Version 1 (modified by Vic Thomas, 11 years ago) (diff)


Internet Scale Overlay Hosting Progress Report 4/1/2010-6/30/2010)

1. Project Activities this Quarter

In this quarter, we continued the transition from primarily a development mode to an operational and support mode. Our main focus is now on ensuring that the SPPs are a reliable and effective tool for researchers interested in developing experimental networks. Most of the activities detailed below are contributing to this transition.

User Documentation

We have greatly expanded the tutorial material on the wiki over the course of this quarter. While there is still more to be done, we now have a solid base of user documentation to aid researchers interested in using the SPPs.

SPP Configuration Tools

We made a number of improvements to the SPP configuration tools during the last quarter to simplify the interface and make them easier to use. We completed work on these improvements early in this quarter.

New SPP Application

We have begun development of a new SPP application to demonstrate some core capabilities and provide a template for researchers interested in using the IPv4 fastpath. The Lightweight Flow Setup (LFS) application is a fast reservation protocol for IPv4 that is implemented using IP options and which triggers the reservation of bandwith along a path and triggers the configuration of packet filters and queues to ensure that a reserved flow gets the bandwidth it reserved. LFS chooses among alternate next hops if the default route to the destination does not have enough capacity to accommodate the reservation. If none of the alternate next hops has sufficient capacity, it makes a partial reservation using the "best-fit" link.

An initial version of LFS that runs only on the GPE has been completed. A version that uses the IPv4 fastpath is currently being tested and debugged.

We expect to demonstrate the fastpath version at GEC-9 and to incorporate it into our SPP tutorial.

GENI Experimenters' Workshop

We partiticpated in the GENI experimenter's workshop at Princeton in June. This leveraged material that was developed for the GEC 7 tutorial, but still required significant preparation. The workshop participants showed a strong level of interest and we expect several groups to start using the SPPs in the coming months.

Improving System Stability

The SPPs have been in operation for over six months now, but we have been experiencing stability issues that require that the deployed systems be rebooted approximaytely every three days. This problem has never manifested itself in our lab systems, so we have had to carry out the diagnosis on the deployed units. We have isolated the problem to the linecard software than manages packet buffers and buffer descriptors. We expect to get it resolved in the next month.

Flow Monitoring

The flow monitoring system is essentially complete, but we have suspended deployment in the running systems as we track down the linecard bug mentioned in the previous item.

While it's now seems likely that the flow monitoring code is not responsible for the problem, we have continued to disable it, to simplify the debugging process. It will be important to get this back in operation as the usage of the SPPs grows.

Development of Second Version of Network Processor Datapath Software

We continue to make slow but steady progress on version 2 of the NPE software. We are now in the process of verifying the software on the SPP hardware (as opposed to the IXP simulation enviornment). We now expect to start deploy version 2 by the end of 2010.

GEC-8 Preparation and Attendance

Preparation and attendance of the GEC consumed about one staff-month of effort. This consititues roughly half of the supported activity during the past quarter.

2. Milestones achieved

Nothing new

3. Deliverables made

Nothing new

4. Project participants

Jon Turner – PI (5% support level)
Patrick Crowley – PI (unsupported)
John Dehart – technical staff (45% support level)
Ken Wong – technical staff (15% support level)
Mike Wilson – graduate student (unsupported)

5. Publications (individual and organizational)

None yet.

6. Outreach activities

Tutorial at GEC-7

GENI Workshop in Princeton

7. Collaborations

Have established connections between SPP nodes and the GPENI and MAX networks, but still need to fully test the connections and verify the performance matches expectations.

Have begun using Plush/Gush tools for configuring SPP nodes.

Have demonstrated experiments that use combination of SPP and PlanetLab nodes.

Have had initial discussions with Jim Griffoen of the University of Kentucky about using SPPs in support of his Post-Modern Internet Architecture.

Started discussions with Brighton Godfrey of University of Illinois on pathlet routing experiments.