Virtual School of Computational Science and Engineering offering new courses

May 13, 2010
NCSA logo

NCSA

The VSCSE, the Virtual School of Computational Science and Engineering, is once again offering courses. For this time, they added quite a lot of new sites, where you can attend the courses – 21 sites in all over the US are now available as classrooms. The VSCSE is provided and funded by the Great Lakes Consortium for Petascale Computation (GLCPC), the National Science Foundation (NSF), the State of Illinois, the Committee on Institutional Cooperation (CIC), and Internet2 Commons.

Their press-release:

Want to learn how to use graphics processors for scientific computing? Scale your parallel code to tens of thousands of CPU cores? Deal with ginormous datasets? The Virtual School of Computational Science and Engineering offers these courses and more during its summer program for 2010!

Since 2008, nearly 250 students and researchers have participated in the annual Summer School offered by the Virtual School. During Summer School, students learn new techniques for applying high-performance computing systems to their work. Due to overwhelming demand for courses in previous Summer Schools, we have added 15 sites (for a total of 21 sites) to the 2010 program in order to accommodate additional students. For each course, students attend on-site in one of 10 state-of-the-art, distributed high-definition (HD) classrooms, located at academic and research institutions across the country. These HD classrooms are equipped with live, high-definition videoconferencing technology that provides a high-quality learning experience.

Students attend technical sessions presented by leading researchers in computational science and engineering and use cutting edge, high-performance computing systems provided by TeraGrid resource providers. Course participants apply the techniques learned in hands-on lab sessions, assisted by skilled teaching assistants who work one-on-one and in small groups to answer questions and solve problems posed during the sessions. This summer’s courses are:

The cost for each course is only $100. To participate, prospective students must first be enrolled in the Virtual School. Enrollment is free and can be completed at https://hub.vscse.org/. After enrolling, students select their courses and indicate which of the distributed HD classrooms they would like to attend.

Snacks and an evening reception will be provided; participants are responsible for travel and lodging costs (low-cost dorm accommodations will be provided where possible). Because of the large geographic diversity of participating sites, it is likely that little travel will be required.

For no additional cost, on-site participants can take online short courses on MPI, OpenMP, and CUDA that are designed to help them meet course prerequisites.

For more information on the 2010 courses, including the sites participating in each course and details on enrollment, go to: www.vscse.org/summerschool/2010


Planning IT-installations – Defining the scope of the project (#2)

May 30, 2008

This is part #2 in the series of postings “Planning IT Installations”.
Part #1 was the Introduction to the series of this postings: Planning IT-installations – Abstract (#1)

Introduction

OK, so you were assigned to this cool new project.

So what now?

First you have to check what exactly your scope of work is. In this series I’ll use an example project, consisting of a Fujitsu-Siemens Primepower 650 cluster in a high-available environment. The purpose of this cluster is to be a database-cluster, serving your company’s most precious secrets.

It’s your task, as the responsible planning-engineer, to make the best possible database-solution of the sold hardware, sail around the problems your sales-department created, deal with all those customer-requests, do all the work which is required so that the installation-staff can build this mother and be responsible for all IT- and engineering related problems which might occur.

Your Mission

You are the responsible planning-engineer for the project “A new high-available database for Antarctica’s ice-core samples“.

You boss gave you the following tasks:

  • A dual-node, Fujitsu-Siemens, Primepower 650 cluster with two EMC CX300 Fibrechannel Storage cabinets was sold to the customer as the cutting-edge database-solution ever. The solution shall be high-available.
  • Go and talk to the customer and check, if the sold solution will really work as sold.
  • Plan and implement the cluster with the delivered components
  • Make it work
  • Make the customer to sign the acceptance-test handbook
  • …profit!

Meet the customer

You need to do two things:

  1. A site-survey to check, if the cluster you’re about to plan, really fits into your customer’s environment
  2. A customer-workshop to sort all all requirements which have to be met

In what order?!

OK, you got lot’s of things to do. What to do first?

  1. Evaluate the bill of materials to see how your hardware really does look like
  2. Think about, how you assemble the racks with all the material you have
  3. Make a questionary for the customer, if the racks, you planned, fit into their data-center.

Evaluate the bill of materials #1 – what did you get from sales?

You really want to know what was sold, don’t you? So the first thing is to get a bill of materials from your sales-department to check out what was sold. Be sure to get the real list. With all details. Up to the last cable.

