Not really a Web site problem, but I do have some questions about Internet connectivity.

Now that Arecibo has a T3/Internet2 connection, does the SETI@home project use that, or do they still rely on tapes by mail? The S@h article in Wikipedia, as well as some things I've come across on the site here (I think, anyway--I can't find them) still seem to indicate the Project still gets the tapes mailed to them.

I would think that Arecibo, the Internet2 administrators and the University of Florida (where the I2 gigapop is) could spare a couple gigs per day--that's only 79 seconds of data transfer.

Also, has Berkeley increased its commodity internet connection[s]' bandwidth? I remember in '02 or so that S@h was using 30Mbps, which was about a third of the campus bandwidth. 100Mbps connections were gargantuan back then, but nowadays, a major research institution like Berkeley should probably have their own OC-12, if not more. (My university system, the University of Alaska, has an OC-12 circuit to Seattle, among other connections, so I would think Berkeley would have far more than that.) And does S@h still have a separate Cogent link, or are they back on the (hopefully newly-increased) university's bandwidth?

Thanks in advance,
Chris

---

MY GUESS would be that they are leery of the transmission errors that could happen causing the data to be corrupted. I am guessing the Scientists could work around this but....

---

Those are VERY good questions! I only hope someone would be willing to answer them either here, or perpahs in an upcoming Q&A to be posted somewhere.

---

Those are VERY good questions! I only hope someone would be willing to answer them either here, or perpahs in an upcoming Q&A to be posted somewhere.

Well, I've come across some of the answers myself, but it required a little sleuthing on my part.

1. Berkeley HAS increased their commodity (commercial) Internet connection. They're connected to CENIC, the Corporation for Education Network Initiatives in California, at 310mbps (two 155mbps pipes). CENIC then connects to Level3 and Qwest. (CENIC's website (cenic.org) doesn't provide any details about its connection to the commodity Internet, but it's gotta be at least a couple of OC12s or so, since they serve pretty much all of California's higher education.) Berkeley's policy since the mid 1990s has been to allocate $250,000 per year for network expenditures. Their commodity Internet connection bandwidth is whatever $250,000 will pay for, and as the price of bandwidth drops, their speed goes up.

2. Berkeley is also connected to Internet2, the network operated by a consortium of higher education institutions, through CalREN. Berkeley's pipe to CalREN is 622mbps (a full OC12). Internet2 traffic is only between participating higher education institutions, and it was designed to share and transfer data for research purposes between those institutions. Since most SETI@home users aren't at other universities (although a sizable percentage are, I'd guess), this really isn't useful for them. I did find it useful last night when I was running speed tests from a computer at my school: I was able to download files from Berkeley to the University of Alaska at 77 mbps. (My connection to the network was GigE. 77 mbps is pretty close to the theoretical maximum accounting for TCP overhead on a 100 mbps link, so I'm thinking that either there's a 100baseT switch or router somewhere between the lab I was in and the connection to Internet2, or the server I was downloading from didn't have a GigE NIC in it. I suspect the former, as I tried downloading two files from two different servers at the same time, and the data transfer was pretty close, though I think I squeezed over 90 mbps out of the connection. Wow. I know the connection to my machine was functional GigE, because I transfered a large file to a neighboring machine at 44 MBps, or 352 mbps. Even more wow. I think that was hard drive data transfer limitations more than anything.)

3. SETI@home still uses the Cogent link, even though Berkeley's general connection has three times as much bandwidth. I suspect a couple reasons for this. First, SETI@home reports that the Cogent link (at 100 mbps) occasionally gets saturated (for example, when everyone was transitioning from Classic to BOINC). If they were using the general Berkeley connection, they could still use upwards of 1/3 of the entire campus's bandwidth. Second, with larger, more multimedia-rich files being transferred from Berkeley's web servers (which Berkeley reports is the primary use of their commodity connection), higher data-rate (HD, even?) streaming media, and of course more activity from the residence halls (think 7 GB DVDs being transferred with P2P software...), I'm sure Berkeley's network usage has increased proportional to SETI@home's. So a 310 mbps connection could easily become saturated. And third, UCB will charge SETI@home to use its commodity Internet connection. SETI@home might be getting a better deal with Cogent than they would get through the university. Some have rumored that SETI@home pays about $1000 per month--not bad for 100 mbps--especially when UCB was charging SETI@home $6000 per month to guarantee them 20 mbps.

Why is the bandwidth so cheap? Perhaps it's a low-cost, low-priority agreement--some have speculated this because their response time in repairs is so slow. Perhaps they're offering it at a special educational/non-profit price, doubly cheap because the connection is located at the already-existing Palo Alto Internet Exchange (PAIX). Or maybe they think the PR they'll get as supporting and being the backbone of the "popular SETI@Home project" will help them as a corporation. I know they keep trying to convince themselves that they're a Tier 1 ISP so they can peer with Level3 instead of pay for data transit, so perhaps this is a way of boosting their data transfer so the big boys in ISP land will respect them.

Some random interesting notes:

-SETI@home's client downloads are handled by a server at MIT
-SETI@home's web site is connected via Berkeley's network
-Only SETI@home workunits and results go over the Cogent link
-Traceroutes from the University of Alaska to SETI@home's web server show the routes going over Internet2. Traceroutes going to the data server (setiboincdata.ssl.berkeley.edu) go over the public network, over two Cogent routers, and then terminate at the server. UCB allows on-campus S@H clients to connect directly without going over the public network. I wonder how hard it would be to allow Internet2 traffic to connect via the campus network. This would be a good thing, I think, since a lot of SETI data crunching is done by university folks. Using Internet2 for this traffic would: a) relieve traffic from the Cogent link, b) allow ultra-high-speed connections for these clients, c) save on other universities' bandwidth connections, since they won't have to send data out on the commodity network, and d) just be cool!)

You can see a map of Berkeley's network here:
http://www.net.berkeley.edu/netinfo/newmaps/ucb-border.html

UCB interface bandwidth usage:
Router between UCB and CENIC (all commodity Internet traffic): http://fragment1.berkeley.edu/~cricket/inr-666-interfaces.html
Router between commodity Internet and SSL lab (where S@H lives): http://fragment1.berkeley.edu/~cricket/inr-202-interfaces.html
Router between UCB and CalREN (Internet2): http://fragment1.berkeley.edu/~cricket/inr-000-interfaces.html
And the biggie, for our purposes: Router between UCB and PAIX, where S@H's Cogent connection terminates: http://fragment1.berkeley.edu/~cricket/inr-668-interfaces.html

Other sources:
http://www.net.berkeley.edu/netinfo/newmaps/ (All campus network maps)
http://www.cenic.org/ (Operator of UCB's commodity and Internet2-via-CalREN connections)
http://www.pacificwave.net/private/docs/technology/PacificWaveswitches.pdf (How UCB and CENIC connect to the Internet)
http://setiathome.berkeley.edu/forum_thread.php?id=14488 (Discussion of Cogent cost)
http://www.planetary.org/html/UPDATES/seti/SETI@home/Update_052002.htm (Background on the Cogent connection)
http://nac.berkeley.edu/bandwidth/ (Network Advisory Committee on UCB's bandwidth usage)
http://qos.internet2.edu/wg/cbm/cbm-matrix.html (Bandwidth controls at various universities, including charging departments for usage)
http://setiathome.berkeley.edu/sah_photos.php?album=cogent_05_30_2002 (The Big Day: when the Cogent connection was activated)

---

Thanks for sharing. It was a good read.

---

I have a correction: S@H's client downloads are handled by several mirrors. I count four: MIT, the University of Wisconsin-Milwaukee, the University of Glasgow, and the Max Planck Society in Germany.

I'm still wondering, though, about the Arecibo tapes being mailed vs. uploaded. Anyone know if and why?

---

CENIC's website (cenic.org) doesn't provide any details about its connection to the commodity Internet, but it's gotta be at least a couple of OC12s or so, since they serve pretty much all of California's higher education.

Just discovered this: according to http://nitro.ucsc.edu/ (the Internet2/Abilene-connected speedtest server at UC Santa Cruz), "[t]he CENIC HPR network is jumbo clean and attaches to Abilene via 10GigE."

CENIC runs three separate networks (the HPR, or High Performance Research, network is the one mentioned above), so perhaps not all of their Abilene-bound traffic runs over that 10GE connection.

Also, I got a cool shot with my camera phone: the University of Alaska's connection to the Pacific Northwest GigaPop. In a small window next to the Cray X1 supercomputer in the basement of the Butrovich building at UAF is a rack with a few routers on it, one of which is labeled "PNW" and has a plugin labeled "OC-12". Kinda cool to see the itty-bitty fiber optic cable that all of the University of Alaska's Abilene-bound data (and there's a lot of it, due to the supercomputers that the Arctic Supercomputing Research Center has) goes over. There's another plug labeled "OC-3", and I wonder if that's UA's commodity Internet connection, but I don't remember specifically if it was labeled such (the UA uses Alaska Fiber Star for their commodity connection).

Still looking for an answer from a S@H staffer why they're still using tapes and mail if Arecibo has a high-speed Internet connection now...

---

I figured (long after I posted the original message in this thread) that I should source where I got that Arecibo now has a T3. See this article at Space.com.

According to this thread (which backlinks to this thread!), S@H's data rate for Arecibo's observations is 32Mbps, which would be a good chunk of the 155Mbps bandwidth that their T3 offers. That may be why we don't (aren't allowed to?) use it and are still using tapes.

---

I must have been half asleep when I wrote the last post, because not only did I not catch the error that the Space.com article contained--I passed it along as well.

The Space.com article claims that Arecibo has a T3 which provides 155mbps. That must be some special T3, because a standard T3 (in North America, anyway) can transfer about 45mbps. Perhaps they were thinking about an OC-3, which carries 155mbps. It's hard to tell, but my guess is that they actually have an OC-3 and that the source Space.com quoted, Arun Venkataraman, Director of Computing at Arecibo, was mistaken. (It's hard to believe that the Director of Computing at Arecibo would be mistaken about this, but perhaps he's just an executive figurehead and really doesn't know all that much about IT...)

---