Bright Computing - Bright Cluster Manager

1. Install your clusters ... Easy
Use your clusters ... Easy
Monitor your clusters ... Easy
Manage your clusters ... Easy
Scale your clusters ... Easy
Bright Cluster Manager
™
Advanced cluster management was never so easy
Bright Computing
Bright Standard
Includes all standard
features for small and
medium Linux clusters.
Bright Cluster Manager Features
Based on trusted Linux distributions
Easy installation on small, medium and large clusters
Intuitive graphical user interface
Powerful, scriptable command line interface
Manage multiple clusters simultaneously
Integrated with workload management systems
Complete development environment for parallel applications
with compilers, libraries and debuggers
Built-in support for GPU computing with CUDA and OpenCL
libraries, and GPU monitoring and management
Multi-level node provisioning and image management for
unlimited number of images and nodes
Monitoring and visualization of many hardware and software
metrics, also allowing custom metrics
Flexible user and group management
Automated cluster management based on preemptive actions
when metrics exceed thresholds
Extensible cluster health checking mechanisms
Native support for redundant head nodes
Comprehensive documentation
...resulting in higher productivity and lower cost.















Bright Advanced
Includes all advanced
features for large and
complex Linux clusters.

2. Bright Cluster Manager™ greatly simpli-
fies the installation, management and
use of HPC clusters. With Bright Cluster
Manager, a cluster administrator can in-
stall, manage and use multiple clusters
simultaneously without the need for
expert knowledge of Linux or HPC.
Bright Cluster Manager provides the system admin-
istrator with a single system view while offering the
user an excellent environment for High Performance
Computing. It has been designed and proven to scale
to thousands of nodes, and to run on mission-critical,
production HPC clusters. Bright Cluster Manager has
also been designed to interoperate with common Grid
technologies and application APIs, and is compliant
with the Intel
®
Cluster Ready specification, which en-
sures compatibility with a wide range of Intel Cluster
Ready applications.
Bright Cluster Manager
™
—

Linux-Based Cluster
Management Software
Bright Cluster Manager has been installed on
hundreds of clusters, including some of the larg-
est and most complex clusters in the world, and
several TOP500 systems. Bright Cluster Man-
ager is available from resellers across the world
and is supported on most x86-based hardware
brands.
Based on Linux
Bright Cluster Manager is based on Linux and
is available with a choice of pre-integrated, pre-
configured and optimized Linux distributions,
including SUSE Linux Enterprise Server, Red Hat
Enterprise Linux, CentOS and Scientific Linux.
Easy to Install and Use
Bright Cluster Manager is easy to install and
use. It provides a single system view for man-
aging all hardware and software aspects of the
cluster through a single point of control. The
graphical user interface (GUI) streamlines the
administrative workflow by allowing all common
tasks to be performed through one carefully de-
signed interface. Furthermore, all tasks and con-
figuration settings are also accessible through a
scriptable command line interface (CLI).
Bright Cluster
Manager can be
used to manage
multiple clusters
simultaneously.
This is the default
Overview screen for
a selected cluster.
When selecting a cluster node in the tree on the left and the
Tasks tab on the right, the administrator can execute a number
of powerful tasks on that node with only a single mouse click.
Bright Computing

