High-Energy Physics Applications on the Grid - Grid Computing: Infrastructure, Service, and Applications - page 433

In the case of ALICE, the system conveniently couples their grid back
end (AliEn) with the ROOT framework [14] (at the basis of their C++ frame-
work ALIRoot). The key component is a very efi cient gateway (a service
used by multiple users) to deal with user commands. This service caches
the authentication state of the clients in order to provide efi cient access
for interactive users. This pattern (described in the paper for the original
ALICE implementation [15]) is actually used more and more in different
areas of the grid middleware since it couples the strict security standards
needed by the grid (basically the usage of X.509 security) with the respon-
siveness needed by any interactive application. This mechanism avoids
excessive load generated by security at the server level since the server
does not authenticate all the clients at each interaction but it basically dele-
gates this to (a set of) trusted services. In particular, the searches in the
ALICE (AliEn) i le catalog can be done in a transparent way from the user
prompt and from ROOT with high efi ciency (also implementing features
like i lename completion, etc.). Again, the complexity of the sophisticated
solution to provide simple and efi cient access is hidden.
In the case of CMS they developed CRAB (CMS Remote Analysis
Builder) [16] an application which is somewhat similar to Ganga. In the
original form it was basically a client tool helping the user to submit and
control jobs on the grid via a convenient set of commands and tools. More
recently the usage of an optional server has been introduced allowing
“disconnected operations” like, for example, automatic intelligent resub-
mission while the user is actually not connected.
CRAB is a also very successful application in terms of user response.
In 2007, 20,000 jobs per day (with efi ciency exceeding 90%) have been
executed by CMS users, making CRAB the most intensively used tool in
the HEP grid environment. In the next chapter, we display a snapshot of
usage of CRAB in Figures 17.6 and 17.7 .

17.5
Monitoring is a vital component in a distributed system. Grid projects had
to invest considerable effort, in particular, when entering a production
phase. The HEP community contributed to this effort, building on previ-
ous experience and adding innovative contributions.
It is clear that a tool like Ganga does not prevent execution problems if
these are connected, for example, to a misconi gured site or to a failure in
the middleware stack. Such investigations need monitoring information.
As a matter of fact, all the different actors in the grid world (operation sup-
port, middleware developers, individual users, application managers)
need easy access to the available information.

The Dashboard and the Grid Reliability Tools