In response to Ryusuke Ogasawara, Pui Hin said she will try to get the 100baseTX going at the same time, but the first priority is to put the backbone in and to install equipment to support the 6 Mbps link funded by the UH HPC grant.
In response to Ryusuke Ogasawara's question on UH's plan to increase bandwidth between UHH and UHM, Pui Hin thought that the plan is to put in a DS3. Subsequently, Pui Hin clarified with UH and it is not clear yet whether it would be a DS3 or an OC3. Pui Hin also said UH has not committed MKOCN to the use of this link. Although if bandwidth is available, we might be able to use it; charges will have to be negotiated at that time. Henry Stilmack remarked that it would be useful at least as a backup.
In response to Dennis Crabtree, Pui Hin said the current plan is for the 6 Mbps to carry both the research and commercial traffic, the traffic will be separated at UH Manoa. Research traffic will be routed to the Mainland through the DREN and commercial traffic through the current GTE DS3. Pui Hin has discussed with UH the possibility of keeping the frame relay as a backup, but that has not been settled.
Towards this effort, Gemini met in Hilo with representatives from a number of the observatories as well as UHH. Subsequently, Bill Chen of UHH, Jim Kennedy of Gemini, David Lassner and Alan Whinery of UH, Bob McLaren and Pui Hin Rhoads of the IfA met to discuss the issues.
Pui Hin gave a summary of what she felt to be the important issues that were considered at the IfA and Hilo Meetings for further discussion. These issues are:
- Further increase MKO Bandwidth, hopefully to DS3
The combination of the UH and Gemini money will still fall short
of supporting both a DS3 to Manoa and a topology that allows
separation of Internet traffic from observatories' private links
to the summit.
Frame relay has been considered, David Lassner has expressed concerns about supporting meritorious applications in his proposal over very high latency link. David is checking on this.
Other options are still being considered. David Lassner is checking with GTE on what they can provide.
In general, the group considered the increase of bandwidth to the Mainland to be a much higher priority than a separation of Internet traffic from the observatories' private links.
- Lower hop count to the Mainland
Concerns over the current high hop count to the mainland was one
of the reasons for the preference to route through MHPCC rather
than UH Manoa. At the IfA meeting, David Lassner clarified that
UH will not be joining the DREN at MHPCC. That was a misconception.
UH will tunnel its traffic through the DREN from Oahu to the
Mainland. This will be a big improvement in terms of hop count
over the current situation.
Ryusuke Ogasawara remarked that a possible real reason for using MHPCC is that they already have the equipment to support the link and that will save money for us. However, Pui Hin pointed out that this is now a non-issue, since the UH has obtained the equipment to support the MKO link through the NSF grant.
- A topology that permits individual choice in separating
Internet traffic from the observatories' private links to the
summit
Pui Hin has learned at the IfA meeting that there is a need for
an architecture that allows separation of Internet traffic from
private operations traffic between summit and base.
Pui Hin presented a viewgraph modified from one presented by Jim Kennedy at the Manoa meeting. This is one of the diagrams presented by Jim Wright at the Hilo meeting.
The high light of this topology is the move of the Internet access point from the summit to the Hilo University Park. This move necessitates the adding of a link between the summit and Hilo to provide Internet access to the Waimea base camps. The modifications Pui Hin made include elimination of elements that are no longer valid, elements that are uncertain, and the end point of the Hilo-MK DS3. Where Jim Wright's diagram showed the link between Hilo and HP, Pui Hin now draws it between Hilo and the summit ATM switch.
Pui Hin sees a disadvantage for the Waimea observatory in this topology in that it will be more expensive for these observatories to provide a separate link.
Dennis Crabtree suggested adding a separate Waimea-summit link as part of the Infrastructure.
Phil Puxley remarked that the link is in one of the options proposed at the Hilo meeting; there likely are others.
Dennis also raised the question of post grant support for the added link but agreed to defer the subject for discussion with other cost sharing issues.
- Router location and Network Management From the standpoint of maintenance by UH personnel, it will be strongly preferable to house equipment in a UH building. Thus a Hilo router ought to go in the IfA building. Gemini has offered to house the equipment temporarily if the IfA building is not ready for occupancy when the money is available. Gemini will also provide full access for service work while equipment is at Gemini. The possibility of relocating the router after its installation suggests that the ethernet connection for Gemini in Jim Wright's diagram should be 100baseFL rather than 100baseTX.
- Post Grant funding and general cost sharing issues
It is not certain at this point what level UH will be
supporting the observatories at the end of the grant. It is
certain that even if UH will support the research traffic
from from the Big Island, there will still be the commercial
traffic to consider. One piece of good news is that UH has
signed a contract with GTE for 2 OC3 circuits to the Mainland.
It was suggested that shared costs be prorated among users according to their usage. Much of the costs, however are usage-independent: maintenance costs for routers at users' site, the central router is necessary whatever the usage, and support staff is required -- perhaps even more for the small organizations who have limited computing staff. If the router is in Hilo, the summit-Hilo link has to be acquired to provide Internet access for the Waimea base camps, so this link needs to be treated as infrastructure, i.e. a shared cost unless Gemini intends to continue paying for this link at the end of the grant. The costs of the Big Island to UHM link can be prorated, if we can figure out a way of measuring the traffic. This is complicated if there are multiple access points.
Phil Puxley remarked that Kennedy & Lassner have a tentative plan for ongoing support.
A general discussion emerged about the technical details that will best serve the community in terms of equity, efficiency, and cost effectiveness. Rather than trying to interpret the remarks, some of the comments follow:
Henry Stilmack:
If we terminate the off-island link at HP, it will be easier to measure packets.Dennis Crabtree:
Or terminate at the summit, run DS3 from both Hilo and Waimea to the the summit as infrastructure. Keck and CFHT would interconnect privately in Waimea.Henry Stilmack:
Yes, that's a good idea. It's fair to all.Phil Puxley:
Yes, it's a possible option.Pui Hin:
The private connection between Keck and CFHT is not cheap, a DS3 might cost more than $3000 since it's not covered by the IfA contract.Jonathan Chock:
Keck is not really interested in a separate Internet link.Ryusuke Ogasawara:
I'm skeptical of the advantage of the Ethernet connections from base facilities to a router. It's a dangerous topology because of the possibility of looping.General:
Yes, it also adds quite a bit of cost, since you have to have a high-end router to support possibly multiple DS3s and 100baseFL.Henry Stilmack:
Well, most routers can handle the traffic, but it's not clear that separating out the Internet traffic is really important. The amount of general internet traffic is quite small, in fact.Antony Schinckel:
What is the time frame for all of this?Phil Puxley:
The NSF money must be spent this FY.Dennis Crabtree:
If the Manoa link comes from the summit switch, then anyone who wants to separate their Internet traffic could add another summit link.Phil Puxley:
NSF is trying to foster Institutional connections.Pui Hin:
The Hilo router could be included in the proposal, as infrastructure. Groups of organization that need separate Internet link can then use the router, but pay for the link to the summit.Dennis Crabtree:
Summit hookup sounds like the best use of funds.Phil Puxley:
We need to cost out various options over the next week.Pui Hin:
For those who are concern with reliability, the summit is also a good choice. With the switching architecture, equipment failure at one site will not affect other sites. With a fully configured hotspare, all observatories can participate in switching cables when it becomes necessary..Henry Stilmack:
Summit staff can be trained to do this.Pui Hin:
The 88" shares the same building as the equipment room, they will be of great help.Don Mickey:
What about providing for future research park tenants? Does this need to be part of our consideration?Phil Puxley:
That's part of the UH interest in the Gemini proposal.Antony Schinckel:
But we have to be aware of possible bandwidth needs of such future users.Henry Stilmack:
The Hilo router would provide an access point for them, plus a new external connection--possibly through UHH.general:
Non-astronomy users may need to arrange separate connection.Pui Hin:
The Hilo router will also provide the Infrastructure for the educational telescopes. Links can be purchased when the need matures.Phil Puxley:
Hilo router layout is looking forward to needs like this.Pui Hin:
So should we bring the 6MB into the summit switch? Aside from cost effectiveness, it's also the most equitable topology, same number of hops for everyone, same opportunity to get a separate Internet link.general:
Good idea.This concluded the discussion. Following is a diagram of what this proposed network looks like:
Ryusuke Ogasawara asked how many DS3 are still available. I don't remember if I answered then, but for the record, three are currently used. There is an allocation policy agreed upon by all the MKO directors: One DS3 is allocated to each organization: CFHT, Gemini, IFA/IRTF, JAC, Keck, and SMA. One will be reserved for contingency. The remaining four unallocated DS3s may be sold to organizations which wish to buy a second DS3.
Pui Hin responded that there is no official policy, but it is obviously appropriate to limit large transfers that would saturate the link. This might not be a problem with the upcoming upgrade.
It was agreed not to implement a formal policy for now. Rather to monitor usage, and at the same time have observatories and users communities work together to make sure that observatories and users home institutions have compatible tape drives appropriate for large format cameras, for the transfer of data between the observatories and the user's home institution or to have observers schedule an extra day for tape writing.
Ryusuke Ogasawara said it will be possible to allocate bandwidth based on protocol with ATM. It works well and is inexpensive. We shouldn't be afraid of making use of the network.
