High-Energy Physics Applications on the Grid - Grid Computing: Infrastructure, Service, and Applications - page 439

17.6
Data management is particularly interesting in the case of HEP. In this
case the quantity of data (every year several PB of data have to be added to
the data store), the replication strategies (multiple complete copies should
coexist over the LCG infrastructure to provide redundant storage and
data access), and the complex access patterns (especially at the level of
end-user analysis) make data management a very interesting problem.
ARDA invested a lot in this i eld, starting from middleware tests to moni-
toring activities. For example, a very important part of the Dashboard
monitors data transfers at the level of the infrastructure services and at
the level of experiment-specii c steering systems.

Data Management

17.6.1 Storage Resource Manager

Due to HEP-specii c requirements (actually much older than the grid idea)
the dei nition of a standard to interface to mass storage has a long history.
In recent years this problem has been discussed in the context of the Open
Grid Forum (OGF), which led to the dei nition of SRM (Storage Resource
Manager). The adoption of SRM within LCG considerably accelerated the
convergence on a workable standard implementation. The deployment of
a nontrivial infrastructure of SRM and the operational experience will in
turn be essential in the further evolution of the SRM concept.
The complexity does not only depend on the difi culty of the perfor-
mance required (data size, number of i les, etc.) but also because SRM is
effectively an interface to be implemented by the different mass storage
systems supported and in use in the grid computer centers. LCG sites use
four systems, namely CASTOR (notably working at the Tier0 and in three
Tier1s), dCache (in use in most Tier1s), StoRM (at the Italian Tier1 and
under consideration in other centers), and DPM (essentially deployed at
Tier2s). Details of the different implementations can be found in [22].
The experiments' requirements are satisi ed with the SRM version 2.2
which is being deployed and now (beginning of 2008) over 160 endpoints
are becoming available for the last round of readiness tests before the data
taking (CCRC'08). Very much like the operations of the i rst services in
LCG back in 2003, this is a proof-of-existence of the viability of the SRM
solution to deliver the base data storage layer for LCG. It is clearly a start,
since this area is in constant evolution, but the fact that this infrastructure
can be actually operated by shift crews and a good service is delivered to
users is clearly very encouraging.

17.6.2 File Transfer Service

An example of a high-level service built on the existing data infrastructure
(and developed in close connection with the HEP community within the