ABC Features

---

ABC functions as an SSL encrypted web portal to access your cluster for day to day user and administrative functions. You can access your cluster from your organizational network, or even the Internet, and use ABC to monitor your cluster, submit jobs, query statuses, and perform many of your daily cluster tasks, all from any java equipped web browser.

---

---

ABC utilizes PAM modules to enable systems level authentication to your clusters underlying user authorization schema, including LDAP, NIS, or Unix password and shadow entries. If a user has an account on your cluster, then they have an ABC account as well.

 

User and Group Admin

 

Within ABC, users can be grouped into any arbitrary group you wish, and specific permissions can be assigned to that group to fine tune what your users can see or control. ABC defaults to having two groups, Administrators, and All Users, but you can define and utilize as many security groups as you wish. Every user is assigned to the All Users group, and root is automatically part of the Administrators group.

---

---

ABC requires IPMI to provide console access directly to any node in your cluster. If IPMI KVM over LAN options are installed on your nodes, ABC allows the video display from the node to be accessed from your web browser through ABC's web proxy service. Serial over LAN (SOL) access is supported and proxied as well. If you simply need terminal access to a running node, use the embedded ABC terminal from your web browser to access any node on the cluster your permissions allow.

 

KVM over LAN Console, BIOS access

 

ABC allows you to directly access any cluster nodes console to troubleshoot or repair the node, set BIOS, change a boot order, or perform other system level functions. Perhaps you wish to boot the node with an iso or floppy image. With virtual media, you can attach that image to a virtual device on the node and boot the system using that image.

 

You no longer have to be in the computer room to troubleshoot a problem node, you can use ABC from the comfort of your office, your home, or even a hotel room half way around the world. (network security policies and routing must allow access to the ABC server web interface from your location.)

---

---

ABC uses IPMI to remotely power off, power on, or power cycle any of your nodes or any group of nodes from the ABC interface. Do you have power outages? Set your ABC control node to power on automatically so that you can access ABC to perform a controlled power up of your cluster without being there physically.

Switched PDU Control

 

For nodes and peripherals that are not equipped with IPMI, ABC can access and control switched PDUs that allow the equipment to be remotely powered on or off.

 

By seamlessly integrating both IPMI and switched PDUs, ABC can give you remote power control of every system in your cluster. This capability has long been a holy grail of HPC cluster administration. (Your cluster must be installed with switched PDUs connected to every peripheral device.)

---

---

ABC provides additional tools to automate common administrative tasks across your cluster. Using the ABC Package Management function, administrators can inventory the installed packages in all nodes, compare package versions within the cluster, and even synchronize the installed packages across the entire environment.

 

Node Detail

ABC also provides a Plugin capability, which allows commands to be ran in parallel across some or all of the cluster nodes. Plugins can be simple shell scripts or complex programs that are added by Aspen or your administrators to automate common tasks you might need to perform.

 

Of course, ABC is also a monitoring tool, automatically polling all your nodes for both hardware and software health and providing displays of this data in different ways. Both real time and historical data collection and display is supported, and each value threshold is configurable as well as the alarm action tied to that value. This allows every ABC installation to be tuned to the specific environment that your cluster is in to satisfy your specific needs.

---

---

If any of your cluster nodes use disks and an installed O.S., ABC provides the capability to easily and quickly restore that node from an image. ABC can snapshot one node, then use that image to restore another, changing node specific configurations so that the newly imaged system returns to service with the correct identity and configuration.

 

A disk failure on a compute node no longer means hours of work re-building the system and re-configuring it to function in your cluster again. Replace the failed disk, select the image you want restored to the node and the node you want restored, and let ABC do it for you.

 

Recovery from a catastrophic hardware failure is easy too. You can utilize images taken of your master, storage, compute, or any other type of node to restore to another system should a failure occur.

 

Need to perform a major upgrade, or radically change your node configuration? Upgrade and customize a single node so that it is set up exactly the way you want, then copy the node and deploy that image to all the other nodes you wish upgraded.

---

---

Our engineers don't want us talking about this yet, but its such a cool feature that We just have to! One of the most time consuming tasks a cluster administrator faces is repairing a node. Aspen utilizes our AIME diagnostics and quality assurance package to ensure that our delivered systems are of the highest quality, and AIME is now coming to your cluster, integrated with ABC.

