Computer memory occasionally flips a bit: a 0 becomes a 1 or the reverse. Causes include electrical noise, ageing components and even cosmic radiation. On a single desktop it might never matter. Across servers running day and night with large amounts of memory, errors are a statistical certainty.
What ECC does
Error-correcting code (ECC) memory stores extra bits alongside your data. When data is read, the memory controller checks it:
- A single-bit error is corrected automatically, and the system carries on.
- A multi-bit error is detected, so the system can stop or report it instead of using bad data.
Corrected errors are logged, which gives early warning of a module that is starting to fail.
Why it matters
Without ECC, a flipped bit can:
- Crash the operating system or an application.
- Corrupt data in memory that is then written to disk, into a database or into a backup, where it may go unnoticed for months.
- Cause errors that are almost impossible to reproduce or diagnose.
For databases, financial systems, file servers, virtualisation hosts and anything you would not want quietly corrupted, ECC is cheap insurance.
When non-ECC is acceptable
Development machines, disposable workloads, and services where data is stateless and easily rebuilt can run without ECC. Some desktop-class processors do not support it at all, which is worth checking before you choose one for an important server.
Monitor the error counters
ECC hides small errors from your applications, but it should not hide them from you. Watch the corrected-error counters, usually reported through the server's management controller or the operating system, and replace modules that show a rising rate.
How IDX helps
When you request a dedicated server from IDX, you can choose ECC or non-ECC memory, or say you have no preference. Our team will tell you if the processor you prefer limits that choice, and our managed services include monitoring of hardware health.
IDX on digital infrastructure
Back to blog →