Lesson11.Glossary

Lesson 11. Glossary

affinity

An association (affinity) between a folder on an Xsan volume and one of the pools that make up the volume. The affinity guarantees that files that users place in the folder are stored only on the pools that have been formatted to meet specific capacity, performance, and reliability goals.

AFP (Apple Filing Protocol)

A network file system protocol used to share files and network services. AFP uses TCP/IP and other protocols to communicate between systems on a network.

arbitrated loop

This term may arise when planning a storage area network (SAN), as the customer may want to save money by deploying a Fibre Channel hub rather than a Fibre Channel switch. The internal wiring of the hub has all the ports connected as a physical loop and uses a special protocol that defines how the connected systems gain control of the loop to communicate with other devices connected to the hub. Note that in an arbitrated loop, the bandwidth is shared among all of the systems and devices. So if system 1 has 200 GB of data to transmit, and system 2 has 350 GB to transmit, and the hub supports 100 GB/s, they arbitrate with each other to obtain control of the 100 GB/s hub and may end up with each one having a certain number of usage spots to transmit 100 GB at a time.

autoloader

A tape backup system that uses a robotic mechanism to automatically load and unload cartridges into a tape drive. When combined with a software solution for hierarchical storage management or information lifecycle management, users may not even be aware that the file is being retrieved from or stored to tape. Autoloaders can also be used to provide unattended data backup and file restoration.

availability

For users with high-reliability requirements solutions or downtime/uptime requirements, availability is a metric they may specify for their solution (see also MTBF). Typically, the number is based on the amount of time that a system is available during those periods when it is expected to be available (for example, excluding scheduled backups). Availability is often measured as a percentage of an elapsed year. For example, 99.95 percent availability equates to 4.38 hours of downtime in a year (0.0005 x 365 x 24 = 4.38) for a system that is expected to be available all the time. This is a key measurement for mission-critical systems. Providing the highest level of availability (99.999 percent) usually requires redundant servers and access to shared storage such as an Xsan solution provides.

bandwidth

Bandwidth (synonymous with data transfer rate) is used to describe the amount of data that can be transferred through a system or network connection in a given time period. Different connection types (Ethernet, Fibre Channel, FireWire) and the protocols supported (TCP/IP and SCSI, for example) on those connection types will dictate the maximum potential data rate. The rate is usually indicated in bits or bytes per second. It is important to assess bandwidth requirements when deploying solutions such as Web servers, network-based storage, and creative workflow applications, to ensure bandwidth to the storage will meet the requirements. For example, a video-streaming deployment will need high-bandwidth networking solutions between the storage and servers to ensure an adequate frame rate. The number of simultaneous streams will need to be assessed to determine the specific network topology.

block

Storage devices contain disks that are divided into tracks and then into sectors. When the disk is formatted, it can be optimized for performance by specifying the number of sectors the operating system can read from (or write to) the disk at a time. This setting is the block factor. By default, a sector of data is typically 512 bytes. The application that the storage will be used for, however, may support larger block sizes for faster performance. Video and database deployments, for example, should include a step to determine the block size needed to sustain the desired I/O.

cache

This refers to special-purpose memory on a server or storage device used to increase performance. Storage cache usually resides on RAID controllers and boosts performance because the CPU doesn't have to wait for a disk head to spin. Data can be written to and read directly from cache. For example, cache can be used to store data fetched from a disk, which is particularly useful if it is data that may be accessed frequently, or by multiple applications or multiple users. Some solutions support prefetching algorithms that essentially guess and fetch anticipated data from disk, based on the original data request. This also decreases wait time. Ideally, the cache size is a multiple of the block size so the operating system can retrieve complete data blocks.

cascading

This refers to deploying switches in a tiered manner to increase the number of simultaneous connections to a server or storage device. This enables reliability and scalability in a network/storage fabric. It is critical to plan this carefully to ensure the desired I/O performance is maintained. For example, an Xserve RAID has two Fibre Channel ports. Each of those ports can be connected to a Fibre Channel switch that may have 16 ports, which would allow 16 connections to the Xserve RAID. If more connections were needed (which may be in an Xsan configuration), then that first switch can connect to up to 16 more switches. Note, however, that the more simultaneous connections are enabled, the greater the potential is for slower performance. For example, the scenario just outlined could enable 256 servers retrieving data from one Xserve RAID, which may be suitable for some transactions, but not for something like video streaming. Some manufacturers refer to this as inter-switch linking or trunking.

client

For the licensing of Xsan, a client is the system directly attached to the SAN via a Fibre Channel connection. Clients can also refer to a system on a network that receives network services from a server, such as client systems connecting to a file or print server. It can also refer to a software program that requests services from other programs or servers. Pay careful attention to the context. For example, someone might describe user systems as clients that can attach to file servers, and those servers in turn may be clients to a SAN file system. It is important to interpret the context as this becomes important when you are discussing licensing costs and planning deployments.

cluster file system

Cluster file systems allow multiple systems to simultaneously access the same storage. This differs from a traditional file system like HFS+, which can be shared only by using network protocols like NFS, AFP, and SMB. Xsan and other SAN file systems are sometimes referred to as cluster file systems.

