Extreme Cloud Administration Toolkit (Xcat) - IBM NeXtScale System Planning And Implementation Manual

8.1 eXtreme Cloud Administration Toolkit (xCAT)

The eXtreme Cluster Administration Toolkit (xCAT) 2 is an Open Source Initiative
that was developed by IBM to support the deployment of large high-performance
computing (HPC) clusters that are based on various hardware platforms. xCAT 2
is not an evolution of the earlier xCAT 1. Instead, it is a complete code write that
combines the best practices of Cluster Systems Management (CSM) and
xCAT 1. xCAT 2 is open to the general HPC community under the Eclipse
License to help support and enhance the product in the future.
xCAT provides a scalable distributed computing management and provisioning
tool that provides a unified interface for hardware control, discovery, remote
control management, and OS diskfull and diskless deployment.
xCAT 2 uses only scripts. This makes the code portable and modular in nature,
which allows for the easy inclusion of more functions and plug-ins.
The xCAT architecture includes the following main features:
Client/server architecture
Clients can run on any Perl-compliant system (including Windows). All
communications are SSL encrypted.
Role-based administration
Different users can be assigned various administrative roles for different
resources.
Stateful, stateless, and iSCSI nodes provisioning support
Stateless nodes can be RAM-root, compressed RAM-root, or stacked
NFS-root. Linux software initiator iSCSI support for Red Hat Enterprise Linux
(RHEL) and SUSE Linux Enterprise Server (SLES) is included.
Systems without hardware-based initiators also can be installed and booted
by using iSCSI.
Scalability
xCAT 2 was designed to scale to 100,000 and more nodes with xCAT's
Hierarchical Management Cloud. A single management node can have any
number of stateless service nodes to increase the provisioning throughput
and management of the largest clusters. All cluster services, such as, LDAP,
DNS, DHCP, NTP, and Syslog are configured to use the Hierarchical
Management Cloud. Outbound cluster management commands (for example,
rpower, xdsh, and xdcp) use this hierarchy for scalable systems management.
An example of such a hierarchy is shown in Figure 8-1 on page 215.
214
IBM NeXtScale System Planning and Implementation Guide