Your list – or better, the list in our example – looks like that:

Partnumber D:GPS65-GM21-K01 – PW650 2 node, 4×1,89GHz, 8GB, Sol9:

Position Partnumber Description Amount
10 D:FCKAB-MM-C20L FC Kabel MMF 20m, Connector LC-LC 2
20 D:GP7KB-2015 UTP CAT 5 Enhanced, 15m 14
30 D:GP7KB-2015 UTP CAT 5 Enhanced, 15m 1
40 D:GP7KB-1420 UTP CAT 5 IC-Kabel 2xRJ45 8pin(S) 20m 1
50 D:GPRAC-BG52 PRIMECENTER Rack 38HE, 1000mm tief 1
60 D:KB258-C250 Netzltg. Kaltgeraet 2,5m (Stift-Buchse) 2
70 D:GP7KB-2005 UTP CAT 5 Enhanced, 5m 2
80 D:GPRAC-KB04RE16 RCA4-Kabel, Port1-16, RJ45/SubD9, 5m 2
90 D:GP7KB-1204 RCI Kabel RJ45 6pin(S)-RJ45 6pin(S) 4m 1
100 D:GPRAC-ZB30 Blindblende fuer 19″ Rack, 1HE 7
110 D:GPRAC-ZB32 Blindblende fuer 19″ Rack, 3HE 1
120 D:GPRAC-SV011 SDL CEE 1x16A > 10xIEC320-10A 2
130 D:GPRAC-SV121 SDL CEE 1x32A > 8xIEC320-10A 2
140 D:GPRAC-ZB5214 Rear-Einbaukit fuer Hubs/Switche 2
150 D:GPRAC-ZB5012 Erdungskit f. Primecenter Rack2 1
160 D:GPRAC-ZB22E Halteboden fuer RCA4 und RPS-Box 1
170 D:GPRAC-RCA4P RCA4plus mit 8 seriellen Ports 1
180 D:GP650-GM15 PW650(EF) GHz-CPU, 8CPU/64GB/8PCI 1
190 D:GP650-ZE84 CPU Submodul 1,89GHz SLC 3MB f.PW650(DE) 4
200 D:GP70M-SP65 Speichererweiterung 8 GB f PW650-1500 1
210 U24529-C423 Solaris 9 RTU PRIMEPOWER 1
220 D:GP70M-SP65 Speichererweiterung 8 GB f PW650-1500 1
230 D:GP71F-FP90 Hard Disk 73 GB 3,5″ 10k rpm 2
240 D:GP70F-CL21 Quad Fast/Gigabit Ethernet Contr., RJ45 2
250 D:GP70F-CL19 Gigabit Ethernet NIC, Fibre Optic LC 2
260 D:GP650-SV24 DUAL POWER FEED OPTION f. PW 650(CDEF) 1
270 U11420-C665 PCLPAS-P V42 (SOLARIS PW) NR -CL8CPU 1
280 U11420-C635 PCLSIS V42 (SOLARIS PW) NR -CL8CPU 1
290 D:GP70M-BG61 Ext PCI box 12 slot f. PW 650/850 f.PCR2 1
300 D:GP650-GM15 PW650(EF) GHz-CPU, 8CPU/64GB/8PCI 1
310 D:GP650-ZE84 CPU Submodul 1,89GHz SLC 3MB f.PW650(DE) 4
320 D:GP70M-SP65 Speichererweiterung 8 GB f PW650-1500 1
330 U24529-C423 Solaris 9 RTU PRIMEPOWER 1
340 D:GP70M-SP65 Speichererweiterung 8 GB f PW650-1500 1
350 D:GP71F-FP90 Hard Disk 73 GB 3,5″ 10k rpm 2
360 D:GP70F-CL21 Quad Fast/Gigabit Ethernet Contr., RJ45 2
370 D:GP70F-CL19 Gigabit Ethernet NIC, Fibre Optic LC 2
380 D:GP650-SV24 DUAL POWER FEED OPTION f. PW 650(CDEF) 1
390 U11420-C665 PCLPAS-P V42 (SOLARIS PW) NR -CL8CPU 1
400 U11420-C635 PCLSIS V42 (SOLARIS PW) NR -CL8CPU 1
410 D:GP70M-BG61 Ext PCI box 12 slot f. PW 650/850 f.PCR2 1

