Database Reference
In-Depth Information
The gLite 17 data management client tools allow a user to move data in and
out of the grid, replicate files between storage elements, * interact with a file
catalog, and more. High-level data management clients and services shield the
user from the complexities of the storage services and catalog implementations
as well as transport and access protocols.
The file transfer service (FTS) 37 allows for the scheduling of the transfer
of data files between sites. The service is configured to allow for transfers
only on predefined channels between peers configured at service startup. The
FTS uses low-level services and tools to perform data transfers and related
operations.
Another data management library worth mentioning is the grid file access
library (GFAL) . 18 It interacts with grid file catalogs and storage services via
the available control protocols. It allows applications to access files using ab-
stractions such as the “logical file name” (LFN), a human-readable identifier
of a file in the grid. Once presented with an LFN, the GFAL library contacts
a grid file catalog (LFC, the LCG File Catalog) to retrieve a handle to the
best replica available. Then, it negotiates with the corresponding storage ser-
vice to determine which file access protocol will be used (POSIX, gsiftp, rfio,
gsidcap, etc.).
Because of historical reasons and the untimely availability of general so-
lutions, the WLCG experiments have developed their own data manage-
ment frameworks to various degrees. For instance, a transfer service such
as PhEDEx, 19 developed by the CMS collaboration, could have evolved to
fulfill the role of the FTS, that is, scheduling not only within but also
between concurrent experiments. Instead, PhEDEx has been modified to
drive the common FTS from the CMS perspective. Another protocol for
ecient file transfer and access has been developed by SLAC. This is
the XROOTD 20
system that is being considered in particular for end-user
analysis. 38
3.4.2 High-Energy Physics Use Cases
Although there are different HEP experiments within the WLCG project, all
of them follow a common way of organizing their basic distributed computing
model. We first describe the general computing and data model that is appli-
cable to all four experiments and outline experiment specific differences later
whenever necessary.
* The term storage element is used to mean any type of grid-enabled storage system and the
software managing it, including multiple storage components such as disks, disk arrays, and
robotic tapes.
Search WWH ::




Custom Search