Skip to content

Hardware Scaling

Observium's RAM and CPU requirements scale linearly primarily around the number of ports for networking-centric installs and the number of devices for server-centric installs.

For Web GUI requirements we use our slowest page, /ports/ as an example.

Base Requirements

For minimum requirements you should consult the documentation of your Linux distribution as well as Apache and MySQL.

A small Observium installation will probably never use more than a few tens of MB of RAM itself, but it relies upon Apache and MySQL which will also use RAM. It should be possible to run Apache and MySQL in ~128MB of RAM, but at least 256MB is recommended for a small installation.

Processor (Web)

Web GUI page generation depends upon a single core, the faster the clock rate of the CPU, the faster the page generation, regardless of the number of cores. Multiple cores will help speed up graph loading, which are loaded in parallel, usually 4 at a time.

  • A 3.2GHz Xeon E3-1225 V2 should take 0.15 seconds to generate /ports/ with 6k ports.

At present single threaded performance is still better on Intel Core i7 processors, so we'd recommend looking at those first if you want to split the web ui onto its own system.

Processor (Poller)

The poller-wrapper runs poller processes in parallel. Depending upon the average latency of the devices you're monitoring and the I/O throughput of your storage you should run ~2 pollers per core so as not to waste CPU cycles.

  • A quad-core 2.13GHz Xeon E5506 should scale to ~20k ports.
  • A quad-core 3.4GHz i7-3770 should scale to ~40k ports.

With the introduction of AMD's Ryzen and Threadripper CPUs, scaling to high core counts on single sockets is now much easier.

Memory

The most RAM-intensive page is /ports/. This page currently uses ~16.5MB per 1000 ports. The PHP process must to be permitted enough memory to generate this page.

  • A 6k port installation uses ~29MB of RAM to generate /ports/

  • Poller memory usage is negligible.

Storage Capacity

Each port will use ~3MB of storage for RRDs, and each host will add another 5-50MB, depending upon the operating system and the number of ports.

  • A 5,000 port installation will generate ~8GB of RRDs.
  • A 11,000 port installation will generate ~23GB of RRDs.

Storage I/O Throughput

Storage I/O throughput is the most serious bottleneck on many large deployments.

  • A single 7200RPM drive will handle the RRD I/O of about 5,000 ports. This can be increased by using RAID-0 or faster (10k, 15k) drives.

We strongly recommend using SSD-based storage. SSDs provide much greater I/O throughput and latency. We recommend good quality SSDs with high endurance and IOPS ratings.