Glossary

active/active cluster

A configuration of two SQL Server installations in a clustered environment where each node is initially configured with one SQL Server installation but where either node can host both active SQL Servers. This term applies to SQL Server 6.5 and SQL Server 7.0 clusters only. It has been replaced by a multiple- instance cluster for SQL Server 2000.

active/passive cluster

A configuration with one SQL Server installation in a clustered environment where one active node runs the SQL Server resources, and the other is waiting for a potential failover. This term applies to SQL Server 6.5 and SQL Server 7.0 clusters only. It has been replaced by a single-instance cluster for SQL Server 2000.

change management

The process allowing changes to applications to occur in a predictable fashion with minimal or no impact on the service. Change management applies to all phases of a lifecycle.

client/application redirection

The process of diverting client or application usage to another SQL Server database in the event of a replication failure ”if replication is the high-availability method employed ”or in the event of a role change.

cluster group

A collection of cluster resources analogous to a folder on your hard drive that contains multiple files.

cluster node

A physical server participating in a cluster.

cluster resource

Object types which may be added to cluster resource groups to provide a specific type of cluster functionality that can be utilized by the cluster. Default cluster resource types include DHCP servers, MSDTC, file shares, generic applications/services, servers running IIS, IP addresses, network names , MSMQ, physical disks, print spoolers, and time services.

cluster resource group

See cluster group .

cluster-aware application

An application that runs on a node, is managed as a cluster resource, and is designed to be aware of and interact with the server cluster environment. SQL Server 2000 is an example of a cluster-aware application. The clustering application program interface provides the programming elements necessary to work directly with cluster objects and to interact with the Cluster Service service.

clustering

Connecting two or more computers in such a way that they behave like a single computer to an application or client. Clustering is used for parallel processing, load balancing, and fault tolerance.

convergence

The process used by Network Load Balancing to reach consensus on what the Network Load Balancing cluster looks like. See Network Load Balancing.

DAS

See direct attached storage .

default instance

An installation of SQL Server 2000 that generally uses the name of the underlying server, or a specific name in a server cluster that is the name of neither an individual node nor the server cluster itself. A default instance is similar to the experience of having one installation of SQL Server 7.0 or earlier per server. There can only be one default instance of SQL Server on an individual standalone server or a single server cluster. All other instances must be named instances.

differential database backup

A type of database backup that only backs up changes made to the database since the last full database backup.

direct attached storage

An external disk array connected to one or more servers via traditional disk protocols such as SCSI.

disaster recovery

The process of executing documented procedures to restore a server or an entire environment after a catastrophic problem occurs.

disk page

The fundamental unit of data storage for SQL Server. The page size in SQL Server 2000 is 8192 bytes, or 8 KB.

downtime

The amount of time that a given system or solution is not available for servicing requests .

extent

The basic unit by which space is allocated to a table or an index. An extent is eight continuous disk pages, which equates to 64 KB.

failover

The process in which, in response to a node or resource failure, resources in a cluster switch ownership and start on another node in the server cluster. This term applies to a server cluster/failover cluster combination only. In log shipping implementations , the similar terms switch and role change are used to describe the change in ownership that occurs between primary and secondary servers.

failover clustering

The SQL Server 2000 implementation of clustering for availability that is built on a Windows Server Cluster.

federated cluster

A grouping of SQL servers used together to achieve scalability by employing a distributed partition view. A federated cluster is not used for availability, only for achieving scalability through scale out.

filegroup

A named collection of one or more disk files that represent a single allocation on or for administration of a database. A database can have more than one filegroup.

five nines

Shorthand for a system s being available, or having uptime, 99.999 percent of the time.

full database backup

A complete point-in-time backup of a database.

heartbeat network

See private network .

IsAlive process

A check in the server cluster, run after LooksAlive, that is an application- specific health check. For SQL Server 2000, this issues a SELECT @@Servername query to determine if the SQL Server instance can handle requests.

load testing

The process of testing an application, entire solution, or hardware component under realistic conditions, or load, to ensure that it will behave properly in a highly utilized production environment.

Log Sequence Number (LSN)

A unique identifier assigned to a record in the transaction log that is incremented automatically.

log shipping

The process of backing up a transaction log on one SQL Server, copying it, and then applying it to another SQL Server where a full database backup of that original database has been applied with the proper settings.

log shipping pair

The log shipping primary/secondary combination.

log shipping primary

The server containing the database source of the transaction logs.

log shipping secondary

The server containing the database recipient of the transaction logs in a log-shipping pair.

logical disk

See logical drive .

logical drive

Any volume created that Windows can see, after it has been formatted and assigned a drive letter. You can have multiple logical drives per volume.

logical unit (LUN)

See volume .

LooksAlive process

A specifically coded application-level health check in a server cluster to ensure that the application is up and running. For SQL Server, this is a lightweight check.

LSN

See log sequence number .

LUN

See volume .

