot disk   Queue disk
<placeholder>     <placeholder>  <placeholder>   <placeholder>   <placeholder>   <placeholder>
<placeholder>     <placeholder>  <placeholder>   <placeholder>   <placeholder>   <placeholder>
<placeholder>     <placeholder>  <placeholder>   <placeholder>   <placeholder>   <placeholder>

Rules of thumb

• Start one size larger than the estimate and scale down after a week of real data.
• Keep the data directory on its own disk so the operating system never competes with ingestion.
• Plan clusters with an odd number of nodes where the product uses coordination.
• Review Usage in your workspace monthly; it shows the real GB/day and streams per deployment.

Next: Upgrades

Verify with XPLG engineering before publishing.