Attendees: Many by VRVS at UERJ, Caltech, CERN, Internet 2, Manchester, AMPATH
In Jan 1, 2003 will have 2*OC48 (5Gbps) via SuperSINET to NY and peer with Geant, Abilene, ESnet, CANET, SurfNet, NorduNET, & Verio. Something about 4 of 5 Gbits/s is dedicated. IEEAF is contributing bandwidth to APAN which is an opportunity for HEP. APAN tried to get delegates from India & Pakistan without success, now getting from HEP society. APAN can be of help to collaborate within Asia, and also to US and Europe & other regions. In Japan universities use SuperSINET not APAN.
There is an APAN conference in Jan 21-25 in Fukuoka at which there is an opportunity for HEP to make presentations (probably via VRVS)
APAN NFS 5 year funded program ends next year. Connects 13 countries in Europe. Australia is a member.
Also presented information on 2 Korean A&R nets KOREN and KREONET.
GE commodity, emergence of 10GE, utilization of dark fiber (e.g. IN2P3, Caltech, SurfNet) & wavelengths
Achieves high stability with high throughput with standard MTUs with larege windows of over 14K packets. Only needs new stack at sender side, good for HEP with a few large sources of data.
This was a Land Bandwidth Speed Record submission. Yotta Yotta is a storage area wide network. Uses a Fiber Channel IP gateway to convert from wide area to disk. The gateway limits the data rate to reduce/eliminate the probability of dropped packets. The test was disk to disk data transfer. Got 93% utilization though a 4Gbit/s path. There are no routers. Each links had 4 streams. Was using TCP but since no packet loss there was no fall-back, since the path was dedicated to this (no cross-traffic) and you tell it how much data rate you can send, has large buffers and FC tells (via a credit rate mechanism).
Tom Kelley from U Cambridge testing a new scalable TCP stack.
The Juniper routers between Japan and CERN re-order packets and may cuase some performance problems.
There are problems at StarLight with power & cooling. Hope to have 10GE interfaces on Dell 6600 which mean some equipment will need disconnecting. The power situation should be solved early next year.
Now have the OK to use the backup link at 2.5Gbps for experiments.
SLAC have 2 OC12 (622 Mbits/s) link. We are sending Traffic has evolved so most of it is going via Abilene. This means we have low load on the ESnet link
They moved from USLIC to GEANT (at 1 Gbps) early this week. Can now send 100Mbits/s. Will increase to 200-300Mbits/s in January. Have requested BaBar physicists check
Link to CERN goes to 1 Gbps soon.
Send a list of countries/sites monitored to the email list.
Add the recommendation on working with ICTP to the report. Include that HENP typically leads developments for other sciences, and the need for outreach. Harvey mentioned the World Summit of the Information Society as a venue for publicizing the results, needs and recommendations.
We need a part of the monitoring report that relates to the needs of major HENP experiments, for example the needs of Tier 1 sites for data replication, in addition to the lighter weight needs for collaboration from smaller groups and individuals sites with poor connectivity. We could/should include the needs of a major experiment such as BaBar as a function of time driven by tier A centers as a natural feature of W. European countries and the US.
Might add/extend a comment on the role/need for new TCP stacks and their relationship to RTT and loss.
Need to address funding problem for PingER. Maybe approach Walt Polansky at DoE to extend the monitoring activity.
The committee needs to increase its activities in getting information to the funding agencies. Might request funding to put together web pages on the status of the digital divide, maps of networks, speeds and feeds, and performance by country & and region.
Need to provide information on best practices on how to achieve high throughputs. Demise of ATM, IP over SONET, IP over lambda, 10Gbits/s backbones moving to 40Gbits/s, causes imbalances, need to match the campus links to so not overwhelmed or unable to get throughput needed. Will need to add something on digital divide, this spans an enormous span of bandwidths. QoS is beginning to be offered by some networks, others simply add more bandwidth to avoid need for QoS. Includes new TCP development by people such as Tom Kelley, Sally Floyd, & Steven Low. There will be a section on packet switching and circuit switching and a discussion of dark fibers. Add information on the emergence of 10GE on the LAN. Add a short section on collaborative systems, UK has access grid and VRVS, and the need to improve such systems.
Will further develop reports, and open up to the whole committee. Need individual report by 2nd weel in January so can put together a full report for ICFA. The ICFA meeting is 13/14 February 2003. Next meeting will occur when we have a better report draft, probably middle to end January, tentative date Wednesday January 15 6am PST.