-
This site is deprecated and will be decommissioned shortly. For current information regarding HPC visit our new site: hpc.njit.edu
ResearcherHardwareStatus
For CSO reference (temporary location for this page)
- The purpose of this page is to provide CSO with a timely status of researchers' hardware requests, purchases, installations, and activations
- The hardware listed here is only for GITC 4320 (physical machines) and GITC 5302 (VMs).
- CSO will be advised by mail to coresys@njit.edu when there is a change to this table
| Line Num |
ID | Date/ By |
Faculty/ Dept |
Description | Follow-up | Comments |
|---|---|---|---|---|---|---|
| 1 | 1 | 11Mar16-16:16/ davidp |
Chase Wu/ CS |
Chase Wu, CS, 2 Dell servers, rabbit.arcs and tiger.arcs. Installation the week of 3/7/2016, GITC 4320 location G-08. Funded by DOE green computing grant. |
Researcher expects to receive further DOE funding in March 2016, with which he intends to purchase 2 more servers, most likely identical to the first two. The additional servers would be named cow.arcs and mouse.arcs. The researcher wants these two servers to also be located in 4320. The OK for this was given at the 2/25/16 meeting between MH, DP, PT. To be done ASAP. |
None |
| 2 | 2 | 11Mar16-16:23/ davidp |
Bin Chen/ Physics |
Researcher wants one server located in GITC 4320. The OK for this was given at the 2/25/16 meeting between MH, DP, PT. The server is to be used for very computationally intensive and high I/O tasks, with 48TB of disk. The price quote for the selected vendor, Microway, was sent to CSO Wed Mar 9 19:25. As of this date, no response. Requested server name : baozi.arcs.njit.edu Server is intended to be installed in GITC 4320 G-08 aka UCS Rack 5 Requires 2U Requires one 1GigE network connection Requires two 120V 7.5A NEMA 5-15 outlets |
When approved, the PR for the server will be entered by the Physics dept. |
None |
| 3 | 3 | 11Mar16-16:31/ davidp |
Bipin Rajendran/ ECE |
Met with ARCS for first time 2/4/2016. Will use Kong right away for research. Intends to purchase nodes in Kong summer 2016, based on usage experience. |
To date, no follow-up. Students are using Kong. |
None |
| 4 | 4 | 11Mar16-16:41/ davidp |
Joga Rao/ MIE |
Met with ARCS 3/3/2016. Wanted to know if faculty purchase an HPC cluster, would ARCS manage it Response was "no", explained Tartan model. From davidp Thu Mar 3 17:01:17 2016 To: raoi@njit.edu Cc: arcs@njit.edu Subject: New MIE faculty member, Jan 2017, HPC hardware Joga -- Per our discussion just now, from HPC node pricing we did in Dec 2015, fo ballpark pricing only : 2 CPUs per node, 6 cores per CPU, 10.7GB RAM per core, 500 GB disk per node : the cost for the node is $5100, or $425 per cores (processor). As discussed, we'll wait to hear what hardware the prospective faculty member is requesting. |
None | None |
| 5 | Related to ID 2 | 15Mar16-08:51/ davidp |
Bin Chen/ Physics |
14-Mar-2016 received OK from CSO for insatllation of MWYQ20348-01.pdf. Notified Bin Chen, who asked Cheryl James on 14-Mar-2016 to order this server. |
MT to make sure 1GigE network connection and two 120V 7.5A NEMA 5-15 outlets are in place |
None |
| 6 | Related to ID 3 | 16Mar16-17:28/ davidp |
B. Rajendran/ ECE |
+ Subject: Re: Catalogue + To: David Perel <davidp@njit.edu+ + Cc: arcs@njit.edu + + + Hi David, + I am working from home this week. Can you share the catalog from last year? Hi Bipin -- Let us know is you want to talk by phone. + I will have to provide a quote from the vendor in any case, so I will take + a look at what hardware you supported last year. Here's the most current catalogue we have (very oudated, from May 2014) : https://wiki.hpc.arcs.njit.edu/external/HPC_catalog-20140501-0858-AM-WIP.pdf + + I am actually thinking of getting some high-performance workstations + http://www.thinkmate.com/systems/workstations/hpx For locating where, and for what use ? + + Were any of these in your last catalog ? No. The catalogue contains only rack-mounted hardware. |
None at present | Done |
| 7 | Related to ID 1 | 17Mar16-13:50/ davidp |
Chase Wu/ CS |
Still waiting for the new grant from DOE. Submitted a revised budget on 3/17. | None | None |
| 8 | Related to ID 1 | 25Mar16-18:15/ davidp |
Bin Chen/ Physics |
