Last update:
Apr 1, 2000
|
|
|
|
The Data Access Layer for D0 Run II: Design and Features of SAM
|
Lee Lueking1,
Heidi Schellman2,
Igor Terekhov1,
Julie Trumbo1,
Sinisa Veseli1,
Matthew Vranicar1,
Richard Wellner1,
Stephen White1,
Victoria White1
- Fermilab
- Northwestern University
Speaker:
Victoria White
The Sequential Access Model, or SAM, is the data access layer which being built
for Run II data handling at D0. It requires the mass storage services of a
separate Run II component, the ENSTORE mass storage system, also developed at
Fermilab. SAM's goals are to provide a straightforward interface for users of
both batch and interactive applications, to provide a robust and easily
accessible database for locating and describing datasets, to archive
descriptions of data analysis projects, to promote efficient use of computing
and network resources, and to help optimize performance from the data storage
and delivery system. Although it has been built with D0's requirements for
Run II in mind, its design is not D0 specific.
A commercial RDBMS is used to maintain a carefully designed schema which
includes event, file, processing and required physics information. The
database chosen was ORACLE, since it was realized early on that a robust and
mature product was essential to the longterm stability of the system. Also,
the size of the database is anticipated to be several hundred GB, and many
features in Oracle are useful in managing a repository of this magnitude. The
schema has been designed to catalog event, file, various run and physics
information. It allows tracking data through a wide variety of complex
processing steps, and tracing lineage from detector or Monte Carlo files
through analysis stages. Information crucial to many types of analysis, such
as trigger, luminosity, and accelerator conditions are available through the
system.
SAM is a distributed system using a client-server architecture. The
architecture includes multiple servers which inter-communicate among each other
and to and from clients through CORBA interfaces. One of the principle servers
manages all communication with the database, another manages data delivery to
multiple clients on various processing platforms, and yet another server is
responsible for storing data into the system. Additional servers collect
information from the system at large, and provide other important functions.
The servers are written in c++ and Python. Clients are written using c++,
Python, and some web based GUI's employ Java.
The SAM system has been integrated with the Enstore, and D0 analysis systems,
and is being used for testing and D0 Monte Carlo data management. Its feature
list includes 1) data and data-description storage, 2) list-directed data
delivery 3) command line interface to all functions, 4) Gui interface to
applicable functions, 5)web-based database browsing tools, 6) disk cache and
buffer management, and 6)system monitoring tools. The system is documented and
being used by D0 physicists to perform simple analyses tasks. Several
additional features are planed over the next 6 months, including a facility for
extracting single events, additional error handling, and more support for
remote site implementations.
Presentation:  | Short Paper:  |
|
Bulletins
| Home | Bulletins | Committees | Scientific Program | Docs by topics | Social Event | Conference Location | Secretariat |
Privacy Policy |