Difference between revisions of "Hardware servers"

From Cncz
Jump to: navigation, search
m (Disk[-][ ]servers)
([Reken-][Compute ]servers/cluster)
Line 39: Line 39:
 
[/en]
 
[/en]
  
=== [Reken-][Compute ]servers/cluster ===
+
=== [Reken-][Compute ]servers/clusters ===
  
 
[nl]
 
[nl]
Begin 2006 is door een aantal afdelingen een stapel Sun Fire X4100 dual processor dual core machines aangeschaft om te voorzien in de behoefte aan Linux rekencapaciteit. In 2007 is [[software_cluster_|Cluster software]] geïnstalleerd. Afdelingen hebben met enige regelmaat nieuwe machines aangeschaft, die toegevoegd zijn aan het cluster. Per augustus 2014 telde het cluster 1324 cores en 7914 GB RAM. Diverse clusternodes hebben een GPU met CUDA cores. Het coma-cluster van Sterrenkunde met (per augustus 2014) 320 cores, GPU's en Infiniband interconnectie staat hier los van.
+
C&CZ beheert een aantal rekenclusters, waarbij afdelingen hun eigen compute nodes aangeschaft hebben. In enkele gevallen hebben afdelingen een apart cluster gemaakt door ook hun eigen headnode (master) aan te schaffen. Het grootste <tt>cn</tt>-cluster heeft een door C&CZ aangeschafte headnode.
De [http://cricket.science.ru.nl/grapher.cgi?target=%2Fclusternodes CPU-load van de clusternodes] wordt automatisch bijgehouden.  
+
Voor afdelingen van FNWI is de housing en het beheer van clusternodes gratis. Het <tt>cn</tt>-cluster is heterogeen, de oudste nodes hebben nog [http://fedoraproject.org/nl/ Fedora Linux] met [http://en.wikipedia.org/wiki/Oracle_Grid_Engine Sun GridEngine] als [[Software_cluster|clustersoftware]], nieuwere nodes hebben [http://www.ubuntu.com/ Ubuntu Linux] met [http://www.adaptivecomputing.com/products/open-source/torque/ Torque] / [http://www.adaptivecomputing.com/products/open-source/maui/ MAUI]. Het plan is om over te stappen op [https://computing.llnl.gov/linux/slurm/ SLURM].
 +
Per augustus 2014 telde het cn-cluster 1324 cores en 7914 GB RAM. Diverse clusternodes hebben een GPU met CUDA cores. De [http://cricket.science.ru.nl/grapher.cgi?target=%2Fclusternodes CPU-load van de cn-clusternodes] wordt automatisch bijgehouden.
 +
 
 +
Het coma-cluster van Sterrenkunde met (per augustus 2014) 320 cores, 240 TB storage, GPU's en snelle [http://nl.wikipedia.org/wiki/InfiniBand Infiniband] interconnectie is een apart cluster. Ook voor de RU-afdelingen [http://www.ru.nl/cls/ Centre for Language Studies] en [http://www.ru.nl/donders/ Donders Centre for Cognition] worden aparte rekenclusters beheerd.
 
[/nl]
 
[/nl]
 
[en]
 
[en]
Beginning 2006, a number of departments bought a bunch of Sun Fire X4100 dual processor dual core machines in order to have more Linux computing power. In 2007 [[software_cluster_|Cluster software]] was installed. Departments have regularly bought new machines that were added to the cluster. As of August 2014, the cluster totalled 1324 cores and 7914 GB RAM. Several clusternodes have a GPU with CUDA cores. The Astrophysics coma cluster with 320 cores, 120 TB discs (as of August 2014), GPU's and Infiniband interconnect is a separate cluster.
+
C&CZ manages a number of compute clusters, with compute nodes that departments have purchased. In a few cases, departments have created a separate cluster by also purchasing their own head node (master). The largest <tt>cn</tt>-cluster has a head node purchased by C&CZ.
The [http://cricket.science.ru.nl/grapher.cgi?target=%2Fclusternodes CPU-load of the clusternodes] is being monitored.
+
For departments of FNWI the housing and management of clusternodes is free of charge. The <tt>cn</tt>-cluster is heterogeneous, the oldest nodes still have [http://fedoraproject.org/ Fedora Linux] with [http://en.wikipedia.org/wiki/Oracle_Grid_Engine Sun GridEngine] for [[Software_cluster|clustersoftware]], newer nodes have [http://www.ubuntu.com/ Ubuntu Linux] with [http://www.adaptivecomputing.com/products/open-source/torque/ Torque] / [http://www.adaptivecomputing.com/products/open-source/maui/ MAUI]. The plan is to move to [https://computing.llnl.gov/linux/slurm/ SLURM]. As of August 2014, the cluster totalled 1324 cores and 7914 GB RAM. Several clusternodes have a GPU with CUDA cores. The [http://cricket.science.ru.nl/grapher.cgi?target=%2Fclusternodes CPU-load of the clusternodes] is being monitored.
 +
 