Purchase requisition placed 3/16/2016 | Bin Chen will advise when there is a shipping ETA | None |
| 9 | 5 | 11Apr16-16:17/ davidp |
Researchers using SQL/ Various |
For informational purposes only XtremeData Inc. has offered to make their dbX software available to NJIT at no cost, for research only. Minimum virtualized hardware requirements for dbX, as specified by XtremeData : 16 CPU Cores 32 GB RAM 10gigE network 500GB disk for temporary data 200GB disk for persistent data Disk bandwidth of 1+ GB/s (either local or network-attached), multiple HDDs or SSDs |
Depending on response from researchers, XtremeData may do a presentation and demonstration NJIT. Office of Research is handling this. |
If NJIT goes ahead with this, funding source for hardware needs to be identified. |
| 10 | Related to ID 2 | 29Apr16-16:31/ davidp |
Bin Chen/ Physics |
Server baozi.arcs.njit.edu arrived 4/27/2016. |
From arcs+bncBC25NTO23UGRBWPERC4QKGQEUDDRRGY@njit.edu Thu Apr 28 11:55:20 2016 Subject: Re: MT : Bin Chen's SuperMicro server came in GITC-4320 -----> COR-9148 To: "Thomas, Matthew" <mrt@njit.edu> Cc: Academic and Research Computing Systems <arcs@njit.edu>, Core Systems General <coresys@njit.edu> X-Original-Sender: anthony.parker@njit.edu X-Spam-Checked-In-Group: arcs@njit.edu Matt T., Ok, when ready, you may proceed. Per your request, I checked and as discussed it would be a good idea to clean up the rack. In this case, to make some adjustments with the devices currently in the rack so that the new unit will fit better. |
None |
| 11 | 6 | 19Jun16-15:19/ davidp |
Gennady Gor/ ChemE |
Dr. Gennady Gor, ChemE faculty starting Sept 2016. Dedicated nodes in Kong.njit.edu : 30 nodes, 16 cores per node, 64GB per node Disk : Probably several TBs, split between local disk on Kong and AFS research space Several (3 or 4) computers (Linux) for his 321A Tiernan Hall lab. Dr. Gor would like the Kong nodes to be ready for use in early Sept 2016, but since that timeframe is almost certainly not realizable, is OK with mid-October 2016, assuming that the Kong public queues, and/or a temporary private queue will serve most of his computational needs until his desicated nodes are on-line. |
1. ARCS to provide node price quote to Dr. Gor. 2. Dr. Gor to provide storage space estimate to ARCS; ARCS to provide cost estimate for this space. 3. ARCS will work with CSO on provisioning of approximately 30 nodes for Kong. By removing most of or all of the 30 SB/X6220 and 11 SB/X6220 nodes currently in service in Kong, it should be possible to accommodate the 30 nodes without additional electric power or HVAC provisioning. |
None |
| 12 | Related to ID 6 | 07Nov16-13:36/ davidp |
Gennady Gor/ ChemE |
Dr. Gor met with ARCS on 10/27/16. At that meeting, the following were discussed :
|
There will be follow-up meetings with Dr. Gor to exactly specify the hardwaren he wishes to purchase in FYE17. Dr. Gor expects to start using his Kong nodes as early as February 2017. He was advised that it would take at least 3 months from the time the equipment is ordered to when it is ready to use. |
None |
| 13 | 7 | 07Nov16-14:15/ davidp |
Xiaoning Ding/ CS |
On 11/3/2016, Dr. Xiaoning Ding, CS. informed ARCS that he had puchased a 16-core Dell tower that he wished to be located in GITC 4320. Dr. Ding had not previously discussed this purchase with ARCS, |
None | |
| 14 | 8 | 07Nov16-14:32/ davidp |
Iulian Neamtiu/ CS |
On 11/4/2016, Dr. Iulian Neamtiu, CS. informed ARCS that he had puchased a SuperMicro tower that he wished to be located in GITC 4320. Dr. Neamtiu had discussed such a purchase with ARCS on 6/22/2016, but at that time the computer was not to be located in GITC 4320. |
None | |
| 15 | Related to ID 1 | 21Apr17-16:32/ davidp |
Chase Wu/ CS |
Third Dell server, in addition to tiger.arcs and rabbit.arcs, already in in 4320. The third server, and the existing 2 servers, will be located in a pod. |
Waiting for price quote from Chase Wu. |
None |
| 16 | 9 | 21Apr17-16:49/ davidp |
Kurt Rohloff/ CS |
R730 rack mount chassis w/ 19" rails Two CPUs: E5-2690 v4 2.6GHz,35M Cache,9.60GT/s QPI,Turbo,HT,14C/28t (135W) Max Mem 2400MHz Two GPUs: NVIDIA Tesla K80 Two RAM DIMMs: 64GB LRDIMM, 2400MT/s, Quad Rank, x4 Data Width Dual, Hot-plug, Redundant Power Supply (1+1), 1100W (required for GPUs |
