ABC for You

ABC requires the following configuration to be installed successfully on your system. If you are purchasing a new system from Aspen, your sales engineer will ensure that the minimum requirements are met in order to install ABC on your purchased configuration.

• ABC requires IPMI to be installed and configured on every node in your cluster that you wish to monitor, including all head nodes, storage nodes, or any specialty nodes you might have configured. 3rd LAN or vampired connections are acceptable in all cases except the ABC server node. The ABC server node (the node running the ABC service, and the node that your web browsers connect to), must have a 3rd LAN IPMI connection if it is to poll its own health. This IPMI connection cannot be vampired, as vampired connections are not visible to the node that the connection is on.
• Additional peripherals such as RAID chassis, tape drives, network or low latency Interconnect switches, must be connected to switched PDUs that allow the equipment to be remotely powered on or off if ABC is to control power on these units.
• The ABC server node must have network connectivity to the IPMI interfaces of the polled nodes. If the IPMI network is configured as a dedicated administrative network, the ABC server must be connected to this network in order to successfully control the nodes IPMI interfaces. In some cases, this may require an additional Ethernet card to be installed on the ABC server node.

• The ABC server node and compute nodes must share at least one IP subnet and broadcast domain, and the ABC server node must be the only DHCP server configured for this broadcast domain.

• ABC currently supports the following RPM based distributions. Your cluster must be installed with one of these distribution to successfully deploy ABC.
  • RedHat Enterprise Server
  • SUSE Linux Enterprise Server
  • Centos
  • Fedora
  • OpenSUSE
  • Warewulf / Perceus
  • Scientific Linux
  • ABC does not currently support Debian, Ubuntu, or Rocks clusters.
• Root access to all systems on your cluster is required to install ABC. Two additional users must be installed on your cluster, aspensys and aspenusr. The aspensys account is used for privileged access to some system functions, and must be allowed to utilize sudo for privileged commands.
• Java applet support is required on all browsers accessing the ABC web

ABC offers many advantages to the beginning HPC administrator who has little experience managing an HPC cluster. ABC provides an easy, graphical way to see the current state and load of your cluster, and simplifies harder tasks such as node restore or package management which can be daunting to the beginning administrator. System alerts, combined with ABCs peripheral monitoring and standardized tools can greatly simplify the junior administrators life, making them more productive more quickly and easing their learning curve.

Combined with Aspens Standard Service package and Aspen remote access to your cluster if your security requirements allow, or even an Aspen VPN for additional security, ABC provides a painless coordination and technical interaction path between your junior administrator and our experienced Aspen engineers, who are there to help your administrator through the tough spots.

The mid-level HPC administrator will utilize even more of ABCs feature set to automate their routine tasks and keep their clusters healthy and at peak performance. ABCs customization features and plugins allows the mid-level administrator to automate event responses, react more quickly to system problems, and spend less time with day-to-day house keeping, and more time enhancing the cluster end user experience.

Manage me? Easy, ABC!

ABCs remote power and console capabilities combine with ABCs web proxy service to give your mid-level administrator quick and easy access to every system and peripheral in your cluster, along with the ability to painlessly control and manipulate the cluster as an integral whole. Access to the more advanced ABC management tool set, many of which are often written by individual administrators as they become more experienced, provides a springboard for your mid-level administrator to extend their HPC skill set, and do more with less.

The senior administrator can benefit from ABC on the cluster as well. Using ABCs advanced command line utilities to quickly and easily upgrade, monitor or troubleshoot systems saves time, and most of those tools are available for copying and modification, or as building blocks to use in performing even more advanced functions your senior administrator might wish to automate.

Does your senior administrator really need to waste time watching another node boot failure, or spend their valuable time troubleshooting node hardware? ABCs integrated node diagnostics (Not yet released!) can take the pain away by giving your senior administrator a standardized testing and quality assurance suite, freeing their time for the more important tasks they need to accomplish.

ABC can be used to provide an additional interface to your cluster, or even become the primary way your end users access your HPC resources. The browser interface that allows the user to easily see cluster loading and state can also provide command line and scheduler access for your end users.

ABC Scheduler View

Do your users need some unusual capability or application access, something that isn't quite the norm for an HPC cluster? No problem, your administrators or Aspen engineers can implement simple cgi programs to extend ABC to service unusual or site specific needs, then configure ABC to proxy and authenticate those custom applications to your end users.

Do your users have multiple jobs to run? They can submit jobs directly from the ABC interface (requires torque or Sun Grid Engine schedulers), as well as see current status of running jobs. Do they need access to an X11 desktop? ABC can be configured to proxy VNC sessions of dynamically built virtual desktops from specified nodes inside your cluster.

ABC can put the power of your HPC cluster directly into your users hands, giving them a standard interface and more flexibility, resulting in more computing productivity.