DAS (direct-attached storage)

Refers to disks inside a system or directly attached as through a SCSI, FireWire, USB, or Fibre Channel interface. For example, when Xserve RAID is connected to the Fibre Channel PCI card interface in a server or workstation, the RAID is "direct attached" to that server or workstation. DAS is generally considered very fast since the storage can be accessed at block I/O rates. It can also be shared with other computers using a network sharing protocol, such as AFP, SMB, CIFS, or NFS, but this is at file I/O rates. Xsan solves this performance issue by allowing multiple systems to access the same volumes and files, concurrently, at block I/O rates over a Fibre Channel network.

directory services

A mechanism of managing information about users, groups of users, and equipment. Organizing this information in a centralized database allows administrators to set up relationships between people and resources in one place, rather than controlling access by configuring each user's system. Active Directory and Open Directory are examples. Xsan can use directory services to enable features such as user storage quotas.

DNS (domain name service)

A software service that enables translation of domain names (for example, www.apple.com) into an IP address. Servers with DNS are generally available from Internet service providers to support their users accessing Web sites with names instead of IP addresses. Within a company, a network system administrator may also choose to enable DNS (available with Mac OS X Server) to support fast identification of hosts within the private network or SAN.

ECC RAM (error correcting code)

This type of RAM is used in Xserve G5 CPUs. ECC RAM can detect errors (both single- and multibit) and is able to fix single-bit errors on the fly. This is transparent to the user and is extremely useful in systems that are kept online constantly, such as Xsan metadata controllers.

fabric

In a storage area network, with or without a cluster file system like Xsan, the fabric is the collective term for the Fibre Channel HBAs, switches, and cables that connect the servers and storage devices. A SAN installation may contain multiple fabrics for redundancy or scalability.

failover

A strategy associated with high-availability deployments. A failover configuration increases the reliability of a server or storage solution. Generally, a failover configuration includes two or more systems configured identically but with monitoring between them. For example, a pair of servers configured with failover (the high- and medium-priority controllers) will have a primary server that responds to metadata network requests, and a secondary, identical server that monitors the status of the primary server. The secondary server services requests only when the primary server becomes unavailable.

Failover may also be mentioned in the context of storage solutions. In this case, the high-availability and reliability points a customer may want to discuss include redundancy of RAID controllers and mirrored RAID sets.

Fibre Channel switch (Fabric Switch)

Used to create a high-performance Fibre Channel network, a switch enables multiple connections between servers and storage at full (rather than shared) Fibre Channel bandwidth. So if server 1 needs to transmit 200 GB, and server 2 needs to transmit 350 GB, and the switch supports 100 GB/s, then both servers get a connection of 100 GB/s.

headless system

A system that does not have a display attached. An Xserve G5, for example, does not need a graphics display and can be administered either remotely using the administration software included or from the command line using SSH. In cases where an administrator wants to log in, the operating system creates a virtual frame buffer that allows applications to be displayed remotely on other systems, or displays an entire virtual desktop (as when being accessed through Apple Remote Desktop).

HBA (host bus adapter)

The interface on a storage device, a server, or workstation used to convert outbound block data to packets suitable for transport over the applicable format of the adapter (for example, FireWire, SCSI, Fibre Channel) and to convert inbound packets to block data. Xsan requires Fibre Channel HBAs in order for the Xsan Admin program to recognize storage.

hot spare

This refers to a drive in a RAID configuration that is unused but allocated to be available if a drive in a RAID-1 or RAID-5 configuration fails. The hot spare will assume the role of the failed drive, and the data will be automatically rebuilt. This is a deployment design strategy and is applicable for users with high-reliability and high-availability needs.

iSCSI

A newer storage networking protocol that attempts to transmit blocks of data more effectively over IP (as for Ethernet) than NFS, for example. iSCSI is still limited by issues described in file I/O, although it attempts to approach block VO performance. Users may ask for a comparison of Xsan to an iSCSI solution. These are not really comparable. An Xsan solution provides much more than any iSCSI solution could, including performance.

LDAP

An industry-standard protocol used in directory service tools (Microsoft's Active Directory, Novell's eDirectory, Apple's Open Directory) to locate organizations, individuals, and other resources such as files and devices (for example Cisco supports LDAP for identification of its devices) in a network. By using an industry-standard protocol, vendors are able to offer more seamless integration with heterogeneous environments.

