Transport protocols and bulk file transfer

[Steven Senger <>]

We will be having our monthly call this friday Apr 17 at 1 pm Eastern. Dial-in info remains the same. 

To join both the web and audio conference (recommended), click here:
https://edial.internet2.edu/call/0154927
If you only want to participate in the audio conference, you may dial in with the following numbers:
Call: +1-734-615-7474 (PSTN CALL-OUT DOES NOT WORK) or +1-866-411-0013 (toll free US/Canada Only)
Enter access code: 0154927

The agenda will include

1) Review of previous action items. 

2) Consideration of a recommendation to complete the Phoebus deployment on the I2 backbone. 

3) Discussion of functional requirements for Phoebus data movement service. 

There have been a couple of ToDo's on my list for a while. First, I was supposed to point the group to the documents developed by the "Bulk Transport" instantiation of this group. Those documents can be found at

http://e2epi.internet2.edu/transport/

Specifically, the design space document is at

http://e2epi.internet2.edu/transport/docs/transport-design-space.pdf

Second, a couple of weeks ago I found time to get back to doing experiments with Phoebus and I have had some success. I have been working over a path that goes from my campus through WiscNet to the Chicago Phoebus Depot, to the Los Angeles Depot and then over CENIC to a server we have in Sunnyvale. I started working on linux with a simple pair of client/server test programs and found that Phoebus improves straight tcp data rates over this path by a factor of 3 or so. Hacking out the pertinent parts of the Phoebus simple_wrapper code I was able to get things working on OS X and got similar results. I then incorporated the hacked simple_wrapper code into the OS X client of one of my anatomical visualization apps (the Sunnyvale machine hosts the app server) and got a significant improvement. Without Phoebus the data rate is slow enough that some modes of the app do not give acceptable performance. With Phoebus, performance improves enough that the app performance is very usable. I can update the group in more detail on friday - I have some data and video I will try to post somewhere. I also have a number of questions about things I have observed in doing this. 

- steve