On November 14th, connection to I2/Abilene was down between 18:56 to 20:21 due to a faulty interface on the UH Juniper core router. The problem was resolved by moving the connection to a spare interface.
On the whole, the network has been reliable. From October, 2002 to September 2003, there was a total system-wide downtime of 20 hours. That makes 99.8% availability. Approximately 90% of the downtime is due to DoS attacks related to Microsoft vulnerabilities.
MRTG traffic graphs show that at our current level of usage, there is adequate bandwidth from the summit to Hilo, Hilo to Oahu and to the Internet. Also we are well within the 10 Mbps agreement we made with UH.
UH is just now seting up servers using the iperf tool to measure network latency and throughput. A system-wide test was performed between 11pm and 12midnight on September 26th. The initial results look good; more data is needed however to make a statement on the issues.
At the MKO users' Meeting, Pui Hin also presented the conclusion that the working group made concerning the usage of wireless at HP and its effect on the summit. On behalf of the working group, Pui Hin made the recommention to the directors that instructions on turning wireless capability on laptops off should be included with other instructions and recommendations to observers. Observatories might also want to post signs to remind the observers at their summit facilities.
Pui Hin asked the working group to help in developing a set of instructions. Antony Schinckel bought up the problem of cellphones. Signs should be posted at the visitor center to tell tourists to turn off their cellphones before going up to the summit.
From Greg Oi, Pui Hin learned that Verizon has determined that not all their interisland fibers are suitable for DWDM, this will make a difference in how much Verizon can bring down the cost of interisland services in the near future. However, other interisland services are available. Also, there is a chance that the 10Gbps link from Australia to the Mainland will land on the big island.
It will help speed up the contract renegotiation process if observatories can give Greg Oi an idea of how much bandwidth they will buy; this is not a committment. Pui Hin asked everyone around the table to give a number assuming FastEthernet (FE) to cost around the price of the current DS3 and GigabitEthernet (GE) to be $6k or under:
"Thanks for the information and thanks again for your support. Overall the demonstration was very successful. I have attached a couple of paragraphs describing the demonstration and the results. We did learn a lot from the demonstration, in particular, there were a few configuration/network issues that had quite an effect on the overall system performance. We weren't able to achieve as high a bandwidth as we would have liked to, but we were still quite happy with some of the results we obtained. We also saw a clear demonstration of the performance benefits of using high performance TCP stacks (FAST and HSTCP). These performed much better than traditional TCP over the high bandwidth paths (even with low packet loss).
"Internet2 Demonstration of e-VLBI over High Speed Networks
"The demonstration was highly successful in raising awareness of VLBI and e-VLBI. In the demonstration area, a large display was constructed which described VLBI, e-VLBI, participants in VLBI and the application of VLBI to astronomical and geodetic applications. Two Mark 5's were located in the demonstration area, both connected via optical fiber into a dedicated 1 Gbps stream that was connected into Abilene via a shared 10 Gbps trunk. Three laptop computers were used to display animations obtained from VLBI observations, video footage of the Westford telescope taken during a geodetic observing session and a real-time display of the application transport statistics.
"The demonstration was attended by a large number of people including people from the National Science Foundation and was well received. A demonstration was given of the transfer of VLBI data between the US and another International site using prototype RTP software. This software is representative of the next generation of VLBI transport protocols. A demonstration was also given of the transfer of VLBI data between two Mark5 systems located in the US. A maximum throughput of 207 Mbps was achieved. Demonstrators were also able to test memory to memory transfers between various sites and achieve a maximum TCP throughput of 460 Mbps using FAST TCP, 180 Mbps using TCP Reno and 400 Mbps using High Speed TCP."
Note: The meeting was subsequently re-scheduled for March 19th in order to have the meeting and the Sun/AMD Presentation on the same day.