GGC network traffic can be broadly divided into two types: high-volume and low-volume.
For all types of traffic, the choice of serving location is made at our discretion. Our intent is to serve users from the best available location for each service, which may or may not be a local GGC node.
High-Volume traffic
Measured in Gbps, the majority of GGC traffic is cacheable YouTube content. High-volume traffic can be served to users by a local GGC node, or from Google core data locations via peering and transit.
For high-volume traffic, our systems map user requests to serving locations based on the IP prefixes we see advertised at GGC and peering.
Some high-volume content is not popular enough to be cached on GGC. Non-cached requests may be served from Google core data locations, or proxied via GGC.
Low-Volume traffic?
Low-volume traffic is Google front-end traffic for websites such as google.com, youtube.com or mail.google.com.
User requests for low-volume traffic are typically served from Google core data locations, accessed via peering and transit. Low-volume user requests may also be proxied via GGC.
Data required to answer this type of user request isn't cached by GGC nodes.
The serving location (served directly from Google core, or proxied via GGC) is selected based on the following factors:
- Total application performance for end users (most important factor)
- Legal requirements
- Demand
- Available serving capacity
- Scheduled or emergency maintenance
- Configuration of your DNS resolvers.
Google doesn’t use measurements of ICMP ping latency in the selection of the serving location. ICMP ping latency does not accurately represent application performance for users.