Open Cirrus™ at Intel Labs Pittsburgh
This page contains directions for accessing the Open Cirrus site at Intel Labs Pittsburgh (ILP). If you are interested in using this site, please see the Open Cirrus web site.
Video
The Intel video wizards captured a short video of Michael Kozuch talking about the Intel Open Cirrus cluster at the 2010 Research@Intel day. [click here]
Papers and Presentations
- "Open Cirrus: A Global Cloud Computing Testbed," Arutyun I. Avetisyan, Roy Campbell, Indranil Gupta, Michael T. Heath, Steven Y. Ko, Gregory R. Ganger, Michael A. Kozuch, David O'Hallaron, Marcel Kunze, Thomas T. Kwan, Kevin Lai, Martha Lyons, Dejan S. Milojicic, Hing Yan Lee, Yeng Chai Soh, Ng Kwang Ming, Jing-Yuan Luke, Han Namgoong, IEEE Computer, April 2010.
- "Open Cirrus Cloud Computing Testbed: Federated Data Centers for Open Source Systems and Services Research," Roy Campbell, Indranil Gupta, Mike Heath, Steve Ko, Michael Kozuch, Marcel Kunze, Thomas Kwan, Kevin Lai, Hing Yan Lee, Martha Lyons, Dejan Milojicic, David O’Hallaron, Yeng Chai Soh, HotCloud '08, June 2009 (pdf)
- "Open Cirrus Overview," David O'Hallaron, (ppt)
Hosted Open Cirrus Projects
- Parallel Brain Activity Analysis
- SLIPstream
- Autolab - Autograding handin service for the cloud
- Register Allocation Deconstructed
Connecting Remotely (SSH)
To access the cluster, ssh to bigdata.intel-research.net.
If you are unable to get onto this machine, and you are a previous collaborator who used the cluster, please contact your Intel sponsor to get access to the cluster.
Equipment Setup
The Open Cirrus testbed(also called the bigdata cluster) is isolated from the ILP lab network. At the moment, all traffic must go through the NAT or be tunnelled in through the SSH frontend.
Available Storage
There are several types of available storage:
| Type | Access Point | Properties |
| Local | /x and /y (on a two-disk machine) or /[a-f] (on a six-disk machine) | Don't assume this is reliable |
| Archival | bdstore0n:/[xy], /mnt/glusterfs | This is archival storage on several data nodes using hardware RAID |
| DFS | hdfs://hadoop-master:8547/ | Nice parallel read/write preformance, integrates with Hadoop, can't modify files and limited VFS access |
Computational Systems - Hadoop
The main webpage for the Hadoop project is available here. We're currently running Hadoop version 0.18.0.
Short version: Again, once you've gained access to bigdata, ssh to hadoop-master. On this machine, cd into /usr/local/hadoop/hadoop to work with the most current version of Hadoop on the cluster. From here, you can launch a job, check the status of the cluster, and do other operations.
As a point of interest, it may be useful to forward out certain ports to your client machine, specifically 50030 and 50070. These ports are HTTP servers that provide Hadoop and HDFS status, respectively. An example command would be "ssh -L 50030:hadoop-master:50030 -L 50070:hadoop-master:50070 bigdata.intel-research.net". After this is done, point a browser on the client machine to "localhost:50030" to check the status of the Hadoop task tracker.
Computational Systems - Maui & Torque
In addition to running Hadoop on our cluster, we are running Maui & Torque - a scheduler and a parallel batch system. The current pool of machines includes the 40 nodes in the hadoop cluster and 30 other nodes.
A wiki for Torque is available here.
Other useful resources include the list of Maui Commands and the Maui Users Manual.
Short version: Once you've gained access to bigdata, ssh to maui-torque. From here, you can run qsub, qstat, diagnose, and other Maui and Torque tools. When using a Maui tool, you must "sudo -u maui" first.
Computational Systems - Tashi
Tashi is an Apache Incubator project that manages virtual machines across a cluster. The webpage is here.
Short version: From bigdata, ssh to tashi. On this machine, cd into /usr/local/tashi to interact with Tashi. You can create, destroy, migrate, suspend, and resume virtual machines. Currently, images are stored on an NFS server - merkabah:/export/tashi/images.