 +
The Astrophysics coma cluster with 320 cores, 240 TB discs (as of August 2014), GPU's and Infiniband interconnect is a separate cluster. Separate clusters are also managed for the RU departments [http://www.ru.nl/cls/ Centre for Language Studies] en [http://www.ru.nl/donders/ Donders Centre for Cognition].
 
[/en]
 
[/en]
  

Revision as of 16:33, 18 September 2014

Servers and Workstations

C&CZ manages ca. 900 Linux workstations and servers, and ca. 900 servers and workstations with MS-Windows, primarily in the domain science.ru.nl. Dual boot Windows/Linux machines are counted twice.


Workstations

C&CZ manages (for departments) workstations with Ubuntu Linux or MS-Windows. There are also workstations in the PC-rooms and Study area for general use by anybody with a Science login, but also by all RU employees and students and guests with a temporary login. Almost all those machines are dual boot, the user can choose to boot Ubuntu Linux or MS-Windows.

Login-servers

  • lilo and stitch, the Linux login-server for employees and students.

Ways to log into the login servers:

  • With SSH (Secure Shell), which is an encrypted connection, so passwords can't be sniffed. An SSH connection is possible from all over the Internet. It is also possible to tunnel X11 traffic.

The automatic monitoring of the servers can be accessed on the web.

Compute servers/clusters

C&CZ manages a number of compute clusters, with compute nodes that departments have purchased. In a few cases, departments have created a separate cluster by also purchasing their own head node (master). The largest cn-cluster has a head node purchased by C&CZ. For departments of FNWI the housing and management of clusternodes is free of charge. The cn-cluster is heterogeneous, the oldest nodes still have Fedora Linux with Sun GridEngine for clustersoftware, newer nodes have Ubuntu Linux with Torque / MAUI. The plan is to move to SLURM. As of August 2014, the cluster totalled 1324 cores and 7914 GB RAM. Several clusternodes have a GPU with CUDA cores. The CPU-load of the clusternodes is being monitored.

The Astrophysics coma cluster with 320 cores, 240 TB discs (as of August 2014), GPU's and Infiniband interconnect is a separate cluster. Separate clusters are also managed for the RU departments Centre for Language Studies en Donders Centre for Cognition.

Mail server

On the Do-It-Yourself website each user can see which mail server will store incoming mail for his account. Mail can be accessed with IMAP over SSL. The use of POP3 is supported, but not advised. In every domain managed by C&CZ the alias smtp.science.ru.nl points to the machine that accepts mail with the SMTP-protocol. N.B.: If one uses a (non-RU) Internet provider, our mail server will only accept mail for our (RU) mail-addresses. Most other mail will not be accepted ("Relaying Denied!"). One can prevent this by using the smtp-server of the Internet-provider or by using Authenticated SMTP.

The amount of mail and spam is being logged.

WWW servers

There are several WWW-servers with websites and web applications.

Printer servers

Several servers are used to print jobs. See the printer webpage to see which server is advised in which situation.

Disk servers

These servers share different types of files, e.g. home-directories, application-software and disks for departments and groups. All modern servers only have redundant disks, sometimes by using "mirroring", sometimes by use of a "RAID-array". That means that loss of only 1 disk will not cause any problem for users.

For more information, e.g. about renting disc space for departments, see our disc space page.

FTP server

To make the transfer of large files to others simple, there is an FTP service. It can be accessed through the original FTP protocol as well as through the web HTTP protocol. Anyone with a Science-account who acquired a directory on this server through postmaster, can copy files to that directory in several ways:

  • by logging in to a login server like lilo in /vol/ftp/pub/yourdirectory.
  • by attaching the network drive: \\ftp.science.ru.nl\ftp

By using an unreadable upper directory containing a secret directory, one can even transfer somewhat sensitive data.

Backup servers

Almost all disks that are managed by C&CZ, are being backupped regularly, in order to be able to restore data in case of small or large calamities.


Lijn.png