At IRTF, we found another burned out FDDI card and a bad FDDI cable. Since the FDDI hot spare was already being used, we installed a card that was given to us by Proteon as a Demo a long time ago. This card was later found to have intermittent problems and has been swapped out since.
The ring is now operating at full counter-rotating mode.
- 4% salary increase and increased fringe benefit costs,
- allocation of $2000 of Hale Pohaku electricity charges to MKOCN,
- acquisition of a second spare for the CNX-500 and FDDI interfaces.
General administration charge has increased because of decrease in Constructiion Camp use, and resulting decrease in the camp's share of fixed admin costs.
- Overview of the Expanded Network
- Performance Comparision between ATM and FDDI
- Estimated Cost of the Various Pieces of Equipment
- Cost of Hot Spares and Maintenance
- Cost Sharing
Pui Hin Rhoads added the following comments:
- Routers versus Switches at the Edges
The proposed configuration uses routers at the edges rather than
switches to avoid getting into the situation of a great big emulated
LAN which will need to be centrally managed to different degrees with
the hardware and software available today. This is normally
considered to be an advantage of LAN Emulation. However, in our
situation, autonomy within each facility is preferred. In the
future, products which allow more division of network management
might become available.
As a reponse to Jim Wright's concern about quality of service (QOS), the central switch in the plan is a 5 gbyte switch. At the edges, facilities worrying about QOS will need to look into routers that will deliver QOS. Whereas all ATM switches claim to deliver QOS, only some routers do, for example, the Cisco router 7000/7500. I am convinved that with our applications and the amount of bandwidth that will be available to us, QOS will not be a concern for a while. However, this is a very important issue. Routers supporting QOS are more expensive than switches. The working group should make the final decision after it has had some time to look at the issue.
- Vender Selection In order to come up with a more realistic configuration, Cisco products are being used in the proposed configuration. Members of the working group should remember that the equipment is there as an example. We still need to make a decision as to which vendor to go with. I strongly favor going with a single vendor, whichever it might be, to avoid problems with interoperability now and in the future. Further, a single vendor solution will make it possible for the group to share spare parts.
- On the Cost Sharing In the last viewgraph, everything in dark blue will be paid for by the IfA and everything in cyan color will be owned by individual facilities. In the case where it will be easier and more economical, IfA will buy the equipment and be paid back by the facilities as a buy-in fee as they join the ATM network. An example of this would be a multi-ports ATM interface card for the central switch.
- Hot Spares Policy
Some companies, including Cisco, do not have maintenace contracts
that provide hot spares. One way to solve this problem would be for
the IfA to put up the money at start to buy the backup equipment.
The costs will be amortized over 3 to 5 years. The amortized value
for each year will then be shared by the facilities participating in
the ATM network during that year. To minimize the cost, we will not
buy a hot spare for every piece of equipment used by members of the
group. Rather, we will buy the least number of pieces of equipment
that will keep the facilities operating at a reasonable level in case
of an emergency.
This scheme can also be extended to cover equipment not on the common network itself. Some formula could be set up to pro-rate the annual amortized value according to the number of routers rather than the number of facilities participating in the scheme. In other words, a facility with three routers would pay three times the amount paid by a facility with only one router. Finally, all equipment must be on an advance replacement maintenance contract to be eligible for participating in the hot spares scheme. Obviously, it won't work if facilities have to keep the hot spares a month at a time while they get their equipment fixed.
As to the schedule of the ATM network implementation, JAC would like to see it in place by the middle of 1998, Subaru by next March, and Gemini by the end of this year. Our preliminary goal is to have the common equipment in place in a year. In the interim, Gemini can use two of our dark fibers to get to HP and the Internet.
In response to the concern that we will run out of fiber between HP and the summit communication room. Pui Hin responded that an ATM switch could be installed at HP and depending on the bandwidth demand at the time, an OC3 or OC12 link to the summit can be set up.
Andrew Pickles expressed concern about the over-heating of equipment in the Summit Communications Room. Ventilation and temperature control is part of the design effort.
