Compute Cluster Server Glossary

Applies To: Windows Compute Cluster Server 2003

ABCD E FGHIJ K L MNOP Q RSTU V W X Y Z

A

Active Directory

Active Directory® directory service is a central component of the Windows platform that provides the means to manage the identities and relationships that make up network environments. Active Directory stores information about objects on a network and makes this information available to users and network administrators. Active Directory gives network users access to permitted resources anywhere on the network using a single logon process. It provides network administrators with an intuitive, hierarchical view of the network and a single point of administration for all network objects. Active Directory is required for Microsoft® Windows® Compute Cluster Server 2003. See also: domain, forest, object

Active Directory Users and Computers

An administrative tool used by an administrator to perform day-to-day Active Directory administration tasks. The tasks that can be performed with this tool include creating, deleting, modifying, moving, and setting permissions on objects stored in the directory. Examples of objects in Active Directory are organizational units, users, contacts, groups, computers, printers, and shared file objects.

ASP.NET

A component of the Microsoft .NET Framework for building, deploying, and running Web applications and distributed applications*.*

B

backfilling

A scheduling policy that allows a job or jobs lower down in the queue to run ahead of a waiting job at the top of the queue, as long as the job at the top is not delayed as a result.

C

Cluster Management Service

The essential software component that controls all aspects of compute cluster operation and that manages the cluster database. This service on the head node provides overall cluster management of node discovery, as well as configuration management. Each node in the cluster also runs one instance of the Cluster Management Service, which communicates with the Cluster Management Service on the head node and is responsible for node discovery within the cluster.

Cluster MPI Service

Software that allows MPI executable files to be run as tasks on the compute cluster.

Windows Compute Cluster Server 2003 uses a Microsoft implementation of MPI developed for Windows, called MS MPI, which is based on (and has identical programming APIs) to MPICH2 developed by Argonne National Laboratory. This version of MPI is automatically installed on each compute node when the Microsoft Compute Cluster Pack is installed.

Cluster Node Manager Service

Software that runs on every compute node and is responsible to communicate with the Job Scheduler Service on the head node and to run jobs in their proper sequence.

Cluster System Definition Model (SDM) Store Service

Software that is responsible for maintaining the integrity of read and write data from the SDM data store, which is used to store cluster configuration information.

Command-line interface (CLI)

A form of interface between the operating system and the user in which the user types commands, using a special command language. In Windows Compute Cluster Server 2003, this interface provides a command-line alternative to the use of the Compute Cluster Administrator and the Compute Cluster Job Manager for executing user actions relating to jobs, tasks, nodes, and cluster configuration.

common language runtime

The core runtime engine in the Microsoft® .NET Framework for running applications. The common language runtime supplies managed code with services such as cross-language integration, code access security, object lifetime management, and debugging and profiling support.

Compute Cluster Administrator

Graphic interface used by the administrator to manage all aspects of compute cluster function, including configuration and management of the head node and compute nodes, and review of cluster status. The Compute Cluster Administrator is a snap-in to Microsoft Management Console (MMC).

Compute Cluster Job Manager

Windows application interface where the user or administrator creates, submits, and manages jobs.

compute node

An x64-based computer configured as part of a compute cluster to provide computational resources for the end user to run jobs.

Compute nodes have the following services installed:

  • Compute Cluster Management Service

  • Compute Cluster MPI Service

  • Compute Cluster Node Manager Service

D

DHCP

See definition for: Dynamic Host Configuration Protocol (DHCP)

domain

In Active Directory, a collection of computer, user, and group objects defined by the administrator. These objects share a common directory database, security policies, and security relationships with other domains.

In DNS, any tree or subtree within the DNS namespace. Although the names for DNS domains often correspond to Active Directory domains, DNS domains should not be confused with Active Directory domains.

domain controller

In an Active Directory forest, a server that contains a writable copy of the Active Directory database, participates in Active Directory replication, and controls access to network resources. Administrators can manage user accounts, network access, shared resources, site topology, and other directory objects from any domain controller in the forest.

