Points of presence
The Fastly network comprises a large number of physical servers distributed all over the world, connected to the internet at high density internet exchange points.
Sites and points of presence
Fastly servers are grouped into:
- "sites", which describes a co-located set of machines in a single physical facility, attached to the same internet transit; and
- "points of presence" (commonly "POPs"), which describes the clustering of multiple machines together to create a single pool of cache storage.
In many cases, all the machines that comprise one site also comprise one POP (i.e., POP == site), with the POP and the site sharing the same name. The largest POPs in densely populated metropolitan areas may span multiple sites, with locations and connectivity chosen to serve the same human population. See metro POPs.
In VCL services, the server.datacenter
variable reports the identity of the POP, while server.hostname
includes the ID of the site. The server.identity
variable includes the ID of both the site and the POP.
Due to the effects of clustering, a single request to a Fastly service may be processed by servers in more than one site, even while remaining within the same POP. However, transit between servers in different sites within the same POP is extremely fast, and comparable to the latency between servers on the same site.
If a service has shielding enabled, a request that is not satisfiable from cache will transit two POPs. With clustering also enabled (which is the case by default), the request may encounter a maximum of two POPs and four sites on its way to origin, as illustrated below:
Both POPs and sites use three-letter codes as identifiers, derived from the IATA airport code of a nearby airport. However, since POPs are rotated into and out of service regularly, codes are allocated based on availability and the location of the airport is not necessarily an accurate guide to the location of the Fastly site.
NOTE: Although we have historically used the term data center to refer to both a POP and a site, current use of the term "data center" may refer to POPs or sites depending on the context. Be sure you understand which concept is being referred to. The server.datacenter
variable, for example, refers to a POP.
A list of all POPs currently in service is shown below, and is also available via the API and the Fastly CLI. Our network map also shows locations of all current Fastly POPs as well as planned future POPs and POP expansions. Announcements of POPs entering and leaving service, or of substantial changes to POP capacity, are made on our service status page.
Metro POPs
In the case of most Fastly POPs, the POP occupies a single site, which has the same name as the POP. Where a POP spreads across multiple physical sites it is known as a metro POP. The following are metro POPs currently in service:
Location | POP Identifier | Sites spanned |
---|---|---|
Ashburn | IAD | KCGS720, KIAD700, KJYO710 |
Atlanta | PDK | KATL184, KFTY213, KPDK178 |
Chicago | CHI | KIGQ800, KLOT810 |
Dallas | DFW | KDAL212, KDFW821, KTKI862 |
Frankfurt | FRA | EDDF823, ETOU822 |
Sao Paulo | GRU | SBGR193, SBSP209 |
Seattle | BFI | KBFI740, KRNT730 |
Metro POPs have the largest cache capacity on the Fastly network and are therefore good choices for shield locations.
POP-specific edge behavior
If you need to know which POP is currently processing a request in edge logic, you can do so in both VCL and Compute services.
- In VCL, see the
server.datacenter
variable - In a Compute program, use the
FASTLY_POP
environment variable
Be aware that POPs are added and removed from the Fastly network regularly, and any logic created to vary the behavior of your service based on POP locations may need frequent maintenance.
Effects of POP and site variability
The design of the Fastly network balances issues such as provider diversity, connectivity, traffic volume, and optimum cache size. For Fastly customers, POP variability is most notable in its effect on cache hit ratio (i.e., the ratio of inbound requests that are able to be satisfied from cache).
For example, 100 requests handled by 100 distinct servers in 100 distinct POPs will experience a much lower cache hit ratio than 100 requests handled by 100 distinct servers all participating in the same POP because, in the latter case, all the requests have access to the same shared pool of cache storage.
Other features provided by Fastly are unaffected by POP variability, such as our image optimizer, which is available on all Fastly servers.
The effects of site variability are negligible and likely undetectable.
Complete list of POPs
The following table lists all POPs current active in the Fastly network: