6/18/2010 meeting with c...@columbia CS conference room attending: richard, clint, quy, hitae, daisy, eric, paul, shlomo, micah
CRF was very accomodating and helpful, we had a very productive meeting covering many issues, and also had a tour of their machine room and 'coffee room'. They have an impressive number of machines for various purposes and classes, and a quality simulation setup for their students. Overall networking architecture overview ======================================== uplink rates, peering partners, internet2 ------------------------------------------ uplink from davis/CEPSR to CUIT edge router: 100mbit/sec; from edge to elsewhere is probably 1gig/sec. MUDD ports are planned next month to be gigabit, also CS ports. CEPSR ports will not change. TODO: (ask CUIT about this) - we did ask and CUIT didn't know anything about these port upgrades the "coffee room" (computer room) is gigabit switches with 10gig backbone (internal) then multiple switches to patchpanel limits, monitoring and caps --------------------------- we will be connected to a CS gig switch patched directly to the wall, so we wont saturate CS' network. worst case video streaming: ~29Mbit/sec from each room (if we can do two streams from davis and interschool lab each room). The video streaming is not intended to be live, just for live editing, if this can be done over the network to the server in the coffee room, that would be ideal. The live stream would be much less (300k or so) CRF said that 30mbit/sec would saturate their network: is there a 10gig backbone from CUIT that can handle this? If we can get two 30mbit streams from CEPSR to CS, their machine room can handle it. Discussion about considering running some cables from coffee room to lounge for video. will look into ceiling runs, but if that doesn't work, on the floor. Will confirm this next week physically we will be on a separate switch in the coffee room, so we wont be on their local network and not interfere. We will be running via the buffalos so we will be limited based on their capacity in the first place. Shlomo will have this worked out next week. Paul recalls last meeting with NOC and they were looking at uncapping a number of IPs for this conference. Also that CUIT is looking at making some policy changes for the conference. There are some uncapped IPs that CRF can loan us for the outgoing streams. Paul raises the point that an internal cap might need to be set in CRF to safegaurd the local connectivity policies -------- Basically as long as we are good netizens, we are fine. The CUIT policies apply, which basically means no copyright material can be illegally downloaded and no port scans. ipv6 ---- CRF is only deparment that is using IPv6, but CUIT has it. should be working, if only within the CS network (people connecting to the columbia wireless will *not* have it). our private subnet will have ipv6 multicast --------- NOC has it enabled, but nobody is using it TODO: talk to NOC about this cabling / switches ------------------ do we need our own, can we use our own? there are plenty of stand-alone 8 port switches that could be loaned to us to use for the lounge and carlton cafeteria All the wired jacks that we will be using will have have caps, any servers that we would not want to be capped would need an uncap request filed with CUIT TODO: CRF will check what equipment they have, they seem to have plenty available that they are happy to loan us are there any cabling considerations we need to be aware of? ------------------------------------------------------------ As long as there isn't haphazard cabling that people can trip on, there is no issue here. wireless ---------- We talked about the wireless coverage and deadzones, its mostly was determined that there are a 'lot of deadzones', but we can always put a buffalo in any of these once we know where they are. The wireless routers they have can handle approximately 20 clients at a time, there are no other restrictions except for the per-ip bandwidth caps that are everywhere. We discussed a plan to use the buffalos to setup in the different areas that we will be using (hacklabs etc.) and put them on a single virtual lan and bridge them (rather than use the columbia wireless). We will then NAT those through uncapped IPs. * network services ------------------ We discussed the Columbia Debian mirror and how it is on the CUIT network, so it might be a concern if everyone is having to traverse CRF's network to get at the mirror. We talked about having a local mirror to not clog up the CRF upstream network to CUIT, which will have to be a public IP which will be patched straight to the CUIT router so it is one hop away from the upstream mirror. Likely we would set that up in a transp. proxy configuration so that we can minimize the impact on CRF as much as possible. CRF will setup a "Disney" DHCP server to provide IPs. Disney is just the name they gave to machines that they received from that company. There will not be a local SMTP gateway, they do not want to deal with that. Power concerns --------------- We also picked their brains about the power situation. The different areas have a significant number of outlets available, but we were interested to know the circuit capacities and segmentation. They said that they do not know the different configurations, only that they've tripped some things in the past. When that happens, a ticket is filed with facilities, and it can take an hour or so to react. They indicated that they did not think that the records at facilities would have accurate information for the building circuits, but we can ask. TODO: talk to facilities about circuits, try to get them to provide us with a map. We also want them to be aware that we are going to be potentially running into this problem so that they can be prepared for this. setup times ----------- We talked about when we can do setup, they were interested in knowing how early we could do setup, so we can work out kinks. edrz said he is able to come up sometime next month to work on video encoding server setup. we talked about how debcamp is typically when things get seutp, but we are happy to come earlier. liason information ------------------- CRF provided us with an email address that goes to their team, and emergency phone numbers to D. which are not being sent to the list. We will put them in our orga contact list. CRF will be available during the normal work week and workday. Also, Richard said he would put together a contact list of orga to provide to CRF so they can get at us easily as well. Overnight considerations ------------------------- We talked some about after-hours, D. said that as long as there is a guard its ok, their preference is that we clear out of CS dept. before midnight, but it is ok for the team to be in and out 24hrs. They said that we need to be mindful of courtyard noise considerations. --
pgpRJakMmyWCR.pgp
Description: PGP signature
_______________________________________________ Debconf-team mailing list Debconf-team@lists.debconf.org http://lists.debconf.org/mailman/listinfo/debconf-team
