data for forecasting traffic demand for content providers

In our forthcoming paper, we try to model connectivity demand of content providers, by estimating the traffic for "Content" ASes.

As this data is not usually disclosed, we hacked our way around (by grabbing public statistics in json or processing public graphs) from the IXP websites and we coupled extracts from peeringdb.com API to know what was the bandwidth allocated to Content Providers and CDN (Akamai, Netflix, Facebook).

After grabbing the stats, we forecasted the time series with R using a SARIMA model. The forecast are converted in Service Level Agreements, to be latter embedded into a ISP Network Topology, to realise the vCDN usecase.

Here's a 3D representation of 4 IXP peering point extracts (normalized at 10Go), that has been discretized to produce SLAs.

 

 

Sponsors

Current Sponsors

 

Previous Sponsors In Grenoble:

 

Previous Sponsors In Bordeaux: