October 2008 - Topg-ssf - mailman3.cels.anl.gov

Tomorrow's meeting
by Craig Stacey 23 Oct '08

23 Oct '08

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Tomorrow's meeting we'll finalize the power layout and take a deeper look at the WBS for the move. == Craig -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.8 (Darwin) iEYEARECAAYFAkkAs3cACgkQbVNGA3HOzQv8FgCePJ2+dz9jMDjiRhRxJVp9hMdL R2YAn1OQaiAUqrWYOY4WNFI7c0tYs+nt =eYXH -----END PGP SIGNATURE-----

1 0

Power drawings
by Craig Stacey 20 Oct '08

20 Oct '08

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 I've put up two new PDFs on the wiki. http://wiki.mcs.anl.gov/SSF/images/8/8b/SSF-Day-One-Power-Layout.pdf This is the first pass at a power layout for Day One occupation of the SSF. Missing are specs for the BG/P and SiCortex. I've used the 4 x 20A 208 layout for LCF power, but Bill's looking into going with our standard of 2 x 30A 3 Phase. http://wiki.mcs.anl.gov/SSF/images/f/fd/SSF-Day-One-Power-Layout-With-Machi… This is the above with machines overlaid. The layout will change, since LCRC and the NSF clusters will likely be bigger than the footprint given. To that end, I've put out an extra row of 60A circuits in that area (with the expectation that we're not likely to see a Cell cluster land in the next year either). Comments, input, feedback welcome and encouraged. == Craig -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.8 (Darwin) iEYEARECAAYFAkj3uPAACgkQbVNGA3HOzQsPEACfel4Ysbepdd/J/vg/+oR6W/rv 1TAAn2wzKkMTm18/Yvfj5R6ShDX+avkZ =CGJT -----END PGP SIGNATURE-----

5 8

Topg-SSF
by Craig Stacey 10 Oct '08

10 Oct '08

The following meeting has been modified: Subject: Topg-SSF Organizer: "Craig Stacey" <stace(a)zimbra.mcs.anl.gov> Resources: 221-A261 (Site: ANL, Building: 221, Floor: 2, Room: A261) Time: 1:00:00 PM - 2:00:00 PM GMT -06:00 US/Canada Central Recurrence : Every 2 weeks on Friday. End by Dec 19, 2008. Effective Oct 24, 2008. Invitees: topg-ssf(a)lists.mcs.anl.gov *~*~*~*~*~*~*~*~*~*

1 0

Topg-SSF
by Craig Stacey 10 Oct '08

10 Oct '08

The following meeting has been modified: Subject: Topg-SSF Organizer: "Craig Stacey" <stace(a)zimbra.mcs.anl.gov> Resources: 221-A261 (Site: ANL, Building: 221, Floor: 2, Room: A261) Time: 1:00:00 PM - 2:00:00 PM GMT -06:00 US/Canada Central Recurrence : Every Friday. No end date. Effective Oct 24, 2008. Invitees: topg-ssf(a)lists.mcs.anl.gov *~*~*~*~*~*~*~*~*~*

1 0

Today's meeting cancelled
by Craig Stacey 10 Oct '08

10 Oct '08

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 In case the calendar notification didn't go out, today's meeting is cancelled -- sorry for the short notice. I've got some things I need to work on this afternoon, and there don't appear to be pressing issues (plus a number of people are out). Next meeting is at 10/24. Thanks! == Craig -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.8 (Darwin) iEYEARECAAYFAkjvissACgkQbVNGA3HOzQvLHwCgljn8994TTNOA1J6YsRoZBJFv lEQAn3ckrr4Ahy0Awi0uleoYZtLUf1QK =eFXQ -----END PGP SIGNATURE-----

1 0

Re: [topg-ssf] Meeting notes
by Corby Schmitz 03 Oct '08

03 Oct '08

This is a difficult situation. It would be great to have a single solution, but that's not going to happen. There are very few 60a PDUs available which meet the needs of the average rack. 2x 3-phase 30a is sufficient to run a rack of systems for the average instance. We know that a single circuit will not work of system's resources as we need a one UPS and one COM connection. The nice part is that even with two 30a COM circuits, a pair (fused management interface) of PDUs provides more than enough power. We are talking about ~100a @ 208vac. The same solution with s single 60a does provide the same power, but minimizes our capabilities for getting PDUs to front-end the system. There is an alternative to trying to pick the right receptacle to be in place. We could have power fee troughs under the floor. This would allow us to drop in multiple conductors of various sizes and get them where they need to be. We have such things in the BMR, but they have been pretty much abandoned as they only cover a small portion of the underfloor space. Using a more modular approach would allow us to have one distribution system (cable-centric), with custom accessibility where it needs to be customized. corby ----- Original Message ----- From: "Craig Stacey" <stace(a)mcs.anl.gov> To: "Schmitz Corby" <schmitz(a)anl.gov> Cc: topg-ssf(a)lists.mcs.anl.gov Sent: Monday, September 29, 2008 2:03:20 PM GMT -06:00 US/Canada Central Subject: Re: [topg-ssf] Meeting notes -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 I'm rethinking our assumptions from earlier this year, and we need to settle on this soon. We need to make sure we're choosing the *most useful* configuration for power in the SSF. Nothing we choose will be universally applicable, but as long as we come up with something that's going to be the trend, we can buy the right PDUs to make it useful to whatever going into the cabinets. What concerns me is that the trend in the computing clusters I see coming in is definitely skewing to the 60A range and not in the 30A range -- this is precisely what prompted our discussion in the meeting on Friday about converting to 60A. Should we be looking at going with a 60A 3 Phase run to every cabinet, alternating UPS and non-UPS for each receptacle? Should we further adopt a hybrid strategy, wherein we deploy 4 x single phase 208 to ALCF's area, 2 x 30A 3 phase to a portion of the room, and 60A 3 phase to another portion? It's starting to feel like the latter is the right choice. Incidentally, the proposed (by Teng) SSF PDU schematic and data sheet is now the wiki: http://wiki.mcs.anl.gov/SSF/index.php/Image:PDM_Mk_II.pdf http://wiki.mcs.anl.gov/SSF/index.php/Image:PDM-MK-II-Schematic.pdf == Craig On Sep 29, 2008, at 1:40 PM, Schmitz Corby wrote: > The wiring would need to change. #10 stranded will be too small and, > we should check with Paul, but I believe that using parallel > conductors is not allowed with smaller than a #2. We would have to > consult the tables, but this would be a #4 I believe. There is > another consideration here, the standard conduit size for a 5 wire 3- > phase circuit is 3/4" (squeezing it into a 1/2" is not advisable). > We would need at least a 1-1/2" conduit for 5x #4 conductors to > support the 60a circuits. > I am not sure what we are planning as far as electrical > distribution, but it is something that we should hash out as it > drives the maximum size of circuits going out to our cabinets. > > -corby -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.8 (Darwin) iEYEARECAAYFAkjhJngACgkQbVNGA3HOzQssQACfXPNyKErlWQnIMY4PeV+QVRiR QJwAn3ntnBZ2NqX0X0PSfF3RCofr5Lix =icSa -----END PGP SIGNATURE-----

7 10