3. Image Management & Provisioning
Bright Cluster Manager includes a sophisticated im-
age mangement and provisioning system that allows
the cluster administrator to automatically install a fully
functional cluster from bare-metal within minutes.
A virtually unlimited number of images can be cre-
ated and assigned to as many different groups of nodes
as required. Incremental changes to images can be de-
ployed to live nodes without the need for a reboot or
reinstallation. For large clusters, multiple load-balanc-
ing provisioning nodes can be configured on the fly.
Monitoring
With Bright Cluster Manager a comprehensive set of
metrics can be monitored, visualized and analyzed.
Practically all software and hardware metrics available
to the Linux kernel and all metrics available to hardware
management interfaces, such as IPMI or iLO, are avail-
able. Custom metrics can also be defined.
Examples include CPU and GPU temperatures, fan
speeds, SMART information, system load, memory us-
age, network statistics, available disk space and work-
load management statistics.
Automated Cluster Management
Automated cluster management takes preemptive ac-
tions when set system thresholds are exceeded. Sys-
tem thresholds can be configured on any of the avail-
able metrics.
Standard actions include logging health state mes-
sages, sending emails, powering up or down nodes,
or running custom scripts. This system is highly
configurable and easily adaptable using the cluster
management GUI with built-in configuration wizard.
The cluster man-
agement automa-
tion wizard guides
the cluster admin-
istrator through the
steps of selecting
a metric, threshold
and action.
Cluster metrics,
such as GPU and
CPU temperatures,
fan speeds and
networks statistics
can be visualized by
simply dragging and
dropping them from
the list on the left
into a graphing win-
dow on the right.
Multiple metrics can
be combined in one
graph and graph
layout and colors
can be tailored to
your requirements.
User & Administrator Management
Bright Cluster Manager offers a role-based access con-
trol mechanism which supports an unlimited number of
administrators, whereby privileges can be defined on a
per-role basis. All user and administrator management
operations can be performed through the GUI or the
CLI. An LDAP service is included for user authentication
within the cluster. Integration with external LDAP serv-
ers is also possible.

4. Parallel Shell
The parallel shell allows execution of commands and
scripts across the cluster as a whole or on easily de-
finable groups of nodes. Output from the executed
commands is displayed in a convenient way with vari-
able levels of verbosity. The parallel shell is available
through the GUI and the CLI.
Workload Management
Bright Cluster Manager provides a user-friendly inter-
face for the monitoring and configuration of common
workload managers such as Grid Engine and Torque/
Maui. Various common tasks for manipulating jobs and
queues are available to users and administrators.
Cluster Security
Bright Cluster Manager offers a high level of security
which can easily be tailored to local requirements. Se-
curity features include a firewall, certificate-based au-
thentication to the cluster management infrastructure,
secure shell access and automated security updates.
HPC Development Environment
Bright Cluster Manager is not just a cluster manage-
ment environment, it also provides a comprehensive
HPC development environment for serial and parallel
applications.
The environment includes GPU libraries such as
CUDA and OpenCL, HPC middleware such as MPI librar-
ies, OpenMP, Global Arrays, mathematical libraries and
multi-core threading libraries. Development tools such
as compilers, debuggers and profilers are also included.
Environment Modules provide convenient mecha-
nisms for managing different versions of applications,
compilers and libraries simultaneously without creating
compatibility conflicts.
Advanced Features
Bright Cluster Manager is suitable for clusters of sev-
eral thousands of nodes and includes many advanced
features for large and complex clusters, including:
management daemons with minimal CPU and mem-
ory overhead which are synchronized to minimize
their potential effect on parallel applications;
multiple load-balanced provisioning nodes which
allow provisioning to scale to many thousands of
nodes;
redundant head nodes that can failover from each
other in case of failure;
extensible cluster health checking mechanisms;
centralized BIOS updates and configuration without
the need for keyboard or console access;
diskless slave nodes;
slave node booting over InfiniBand.
Comprehensive Documentation
A comprehensive system administrator manual and
user manual are included.







The status of clus-
ter nodes, switches
and other rack-
mounted hardware,
as well as one or
two metrics can be
visualized in the
Rackview. A zoom-
out option is avail-
able for clusters
with many racks.
The parallel shell
allows for simulta-
neous execution of
commands or scripts
on multiple nodes in
the cluster.
Bright Computing
Bright Computing, Inc.
2880 Zanker Road, Suite 203
San Jose, California 95134
United States
Tel: [phone]
Fax: [phone]
[email protected]
www.BrightComputing.com
Bright Computing Terms & Conditions apply. Copyright © 2009‑2010 Bright Computing, Inc. All rights
reserved. While every precaution has been taken in the preparation of this publication, the authors
assume no responsibility for errors or omissions, or for damage resulting from the use of the informa‑
tion contained herein. Bright Computing, Bright Cluster Manager and the Bright Computing logo are
trademarks of Bright Computing, Inc. All other trademarks are the property of their respective owners.