wiki:CmuLab-4Q08-status

CMU Lab Quarterly Status Report 4Q 2008

I. Major Accomplishments

A. Milestones Achieved

  • CMULab Bare Bones Node Configuration

The homenet nodes can boot and retrieve their operating system image remotely, on the Internet as well as behind a residential NAT gateway. We have customized an Ubuntu Linux image to support the homenet wireless and video capture hardware and to use the Emulab remote account and experiment management software.

We have integrated about half of our changes back into the ProtoGENI codebase.

  • Emulator nodes integrated

The laptops in the Emulator rack are now fully managed by Emulab software running on boss.cmcl.cs.cmu.edu. The changes to the ProtoGENI software to make this work properly (making Emulab work properly with nodes that have only one interface; creating hooks to power cycle the laptops using their wake-on-LAN capability) have been integrated into the ProtoGENI codebase.

B. Deliverables made

We have integrated about 2/3rds of the changes we have made to the ProtoGENI software with the main ProtoGENI development branch, as noted above. The remainder of the deliverables, in which CMU was contributing to design efforts headed by the Utah group, are documented in the Utah quarterly report and documented on the ProtoGENI wiki.

We have created internal documentation reflecting the mechanisms used for node control and boot management; if this is of interest to GPO or other clusters, we can turn it into something suitable for external consumption. In the next quarter or two, we plan to distill this (extensive and detailed) documentation into a level suitable for integration with the ProtoGENI docs.

II. Description of work performed during the last quarter

A. Activities and findings

Our major efforts this (partial) quarter focused on the individual node management software for the Emulator nodes (rack-mounted laptops for the wireless emulator testbed) and the HomeNet nodes (small form factor PCs with multiple wireless interfaces that will be deployed in residential settings).

Towards this end, we have completed the basic functionality needed to deploy, use, and upgrade nodes in the field. Much of this work was time-consuming but, in the end, straightforward; the largest amount of time taken was that needed for the CMU team to become familiar with the internal workings of the ProtoGENI software well enough to begin to extend it to support new boot mechanisms and hardware. As part of this process, we've committed temporarily to several design decisions for the HomeNet nodes:

  • Whole-image software updates only; no incremental updates.

Our experience with the previous generation of testbed management software with the RON testbed illustrated a surprising danger of providing easy incremental updates: the whole-image update mechanism can easily bit-rot, and the testbed maintainers lose the experience they need to easily keep new images flowing. We therefore are beginning the rollout of the HomeNet testbed supporting only whole- image software updates, and are beginning to design automation mechanisms to make this process easier. We are also streamlining the ProtoGENI image update support towards the same goal.

  • Each HomeNet node acts as an independent wireless access point

Because our initial HomeNet deployment is within a single administrative unit (a large apartment building consisting of many individual apartments that will host HomeNet nodes), we considered creating a single-ID wireless deployment that would allow users to move transparently from node to node. We have instead decided to opt for the simpler mechanism of having each node act independently, deferring issues of transparent mobility to research projects that will be run atop the HomeNet nodes.

We participated in GEC3 with a talk and a poster; we had fruitful discussions with several other projects.

C. Publications (individual and organizational)

None this quarter

D. Outreach activities

None this quarter

E. Collaborations

As documented in the Utah quarterly report, we have been participating in bi-weekly conference calls with the other members of our cluster. PI Andersen visited the Emulab team in Utah to coordinate on our development efforts. Our lead staff developer, Pat Gunn, has had extensive contact with the Emulab team as he immersed himself in the code.

We also have frequent discussions with the PSC team, regarding both their GENI activities for user opt-in and with respect to gaining access to the Emulab Internet 2 resources.

F. Other Contributions

Last modified 11 years ago Last modified on 01/05/09 16:00:00