Overview
The current traffic view provides basic traffic metrics to ISPs who have at least one interconnect asset: IX peering session, PNI or GGC.
Where does the data come from?
Google continuously monitors traffic across its edge network. Every few minutes, data is pulled from the company's network monitoring systems to generate a per-asset time-series of various traffic metrics.
Metric details
Asset egress
Egress measures the flow of traffic from Google to your network as measured by Google's serving hardware.
Asset ingress
Ingress measures the flow of traffic from users of your network to Google.
GGC cache efficiency
Cache efficiency measures the amount of cacheable user demand that the GGC cache node is able to offload.
If a new serving point is in the process of being turned up, you can expect this metric to be low. As the cache fills with content this metric will rise. Generally, this value will be above 60% for caches that are well-utilized.
There are several reasons this metric may go down:
-
Users are requesting unpopular videos
-
Users are abandoning videos mid-stream (possibly due to bad network connectivity)
GGC cache fill
Cache fill is the rate at which a GGC cache node is retrieving video data from another source (which may either be another cache node, or Google's network) to fill its local cache.
The cache fill rate may increase when a new serving point is in the process of being turned up.
GGC egress-ingress ratio
Egress-to-ingress ratio measures the ratio between the asset egress and the cache fill for a single GGC serving point.
A simple interpretation of this value is that the node is serving X many times the amount it is downloading at this point in time" where X is the value observed on the graph.
This ratio is computed by dividing the bandwidth served (egress) by the bandwidth used for content download by the node (ingress) at a certain point in time.
For example, for a ratio value of 2, it means that the node is serving twice the amount of content it is downloading at that given point in time.