AIME Node Diagnostics

 

AIME runs multiple automated tests on a nodes hardware, including disks, memory, CPUs, and networking components. AIME will soon be integrated into ABC to perform this same function for installed nodes in the field. If you have a node failure, or a node that just appears to be a little slow or cranky, you'll be able to put the node into AIME testing to determine what the problem is directly from your ABC web interface. (and the command line too. your friendly engineering team)

 

AIME will show you the results of these tests, and if there is something wrong and your security allows it, open a support ticket directly with Aspen support, detailing all the tests ran and their results so that we can quickly replace hardware components or take corrective action. See? We said that this was a cool feature!

---

---

ABC isn't just for your administrators, it simplifies your end users lives as well. If you use either Sun Grid Engine or Torque/Maui , your cluster users can submit and monitor their jobs directly through the ABC interface.

Scheduler Monitor

 

ABC can be used to proxy VNC servers for your individual users X server needs, as well as provide access to Ganglia pages on your cluster. Many users find Ganglia familiar, and helpful when studying the effects of particular code runs on their cluster.

Submit User Job

 

And of course your users have access to all the monitoring and real time status tools ABC provides, allowing them to quickly see how much of your cluster is loaded, and if enough nodes are free for a particular job. If command line access is needed or desired, users can be allowed to open an ABC SSH terminal to the master or other nodes to perform code compilation or other tasks directly out of their web browser.

---

---

ABC monitors your clusters health, and provides system alerts when things go wrong. For instance, your system and CPU temperature in every node is monitored, and an event is generated when maximum temperature is exceeded. The maximum temperature value is totally configurable within ABC, although ABC comes with reasonable defaults for each of these settings.

Temperature Event

 

You can define custom actions for each event as well. In some facilities with frequent cooling outages, Aspen customers have ABC power off nodes that exceed a maximum temperature and notify them of the problem. This can save the nodes under ABC control from continuing to operate under high heat conditions, and remove heat load to help save other less well-monitored nodes within the facility.

 

ABC can be configured to perform periodic S.M.A.R.T. checks on all hard drives in your cluster and report back the results, removing the need for manual checks or cron driven scripts to perform the same function. ABC also monitors free space on your partitions, and warns you if the percentage of disk usage you set is exceeded.

 

ABC monitors many other things about your cluster, including system fan status, motherboard voltages, and other critical system values by default. It also integrates with UPS systems that service your cluster. System warnings are also generated in the event of power events, and you can define who you want notified when critical system events occur.

---

---

HPC clusters normally operate with all the compute nodes and peripherals connected to an internal network. ABC allows the web interfaces of internal peripherals, such as node IPMI interfaces, RAID devices, UPS systems, PDUs, tape backup units, environmental monitoring systems, cooling system interfaces, and other devices to be integrated directly into your ABC web interface. Additional web services on your master or any other internal node are easy to proxy as well.

ABC Web Proxy for Ganglia

 

ABC does this by securely web proxying the internal network web output of the device through ABC and to your browser. This allows you to seamlessly integrate additional peripherals as needed, and access and authenticate them through a single ABC interface.

 

This capability makes ABC customizable and extremely extensible. Not only can additional device web interfaces be added, but also additional custom web pages that can be used to perform site specific tasks. One Aspen customer utilizes this capability to operate a web control panel accessed only by a specific user group. They in turn use this interface to control and boot specific cluster nodes into different environments. The possibilities are endless, and allow you or Aspen to customize your ABC environment to your specific needs.

---

---

What if you're an experienced administrator or a command line zealot? Is ABC for you?

 

Many experienced HPC engineers and administrators prefer using command line utilities to graphical user interfaces for speed and scriptability, and Aspen engineers are no exception. ABC comes with many command line tools specifically targeted to the more experienced administrator. ABC CLI tools exist to remotely control power on all nodes, retrieve sensor data, and even to automatically restore a node!

CLI Tools

 

We may not have all the functionality of our web interface replicated to command line utilities, but we're working on it. Tell us what function you want to see. We may already have it, and if not, we have plenty of engineers who love to add more cool features and command line utilities.

 

Most of our command line utilities are written in PERL, bash, or python so that you can easily see what they're doing, and modify or copy the code to fit your own preferences or customizations. We love to get your feedback. If you want a command line tool that we haven't written yet, it's likely that another customer would like that tool as well, so we'll be more than happy to put your tool on our feature request list.

---