Wating for confirmation from Kurt Rohloff |
None |
| 17 | 10 | 21Apr17-16:56/ davidp |
Usman Roshan/ CS |
3 GPU nodes in Kong Rack + switch + power distribution + cables |
Waiting for confirmation from Usman Roshan |
None |
| 18 | 11 | 21Apr17-17:02/ davidp |
Iuliam Neamtiu/ CS |
Server in the the process of being purchased. Configuration in process of being finalized with vendor. Will go in GITC 4320 pod along with Lousie, already in 4320. Skeleton description follows. Same rack * Same PDU * 16-port version of the switch * No compute nodes The second quote should be for: * One quad-server with network and power cables * Info on quantity discount break points (if some multiple of them are purchased) * No rack, PDU, or switch With this setup we would most likely install servers into the rack locally. |
Waiting for final price quote. |
None |
| 19 | Related to ID 6 | 21Apr17-17:19/ davidp |
Gennady Gor/ ChemE |
Goes in Kong. Attached for discussion is a revised quote from Microway. At $109K it is quite over budget. Of the total, $90K is for nodes. The remaining $19K (rack, switches, power distribution, etc) is quite good for future expansion except possibly needing another IB switch. I have several ideas to bring this to budget; in all of the following considerations I've left the $19K constant. To come in at $85K we would have to cut the number of nodes down from 30 to 22. The servers are 1U twin node, so 11x6K=$66K; total $85K. However, additional consideration of CPU and RAM (and in which I show dual-node servers, not nodes, because we cannot buy 1/2 server)... In the above 11x6K I switched to from the previously quoted dual 10-core to dual 8-core node CPUs following our discussion with your associate Noam Bernstein on Nov 10th, in which he suggested that most users are happy with 16 core nodes as they are cost effective for how they scale and may have faster clock speeds. In actuality the 8-cores are 2.10 v.s. 2.20 GHz for the 10-cores, an insignificant difference, but the servers are $1K more expensive. The quote includes an option for 10-core processors at 1K per twin server. You would have to cut to 9 servers (18 nodes) to make budget: 9x7K=$63k; total $82K. The quote also includes an option for 128GB RAM instead of 64GB originally discussed. We unfortunately did not discuss RAM with Noam (at least not in my poor recollection but detailed notes). Locally we have been keeping 128GB as a local standard for RAM and it does provide you with some future-proofing of your servers. At $700 per twin server this is fairly cheap future-proofing. With the 9x 10-core option above this comes to 9x700=$6K; total $88K (not sure if you have wiggle room on your budget, but we might be able to bet $3K break from vendor). With the 8-core option we'd have to go with one less server to make budget: 10x6.7K=67; total $84K. (Effectively 20 nodes but nicely inside budget.) |
Waiting for confirmation from Gennady Gor. |
None |
| 20 | 12 | 24Apr17-09:19/ davidp |
Xiaoning Ding/ CS |
Model number: Dell PowerEdge T630 Voltage and watts: 120V, 495W. Physical dimensions: Height 443.5 mm (17.5 inch) with feet. Can be reduced to 434.5 mm (17.1 inch) without feet Width:48.2cm (18.98") Depth:75.09cm (29.56") Processor type and count: Intel Xeon E52620 v4 2.1GHz, 2 processors, 16 cores GB RAM: 128 GB GB storage: 4.6TB What version of Ubuntu will be installed: Ubuntu 14.04 |
Awaiting pod |
Existing system phoenix in 4320 and sysmen is same pod |
| 21 | 13 | 24Apr17-09:47/ davidp |
Yuriy Polyakov/ CS |
Name: crc-esx-01.njit.edu Model: IBM x3550 M2 (7946-AC1) CPUs: 2 Xeon x5570 @ 2.93 Ghz; Each has 4 physical cores with two threads each. So 8 cores/16 logical processors in total. 36GB of RAM running in the DDR-800 mode RAID 5 consisting of 4x136 GB 15K RPM drives (408 GB in total) RAID 1 consisting of 2x67GB 15K RPM drives (67 GB in total) 2x1Gbit Ethernet ports |
Awaiting pod |
Share pod with K. Rohloff |
| 22 | 14 | 24Apr17-09:55/ davidp |
Joerg Kliewer/ ECE |
Joerg Kliewer, ECE faculty, requests locating 2 servers, one rackmount, one tower, to GITC 4320 (pod) while his lab in FMH undergoes renovation from June 2017 to December 2017. |
Awaiting approval for re-location. |
None |
