The Paracel
BlastMachine runs parallelized versions of NCBI BLAST, PSI-BLAST,
and MegaBLAST. It is an Intel-based cluster, currently consisting
of 22 dual-processor nodes. Each node contains two 933 MHz Intel
PIII processors and 2 Gigabytes of memory. The operating system
is a proprietary cluster variant of Linux. One dual-processor
node acts as the head node, scheduling and controlling jobs. 42
processors are available to run sequence search jobs. Communication
between nodes and with the fileserver is via a 100 MHz Ethernet
connection.
The BlastMachine accelerates the BLAST algorithms
by dividing a job across multiple processors. It can divide both
the database and the query among groups of processors. The number
of pieces a database will be divided into is determined at the
time the database is formatted on the BlastMachine. The BlastMachine
always divides a given database the same way to insure precise
repeatability of results.
The BlastMachine has its own queueing
and job control system. This system is accessed from
a host computer using the client program pb.
The system attempts to maximize total throughput;
different search jobs can and will run simultaneously.
The actual number of processors assigned to any given
job dynamically varies dynamically with machine load.
See the Using
the BlastMachine section for further details.
Command-line
and web access to the BlastMachine is available.
pb
is the command line interface to the BlastMachine. The following
usage information is based on the
pb command version 1.2.3 help output:
pb
<command> [arguments...]
For additional help on a command, type "pb
<command> --help".
Arguments to pb
are specified as --<flag> or --<parameter>=<value>.
For compatibility with other search tools, specific search commands
support arguments in other formats (for example -i
<queryfile>).
The following <command>'s are supported:
blastall
- submit a BLAST search
megablast - submit
a MEGABLAST search
blastpgp - submit
a BLASTPGP (PSI-BLAST) search
chgrp - change
the group owner of files and directories
chmod - change
the permissions of files and directories
chown - change
the owner and the group owner of files and directories
cp - copy the specified
PBFS file or directory
df - list file
systems and usage statistics
formatdb - format
a database for BLAST searching
killjob - cancel
the specified job
ls - list the files
in the specified PBFS directory
mkdir - create
the specified PBFS directory
mv - move or rename
the specified PBFS file or directory
rm - remove the
specified PBFS file or files
status - display
status information about the Paracel BLAST system
dbinfo - display
information about a BLAST database
shutdown - shut
down the Paracel BLAST system
The following arguments are applicable to most commands:
--config=<configfile>
specifies a file containing additional arguments
--help print help
and usage information
--host=<hostname>[:<port>]
specifies the name and port of the Paracel BLAST server
--priority=<priority>
specifies an integer priority for a job
--stats display
resource usage statistics to stderr
--quiet disables
display of status information to stderr