Partnumber D:GPSRC-CX3B1-K01 – 2x CX300_S per 15x 73GB 15k

10 D:FCKAB-MM-C20L FC Kabel MMF 20m, Connector LC-LC 4
20 D:FCKAB-MM-C05L FC Kabel MMF 5m, Connector LC-LC 4
30 D:GPRAC-ZB5211 Tragewinkel 1HE für 19″ Einbauten PCR2 1
50 D:FCX32R-73FH FC CX300_S Basism.15x73GB/15k SI COM 1
60 D:FCRAIL-2 Montage-Satz FibreCAT f. PrimeCenterRack 1
80 D:FCNAV300-WG FibreCat SwLic NaviMgmt f CX300 WG 1
90 D:FCNAV-WGKIT FibreCat SwKit NaviMgmt WG 1
100 D:FCAGSUN-KIT Navisph.Agent Sw.Kit Solaris (per site) 1
110 D:GP70F-CF34 Fibre Channel Cont. Dual Port 2Gbit/s 4
120 U24529-C530 Multipath V1.1B NR (Sparc) 2
130 D:GPRAC-ZB5211 Tragewinkel 1HE für 19″ Einbauten PCR2 1
150 D:FCX32R-73FH FC CX300_S Basism.15x73GB/15k SI COM 1
160 D:FCRAIL-2 Montage-Satz FibreCAT f. PrimeCenterRack 1
180 D:FCNAV300-WG FibreCat SwLic NaviMgmt f CX300 WG 1
190 D:FCNAV-WGKIT FibreCat SwKit NaviMgmt WG 1
200 D:FCAGSUN-KIT Navisph.Agent Sw.Kit Solaris (per site) 1

Evaluation the bill of material part #2 – So what do we have?

The first list, about the machines, tell us the following facts:

  • We have two 38-RU racks (Pos. 40 and 390)
  • We have two Primepower 650 machines with 8 CPUs and 64 GB of RAM each (Pos. 270 and 600)
  • Those PW650 machine were ordered with the dual-power feed option (Pos. 340 and 670)
  • We have two DN30 SCSI-cabinets (Pos. 200 and 530)
  • We have one dumb Ethernet-switche for networking (Pos. 190)
  • We have, for each rack, an IEC320 8-port power-outlet (Pos. 130) and a 3-port Schuko power-outlet per rack (Pos. 120)
  • Lot’s of misc. stuff

The second list, about the EMC Clarions, tells us:

  • We have two CX300 base-cabinets with 15x73GB (Pos. 50 and 150)
  • Both cabinets have reduntant FibreChannel ports (Pos. 110)
  • Everyhing, like cables and licenses, is included

Bringing it all together

Now that you checked what sales sold and groked, what’s really going to be delivered, you have to make up a rough proposal which you can present to the customer. You have to make up an idea how the racks exactly have too look and what components are included.

Now you have to figure out what exactly you have to do.

What Do you have?

The bill of materials tells you a lot of things. It tells you that you have…

  • Two racks where your two cluster-node will be builtin
  • All the necessary components to power up the system when thinking of electricity (Pos. 130 and 120)

All those facts tell you that you’re pretty much independant of all the problems which might occur.

What do you need?

In this certain case, you, as the planner, are pretty much independant now. But yet you need to persuade the customer to include your two racks into their datacenter.

So, when talking about those requirements, you need to check out what exact things the customer has to provide to meet your expectations to make the cluster running.

That include things like:

  • How many power feed to provide?
  • What’s the footprint of my racks? How heavy are they? What’s the maximum heat-dissipation – can the customer’s air-conditioning cope with our system?

The scope

Now that you figured out what was delivered and got a basic idea of what you have to do to implement this solution on your customers’s site, you have to define the scope of your project.

  • The scope of your project is your assigned mission.
  • The scope includes to evaluate, if the given hardware is enough to be set up in your customer’s datacenters.
  • You need to check if the hardware is good enough to be set up in your customer’s datacenter; (Example: You’re about to set up two cluster nodes in two different server-rooms. Those rooms are 100 km apart. Implicite action: Go and persuade project-management ro buy FC-switches and LAN-Switsches witj LH-modules.)
  • You’re responsible to get the cluster working.

This is the end of this posting. All remaining questions will be answered in the next postings.

The next posting is due by this Sunday, July 1st.

If you have any questions, do not hesitate to get in touch with me.


NSA: Ford Meade running out of juice

August 6, 2006

NSA logoVia Slashdot:

The Baltimore Sun reports that the NSA‘s baracks in Fort Meade are apparently running out of juice. It’s reported that the power-shortage will delay “the installation of two new, multimillion-dollar supercomputers“. It was not mentioned what vendor and brand of supercomputers are meant, but that’s not suprising – lot’s of supercomputer-systems are classified.

They also seem to have serious problems with the air-conditioning in their server-rooms; it’s not mentioned if this is also due to the lack of power or due to under-dimensioning.

The NSA is now considering shutting down legacy systems and installing additional generators.

Baltimore’s Constellation Energy reminds it’s customers in the meantime to use energy wisely. They advice “Set your air conditioner thermostat higher than usual, if health permits“.

Interesting sidenote for me as a german: “Fort Meade is not the only NSA outpost facing limitations on its ability to upgrade electrical infrastructure. Listening posts around the world, such as Menwith Hill in Britain and Bad Aibling in Germany, are ailing.” – What the hell do they have in Bad Aibling? I mean everyone knows that Echelon Field Station 81 is there, but we all thought it already out-of-service.

So: What’s in Bad Aibling?

Tags:


Follow

Get every new post delivered to your Inbox.

Join 120 other followers