Dynamic Host Configuration Protocol (DHCP)

A TCP/IP service protocol that offers dynamic leased configuration of host IP addresses and distributes other configuration parameters to eligible network clients. DHCP provides safe, reliable, and simple TCP/IP network configuration, prevents address conflicts, and helps conserve the use of client IP addresses on the network.

DHCP uses a client/server model where the DHCP server maintains centralized management of IP addresses that are used on the network. DHCP-supporting clients can then request and obtain lease of an IP address from a DHCP server as part of their network boot process.

E

There are no glossary terms that begin with this letter.

F

forest

One or more Active Directory domains that share the same class and attribute definitions (schema), site and replication information (configuration), and forest-wide search capabilities (global catalog). Domains in the same forest are linked with two-way, transitive trust relationships.

G

Gigabit Ethernet

The Ethernet standard that transmits data at one billion bits per second or more. GigE

See definition for: Gigabit Ethernet

H

head node

Server that provides management and job scheduling services to the compute cluster. User interfaces consist of the Compute Cluster Administrator, the Compute Cluster Job Manager, and a Command Line Interface (CLI). Management services include job scheduling, job and resource management, Remote Installation Services (RIS), and a remote command environment. The head node can also serve as a network address translation (NAT) gateway (part of Internet Connection Sharing) between the cluster private network, if one exists, and the public network.

heartbeat

Network signal sent out by the head node once every minute to each compute node in the cluster. This heartbeat signal verifies node health and availability. If a compute node cannot be contacted after three tries, the node status shown in the Compute Cluster Administrator console changes to Unreachable.

I

InfiniBand

Networking architecture used in high-end servers. InfiniBand uses switched, point-to-point channels similar to mainframes and also similar to PCI Express, the switched version of the PCI bus. InfiniBand provides data bandwidth from 500 MB per second to 6 GB per second at very low latencies, and is designed as a fabric architecture that can extend connections via external networks.

Internet Connection Sharing (ICS)

Microsoft Windows technology that allows computers on a network to access online services through a single Internet connection.

J

job

A container that contains a task or sequence of tasks that runs on the cluster. Jobs can take the form of serial tasks, parametric sweeps, dependency-based workflows, or one or more parallel tasks using MS MPI or other message passing mechanisms.

Job Manager

See definition for: Compute Cluster Job Manager

job object

A system-level structure that allows processes to be grouped together and managed as a single unit.

job owner

Person who submitted a job to the compute cluster.

job priority

A property assigned to a job that helps to determine its place in the queue. Job priorities include: Highest, Above Normal, Normal, Below Normal, and Lowest.

job queue

List of jobs that have been submitted to the job scheduler. The order in which jobs in the queue are run is based on job priority class and, within priority class, time of submission. Queued jobs move through the life cycle Queued-Running-Finished or Queued-Running-Failed but can be cancelled at any time during the Queued or Running state. By default, jobs remain in the queue for five days regardless of status.

Job Scheduler Service

Software responsible for job queuing, scheduling, resource allocation, and job execution.

Job status

State of a job with respect to the job queue. Job status types include:

  • Not Submitted — Job that has been created but not added to the queue. In this state, a job is not yet eligible for resource allocation and activation.

  • Queued — Job that has been submitted to the queue but not yet started because it is waiting its turn for the required resources.

  • Running — Job that is being processed.

  • Finished — Job that has successfully completed all of its tasks and has released its resources.

  • Cancelled — Job that has been removed from active status (Queued or Running) and will not be completed. A job can be cancelled by a user or an administrator.

  • Failed — Job that has at least one task that has not been successfully completed, due to an error. A failed task can go back to Queued state if the error was not caused by the application itself.

job template

An XML file to which users can save job parameters for later submission or reuse.

K

There are no glossary terms that begin with this letter.

L

There are no glossary terms that begin with this letter.

M

Media Access Control (MAC) address

The address that is used for communication between network adapters on the same subnet. Each network adapter is manufactured with its own unique MAC address. Message Passing Interface (MPI)

Standard that defines a software library used by two or more processes running on the same or different physical computers to communicate. Typically, these systems are clusters of servers or networks of workstations. The standard was created by the MPI Forum in 1994 and is now the de facto standard for high-performance computing.

