Skip to content
Rows of server racks in a dark data centre aisle, small status lights glowing along each rack.

OTT & Media · Partly AI

Transcoding errors forecast a month ahead

A monitoring system for more than 7,000 channels that reads early warning signs in the broadcast metrics and forecasts transcoding errors up to a month before viewers would see them.

Client
An IT services provider that monitors transcoding and online broadcasting
Segment
OTT & Media, broadcasting
Scale
7,000+ SD and HD channels
Hosting
A closed data centre, all traffic on internal infrastructure
  • VictoriaMetrics
  • Grafana
  • GStreamer

Before

What was true before

The client’s previous system for aggregating transcoding and broadcast data had been hacked. It needed a replacement it could trust, and a way to see a failing channel before its viewers reported it.

After

What we did

A monitoring and aggregation system runs inside a closed data centre and collects bitrate, memory, CPU and device state from every channel. An analytics layer reads the current metrics for early signs, such as a rising error rate, and forecasts transcoding errors up to a month ahead.

Build log

How it was built

work log · OTT & Media · Broadcast monitoring provider7 steps
  1. 01

    Rebuilt monitoring and aggregation inside a closed data centre

  2. 02

    Pushed metrics from edge locations into several local data centres

  3. 03

    Secured the system with certificates and closed ports

  4. 04

    Tracked bitrate, memory, CPU, state and device status per channel

  5. 05

    Built predictive analytics that forecast transcoding errors from early signs

  6. 06

    Added automated QA tests that check databases, components and metrics on each update

  7. 07

    Audited the GStreamer framework to make transcoding and live streaming more efficient

Measured

Results

As reported in the published case.

SD and HD channels monitored
7,000+
reported
Transcoding errors forecast up to 1 month ahead
reported
Big data stored each month cut 10x
reported
Routine issue requests from end users cut 10x
reported
Issues addressed in about 2 minutes
The published case on oxagile.com

Engagement

Where this work sits

Next step

Yours would be the next one.

Tell us where it breaks, and you get a written answer on what breaks first and what it costs.