Frontier Squid Cache¶

The Frontier distributed database caching system is common in the HEP comunity as a way to distribute data to many clients around the world. The Frontier system uses the standard web caching tool squid to cache objects at every site, so that users access a cached copy much closer to their compute job, and the downloading from the server only has to happen once even when running. The Frontier caches are often used to help with distobution of content from CVMFS

Using the NERSC Frontier Cache¶

To use the frontier cache at NERSC you can include frontiercache.nersc.gov as part of your caching services.

Technical details¶

At NERSC there are two caching servers fs1.nersc.gov and fs2.nersc.gov which are setup to be used round robin as frontiercache.nersc.gov. The servers are monitored using the cern squid monitor, and can be found under with the name NERSC-Infra with the two coresponding servers fs1.nersc.gov and fs2.nersc.gov.

The caching servers are physical servers with 64GB of RAM and 120GB of SSD allocated to the cache. Each server is running four workers each, with the cache divided equally among the workers. These servers have 25Gb NICs for connecting to the HPC environment for compute related work, and 10Gb NICs for connecting with the outside world for downloading data.