Windows Compute Cluster Server 2003 uses a Microsoft implementation of MPI developed for Windows, called MS MPI, which is based on (and has identical programming APIs) to MPICH2 developed by Argonne National Laboratory. For more information about MPI, see The Message Passing Interface (MPI) Standard (https://go.microsoft.com/fwlink/?LinkId=51266).

Microsoft Management Console (MMC)

A Microsoft management display framework that provides a unified view for administering multiple integrated network applications. In Windows Compute Cluster Server 2003, the MMC snap-in (called Compute Cluster Administrator) that is used by cluster administrators to manage the hardware, software, and network components of the cluster.

MPI

See definition for: Message Passing Interface

MPI network

A cluster network that supports Message Passing Interface (MPI) and other parallel processes. An MPI network is used for parallel application communication among the compute nodes in the cluster, separating latency sensitive MPI traffic from cluster management traffic on the network.

MSDE

Microsoft® SQL Server™ 2000 Desktop Engine (MSDE 2000). A data store based on Microsoft SQL Server™ technology, but designed and optimized for use on smaller computer systems, such as a single-user computer or a small workgroup server.

MS MPI

Software that allows MPI executable files to be run as tasks on the compute cluster.

Windows Compute Cluster Server 2003 uses a Microsoft implementation of MPI developed for Windows, called MS MPI, which is based on (and has identical programming APIs) to MPICH2 developed by Argonne National Laboratory. This version of MPI is installed automatically on each compute node (and optionally on the head node) when the Microsoft Compute Cluster Pack is installed.

N

NAT

See definition for: Network Address Translation (NAT)

.NET Framework

An integral Windows component that supports building, deploying, and running the next generation of applications and XML Web services. It provides a highly productive, standards-based, multilanguage environment for integrating existing investments with next generation applications and services, as well as the agility to solve the challenges of deployment and operation of Internet-scale applications. The .NET Framework consists of three main parts: the common language runtime, a hierarchical set of unified class libraries, and a componentized version of ASP called ASP.NET. See also: ASP.NET, common language run time, .NET Framework class library*.*

.NET Framework class library

A CLS-compliant library of classes, interfaces, and value types that are included in the Microsoft .NET Framework SDK. This library provides access to system functionality and is designed to be the foundation on which .NET Framework applications, components, and controls are built.

Network Address Translation (NAT)

The translation of an Internet Protocol (IP) address in one network to a different IP address in another network. One network is designated the internal network and the other is the external. The internal network appears as one entity to the outside world. The NAT used by Compute Cluster Server 2003 is provided by installing Internet Connection Sharing on the head node.

node

Generic reference to the head node and compute nodes in a cluster.

Node Manager Service

Software that runs on all computers in the cluster, including the head node if the head node is configured as a compute node. The Node Manager runs jobs on the node, sets task environment variables, and sends a heartbeat signal to the Job Scheduler (by default, every one minute).

node status

Describes the state of a node in a cluster. Possible node states includes:

  • Pendingfor Approval — The default node status after running Microsoft Compute Cluster Pack setup on a compute node. The cluster administrator must approve or reject the node for inclusion in the cluster. When approved, the node status changes to Paused.

  • Paused —A state in which the node is reachable and can run applications, but is not available for allocation to a job. When a node is in the Paused state, a cluster administrator can run scripts, install software, and perform other tasks on the node. If a node is placed in a Paused state while running jobs, those jobs will complete, but no new jobs will be accepted from cluster users. However, cluster administrators can submit a job to a paused node, although the job will not appear in the job queue. When the node is resumed, status changes to Active and jobs will be assigned to that node by the Job Scheduler.

  • Active —Normal operating state of a compute node. A node with this status runs jobs and receives new jobs from the Job Scheduler.

  • Unreachable —Cluster node cannot be reached by head node. The head node sends out one heartbeat per minute to check on the cluster nodes. If the head node has been unable to contact a compute node for three minutes, it places the node in the Unreachable state.

nonexclusive scheduling

A scheduling policy that allows tasks from different jobs to share a node.

