A new problem has developed about a week ago: the link between CFHT and HP has been going down at intervals. Pui Hin Rhoads and Don Mickey will be over next week to look at the problem.
On Feb 25, as requested by Torben Nielsen of the UH, the full T1 between HP and UH Manoa was replaced by a T1 frame relay circuit. Since frame relay does not support RIP broadcast, sites need to add an additional static route to the class B UH subnet 188.8.131.52, netmask 255.255.0.0 to get to UH Manoa. After some teething problems, the new circuit is working ok. No significant performance change has been observed.
Installation of all MKOCN and requested DS3 fibers have been completed. Termination of MKOCN fibers for SMA (at Subaru) and Gemini (at the generator room and Gemini) and all HawTel DS3 fibers are scheduled to happen in the next two weeks.
Requirments are as follow:
TO:Laser guide traffic control computer (initially, this will be at Keck)
About 1 kbyte per telescope, 0.1-1 Hz.
Each camera produces a 640x480 image (1 byte/pixel). The cameras do not have hardware compression. 1 Hz readout is desirable (600 kbyte/sec) but 0.1 Hz is probably acceptable, switching to streaming at 1 Hz if something is detected.
Latency - up to 1 second is probably OK
Reliability - item 1 MUST be fully operational for laser guide stars to be projected. Item 3 could, in case of a failure, be replaced by people placed outside to spot planes.
Of the four types of monitoring, only Item 3 is bandwidth demanding. The group decided that the requirements are well within the performance of the FDDI. Pui Hin suggested that we can run some simulation just to verify. Other suggestions are:
In addition, if the current FDDI (shared 100 mbits/s) is replaced by ATM (switched 100 mbits/s), the requirements will easily be met.
On the issues of emergency backup, Pui Hin reported that HawTel does not provide ISDN service on Mauna Kea. Jonathan Chock suggested that Keck might be able to help. Keck has an extra T1 port which can be used to route the MKOCN internet traffic to Keck's Waimea headquarter where ISDN service is available. However, ISDN is also not currently available to MHPCC, which seems to be the most logical alternate Internet access provider.
Pui Hin pointed out that we might learn something from the upcoming BOF session for remote observing at the SPIE.
After some discussion, the group decided that what we need is something like an intercom system or a paging system. Jonathan Chock volunteered to look into it.
Note: Since the meeting, Clayton Yugawa has proposed an alternate system.
This system will cost each participant an one time installation fee of $50.00 and a monthly fee of $17.20. There might be additional intallation charges at HP.
For this purpose, Pui Hin has provided the prices for equipment that we are interested in. According to Pui Hin's calculations, with two partcipants, cost per participant is about $14k a year. With three participants, the cost will drop to $9k a year. These numbers include only the maintenance cost for the public equipment and hotspares. Each facility still needs to pay for the maintenance of their router or routers - about $3k for each router directly through Cisco.
Currently, facilities with Cisco equipment pay about $4k to $6k a year for the maintenance of each Cisco router through HawTel plus another $3k for the maintenance of the public Proteon equipment. The HawTel maintenance includes the use of a hotspare unit.
Pui Hin pointed out that the cost per participant with three participants is really not far from what the facilities with Cisco equipment are currently paying.
Instead of buying our own backup unit, Jonathan chock suggested buying maintenance for the public equipment through HawTel which includes the use of a hotspare unit. This will simplify the cost sharing.
Henry Stilmack however would prefer to do the maintenance ourselves. He thinks we might be able to do a better job than Hawtel if hotspares are kept at the utility room on the summit.
Pui Hin will make further analysis and hopes to continue the discussion through email.
Again even though there is currently no performance requirement for ATM, ATM remains attractive for its switched architecture. Other than the central switch, equipment failure at one site does not affect other sites. This will make trouble shooting a lot easier. Further, the Proteon equipment has been failing more often than before. Since Proteon is no longer manufacturing the CNX500, the situation will only get worse even with a maintenance contract.