LUN (logical unit number)

A LUN may be a disk, a RAID set, or a slice of a RAID set. Xserve RAIDs can be subdivided into many LUNs. A LUN will typically be formatted with a specific RAID strategy to achieve performance or data redundancy goals. Even though access can be assigned to a LUN, and this seems to be the same as access to a volume, a LUN and a volume are different. A LUN is basically the address of a physical disk, and while this address can be used for access control, this is a manual strategy that is less often used as it can be somewhat less flexible in capacity management. In contrast, a volume can consist of multiple LUNs.

metadata controller

This software component of a SAN file system manages access to files on the volumes that are part of the SAN file system. Client systems send file requests to the metadata controller, which determines whether the requesting system has permission, and if so, whether the file is already being used for read or write, then grants an access token along with the file location. The requesting system can then access the file. The metadata controller typically resides on a system that is used for no other purpose and communicates to all client systems over a private Ethernet network.

mirroring

A storage strategy used for high-availability, redundancy, and disaster-recovery planning, in which data is written to more than one location, generally at the same time. Mirroring can be set up within a single RAID array by formatting the array with RAID level, or an entire rack of RAID arrays can be mirrored by point-in-time copying to another identically configured rack of arrays.

multipathing

The utilization of both paths (two) between the Fibre Channel ports of a system and the Fibre Channel switch(es) to the storage. When both paths are connected to the same switch, the system will automatically utilize both of them for either high availability or to increase bandwidth. If they are connected to separate switches, the system will utilize them only for high availability. The system will automatically determine which utilization model is most appropriate. If both paths are connected to one switch, then their combined bandwidth will be used for greater throughput. In either configuration, if one path becomes unavailable, the system will automatically fail over all data traffic to the remaining path (high availability).

NAS (network-attached storage)

A general term for stand-alone storage devices connected directly to a local area network (LAN) or a wide area network (WAN) for easy access by multiple servers, multiple operating systems, or multiple clients. These devices typically don't include a full operating system; rather they have a special-purpose operating system designed to handle only specific data read, write, and sharing functions.

NAT (Network Address Translation)

This is a service usually offered on a router, a network device that translates computer addresses between networks, but is also included with Mac OS X Server. NAT is used to convert a public Internet address (as used on the Internet) to an internal network address. This process allows a single unique IP address to represent an entire group of computers to anything outside their network The benefits are increased security since the private network addresses are not shared outside the company, and a reduction in the number of Internet addresses that a company will need.

NFS (Network File System)

A file sharing protocol used on Unix and Linux systems. NFS allows users to access shared folders over Ethernet (TCP/IP).

node

This term is used to refer to a device on a network, such as direct clients of an Xsan storage area network.

partition

Refers to the division of a disk drive into more than one logical unit. This allows a number of activities, such as different operating systems being installed (and booted from); sizing of the disk to match specific operating system requirements; and isolating data storage from the operating system drive to help minimize data corruption. Partitioning is usually done on a laptop, workstation, or perhaps on a server with a single disk.

RAID (redundant array of independent disks)

A collection (array) of disk drives housed in a single chassis that allows management of the array in ways that support requirements for increased data integrity, fault tolerance, and performance. The drives may be configured into one or more LUNs, and collections of LUNs from that array or from multiple arrays can be grouped as volumes.

RAID controllers

Integrated within the RAID array chassis, RAID controllers handle I/O requests (translate the file request into a block location, retrieve it, and vice versa), monitor the status of drives and other array components, and have specialized processors to manage data writes according to the RAID level specified during the formatting of the array.

replication

Refers to the duplication of data (usually a database) to another site, as for backup or access from multiple sites. For example, directory services can be replicated to multiple geographic locations to increase performance for users accessing the data remotely. On a scheduled basis, data is pushed from the master server to the replication servers. Typically only changes are pushed, rather than an entire database.

Samba

A freeware file-sharing solution based on SMB/CIFS that allows UNIX, Linux, IRIX, Mac OS, and other operating systems to provide file or print services to Windows clients or servers.

SMB/CIFS (Server Message Block / Common Internet File System)

A protocol developed by Microsoft for Windows systems, SMB enabled client systems to access network resources such as shared files, printers, and so on. Leveraging an open-source version of SMB, the Internet Engineering Task Force developed a complementary protocol (CIFS) to support Web applications that need more flexible file exchange than offered by FTP.

zoning

An access control mechanism similar to LUN masking, zoning is a manual method to create logical groups in a SAN. Devices attached to the Fibre Channel fabric of a SAN may be grouped into zones based on their LUN number, host name, physical device ID (for example WWN), port number, and so on. Devices in the same zone have visibility and access to each other; devices located in different zones do not.