O

object

An entity, such as a file, folder, shared folder, printer, or Active Directory object, described by a distinct, named set of attributes. For example, the attributes of a File object include its name, location, and size; the attributes of an Active Directory User object might include the user's first name, last name, and e-mail address.

P

parallel processing

The simultaneous use of more than one CPU to run a program.

parallel tasks

Tasks whose computations are independent of each other, so that all such tasks can be performed simultaneously with correct results.

parametric sweep

Parallel computing job that consists of running the same executable file on multiple nodes with different input and output files.

paused

See the definition for: Node status

Preboot Execution Environment (PXE)

A DHCP-based remote boot technology that is used to boot or install an operating system on a client computer from a remote server. A Remote Installation Services server is an example of a PXE server.

preemption scheduling

A scheduling policy that allows a pending high-priority job to take resources away from a running job of lower priority.

private network

An intracluster network connecting the nodes of a cluster. This network is optional, and when present, carries intranode cluster network traffic and in the absence of an MPI network, also carries any MPI network traffic that occurs between nodes.

public network

The larger organizational network to which a cluster connects. The public network generally hosts the users and resources (such as file and database servers) required by the jobs submitted to the cluster.

PXE

See definition for: Preboot Execution Environment

Q

queue

See definition for: Job queue

R

refresh interval

The interval at which job and task status is refreshed at the Compute Cluster Job Manager. Status can be set by the user to automatically refresh at the following intervals: 15 seconds, 30 seconds, 1 minute, 5 minutes, and 10 minutes. The default is every 15 seconds.

Remote Installation Services (RIS)

Remote Installation Services (RIS) is an automated installation technology that can be used to create installation images of operating systems. These installation images can then be made available to users at client computers. The destination clients must support remote restarting using Preboot Execution Environment (PXE).

RIS

See definition for: Remote Installation Services

S

scheduling policy

Scheduling rules implemented by the Compute Cluster Job Scheduler that determine the order in which submitted jobs are processed.

snap-in

A software component that provides system administration and system management capability within the framework of Microsoft Management Console (MMC). A stand-alone snap-in can be added by itself; an extension snap-in can be added only to extend the function of another snap-in. The Compute Cluster Administrator is the MMC snap-in that is used to manage a Windows Compute Cluster Server 2003 cluster.

See also: Microsoft Management Console (MMC)

stack

A data area or buffer used for storing job requests that need to be handled.

submitting user

Person who submits a job to the cluster. The submitting user is the owner of the job.

T

task

A logically discrete section of computational work. A task cannot be run independently of a job, but a job can consist of only one task.

throughput

A network's data-transfer rate, which is usually measured as the number of transmitted megabits per second (Mbps) or kilobits per second (Kbps).

U

user account

In Active Directory, an object that consists of all the information that defines a domain user, which includes user name, password, and groups in which the user account has membership. User accounts can be stored in either Active Directory or on your local computer. For computers running Windows XP Professional and member servers running Windows Server 2003, use Local Users and Groups to manage local user accounts. For domain controllers running Windows Server 2003, use Active Directory Users and Computers to manage domain user accounts.

user rights

Tasks that a user is permitted to perform on a computer. User rights are controlled by security settings on the computer and are usually set by the system administrator.

V

There are no glossary terms that begin with this letter.

W

Web Services Enhancement (WSE)

Web Services Enhancements (WSE) is a fully supported extension of Visual Studio and the .NET Framework that provides an implementation of the latest advanced Web services capabilities and industry supported Web services protocols (WS-*). Win32 API

A 32-bit Windows operating system application programming interface (API) that enables applications to use the 32-bit instructions available on 80386 and later processors. The Win32 API includes sophisticated operating system capabilities, security, and API routines for Windows-based applications.

WSE

See definition for: Web Services Enhancement

X

There are no glossary terms that begin with this letter.

Y

There are no glossary terms that begin with this letter.

Z

There are no glossary terms that begin with this letter.