LUN masking

A relationship between a specific volume and a host at the disk controller level.

mean time between failures (MTBF)

The average expected time between failures of a specific component.

mean time to recovery (MTTR)

The average time it takes to recover from a failure.

Microsoft Cluster Service (MSCS)

The original name for a server cluster under Windows NT 4.0.

MSCS

See Microsoft Cluster Service .

MTBF

See mean time between failures .

MTTR

See mean time to recovery .

multiple-instance cluster

A SQL Server 2000 failover cluster with more than one SQL Server virtual server.

named instance

An installation of SQL Server 2000 that has a unique name to differentiate itself from other SQL Server instances. The name must be unique for both the server as well as the domain. There can be multiple named instances of SQL Server on an individual standalone server or a single server cluster.

Network-Attached Storage

Low-cost, high-density storage generally used to store files; generally not used with databases.

Network Load Balancing

The type of clustering under Windows Clustering that provides scalability and availability for IP-based services and some SQL Server uses.

nine

Refers to the digit 9 in a percentage that represents availability. For example, three nines is 99.9 percent.

node

See cluster node .

perceived downtime/unavailability

The amount of time that a system or application is unavailable to those who normally access it from a business standpoint, although the server itself may be up and running and having normal maintenance performed on it, or one part of the overall solution is down, preventing access to that server.

physical disk

The actual disk used internally in a system or as part of a disk array.

primary

See log shipping primary .

private network

In a cluster, the dedicated intra-cluster network used to run processes that check to see if the cluster is up and running. For non-clustered machines or other uses, a private network is one that is segmented off from the public network on a different subnet to offload tasks such as large file copies.

public network

On a standalone server or a cluster, the network that allows clients and applications to access the server itself.

quorum

In a server cluster, contains the master copy of the server cluster s configuration and is also used as a tie-breaker if all network communication fails between the nodes. Depending on the type of server cluster implemented, this may or may not be a physical disk on the shared cluster disk array.

RAID

See redundant array of independent disks .

recovery model

The database-level setting ”either Bulk-Logged, Full, or Simple under SQL Server 2000 ”that controls the amount of logging that occurs.

redundant array of independent disks (RAID)

A specific fault-tolerant disk array system design strategy that takes into account issues of cost benefit, reliability, and performance. It can be implemented at a hardware or a software level; each provides a different profile of cost, reliability, and performance. Depending on the person defining RAID, the word independent may be substituted with inexpensive .

resource group

See cluster resource group .

role change

The process of switching from the log shipping primary to the secondary; applies to log shipping only. This is a manual process.

run book

A centralized and continually updated collection of information such as contacts, phone numbers , disaster recovery plans, software license information, support information, and so on. It is used in the event of an emergency to ensure the system can be brought back online in a timely fashion.

SAN

See storage area network .

secondary

See log shipping secondary .

server cluster

The type of Windows Clustering that provides availability only. It is a collection of nodes that allow resources to be failed over to another node in the event of a problem. SQL Server 2000 failover clustering is installed and configured on top of a server cluster.

service level agreement (SLA)

A signed agreement of system service requirements between two parties (such as your company and an ASP or between your department and end users) that defines the guidelines, response times, actions, and so on, that will be adhered to for the life of the agreement.

shared cluster disk array

In a server cluster, the DAS or SAN that contains the disks used by the nodes.

single-instance cluster

This is a SQL Server 2000 failover cluster configured with only one SQL Server virtual server. This replaces the term active/passive.

SLA

See service level agreement .

spindle

Another name for an individual physical disk.

standby server

A server configured to be brought online in the event of a failure of the current server used in production. Standby servers are commonly referred to as hot or warm standby servers to reflect how quickly the server can be pressed into production.

storage area network (SAN)

An external disk array connected to one or more servers via underlying networking protocols, such as fibre, and requiring specialized hardware. SANs provide flexibility, availability, scalability, reliability, and ease of configuration.

switch

The process of switching from one server to another in the event of a problem. This is usually used for non-clustered implementations where a manual process is implied .

transaction log backup

A backup of the transaction log that flushes the transactions from the transaction log to a file. To have transaction log backup integrity, each consecutive file must not break the LSN chain.

uptime

The amount of time that a given system or solution is available for servicing requests.

virtual server

The combination of the network name and IP address that makes up the clustered application or server that a client or application will access. This appears as a normal application or server. Both the server cluster itself, and a SQL Server 2000 instance installed on a cluster, can be considered virtual servers.

volume

The amount of physical disk that Windows recognizes after the storage is configured at a hardware level (whether one individual disk or a RAID subsystem). The volume is then formatted and configured in other ways so Windows and applications can use it in different ways.

warm standby

A server waiting to be brought online; generally refers to the secondary in a log shipping pair.

Windows Clustering

The umbrella group of clustering technologies for the Windows platform, which currently includes a server cluster and Network Load Balancing.

zoning

Isolating volumes on a SAN so that only designated computers that have access to the zone can see them and activity in that zone will not affect other volumes and users on the SAN.