|
WELCOME TO FRONTIER |
|
|
The Frontier distributed database caching system distributes data from
data sources to many clients around the world.
The name comes from "N Tier" where N is any
number and Tiers are layers of locations of distribution. The protocol is http-based and uses a
RESTful
architecture which is excellent for caching and scales well. The Frontier
system uses
the standard web caching tool squid
to cache the http objects at every site. It is ideal for applications
where there are large numbers of widely distributed clients that read
basically the same data at close to the same time, in much the same way
that popular websites are read by many clients.
For Large Hadron Collider (LHC) projects, the database is located at the CERN Tier 0, and the Frontier system distributes data to to all the Tier 1, Tier 2, and Tier 3 sites around the world. Within each site the Frontier system also distributes the data to all the worker nodes (typically hundreds or thousands) that need to read the data. The Frontier system was developed for the CDF experiment at Fermilab and is heavily used there. Fermilab also adapted Frontier for the CMS experiment at the LHC at CERN and it is used to access "conditions data" (mainly detector calibrations and alignments) at all its sites worldwide. The ATLAS experiment at the LHC later also adopted Frontier to distribute conditions data to all of its sites. The JUNO neutrino project based in China also adopted Frontier much later. Even though this web site is hosted at CERN for convenience of the LHC projects, Frontier is maintained by Fermilab. The LHC projects choose the level of abstraction between client and server to be SQL queries, but the Frontier system is designed with plugins so applications can put more of the logic on the server side, and that is how CDF uses it. There is also a plugin that reads from files on the local disk or other http web servers. The main advantage of using Frontier instead of directly using a http web server and squid proxy is that years of experience have resulted in Frontier supplying many features that are important for robust and flexible operation on computing grids. A summary of the features and components of the Frontier system is available in this overview webpage. |
|
|
| Source code and other project information for frontier client (implemented in C and C++) and frontier server (implemented as a java tomcat servlet) is freely available as open source, mostly under the Fermilab Fermitools license (a BSD license). The Frontier project maintains a squid distribution that has some big fixes and pre-configuration for supporting the Frontier application. Here's a brief introduction to the CMS Frontier architecture. |
|
|
These are the CMS Offline Frontier monitoring systems:
|
| For general discussions about Frontier subscribe to the frontier-talk mailing list. If you don't have a CERN account, on the login screen click on Create/Check your account and register for a Lightweight Account in the right column. Once you're logged in, click on the 'Join/leave group' link to the right of the frontier-talk search result. |
| Discussion also happens in the CERN Mattermost Frontier team. If you don't have a CERN account for that, click on "CERN Single Sign-On" and use "Guest access" in the lower right, click on "Register", and fill in your information. After verifying your email address, go back to the original link, use "Click here to sign in" and then on "Guest access" again, but this time login with your email and password. If you don't already have a Mattermost app you might also want to install that. |
If you have support questions for CMS or ATLAS Frontier, send a message to one of these:
(these are images to prevent spam so you can't click on them or copy/paste them, sorry). |
Some publications about Frontier:
|