Glossary

A

ACE An access control entry (ACE) is an entry on an ACL that defines what permissions a security principal has on a resource. For example, an ACE may describe what permissions a user has on a file or folder.

ACL An access control list (ACL) defines what permissions security principals have on resources.

Alias Aliases apply human-friendly identifiers to more-complex and obscure entities. In FC SAN fabrics, aliases are used in zoning to apply human-friendly names to things like 64-bit hexadecimal WWPNs.

ALUA Asymmetric logical unit access (ALUA) is a form of load-balancing technology employed by many midrange storage arrays. ALUA allows all controllers on a storage array to be active, while any single LUN is active on only one controller.

Annualized failure rate Annualized failure rate (AFR) is an industry-accepted calculation that estimates the likelihood of a disk drive failing in a year of sustained use.

API Application programming interfaces (APIs) allow simplified and effective communication between an application and a developer (or the tools a developer will develop).

Archiving The process of storing data that is no longer used or infrequently used. Compliance archiving is the process of securely storing data for legal reasons. Archives are normally full-text indexed with additional metadata applied to assist in searching the archive.

Array In the context of storage and storage networking, an array is a specialized computer system with a lot of storage capacity that is shared out to clients over a storage network.

ASIC An application-specific integrated circuit (ASIC) is a custom-designed chip. A well-known example in the storage world is the ASIC that has been at the center of the 3PAR architecture for a lot of years. This ASIC is specially designed to process storage functions such as zero space detection.

Asynchronous replication Replication technology that offers extremely large distances between two replicating devices and introduces no replication-based latency into the solution. However, asynchronous replication is prone to data loss in the event of an outage.

ATS Atomic test and set (ATS) is a hardware-assisted locking mechanism that speeds up locking and supports extent-based locking, whereby locks can be placed on portions of a volume rather than the entire volume. ATS is now a formal T10 standard that requires OS/hypervisor support, as well as storage array support.

Autonegotiate A network protocol that allows two connected devices to determine the best possible speed to connect with. Sometimes referred to as AN.

Availability A measure of the uptime of a system. 99.99 percent availability in a year tells us that a system has been down for less than 53 minutes in the year. 99.999 percent availability is less than 6 minutes of downtime a year.

AWS Amazon Web Services is the largest cloud computing platform in the world.

B

B2B Buffer-to-buffer (B2B) is a sending credit system employed on FC fabrics. A sender of a frame can send frames along the next hop of a journey only if the sender has a buffer credit available.

B2D Backup-to-disk (B2D) technology allows backups of applications and data to be sent directly to disk rather than to tape.

B2D2C Backup-to-disk-to-cloud (B2D2C) is a combination of backing up applications and data directly to disk, and then at a later date moving those backups to the cloud. Sometimes called disk-to-disk-to-cloud (D2D2C).

B2D2T Backup-to-disk-to-tape (B2D2T) is a combination of backing up applications and data directly to disk, and then at a later date moving those backups to tape. Sometimes called disk-to-disk-to-tape (D2D2T).

Backend The backend of a storage array is where the disk drives and solid-state drives exist.

Backup A copy of data taken specifically for potential recovery purposes and stored securely off-site at a remote location.

Backup network A network dedicated to backup and recovery traffic.

Backup retention The length of time that you keep a given backup before discarding the data in the backup.

Backup window A time window during which backup operations can take place.

BBWC Battery-backed write-back caching (BBWC) is a cache memory subsystem that is protected from data loss during a power failure; a battery maintains power to the cache memory.

BC Business continuity (BC) refers to the ability of the technology systems to survive disasters and keep the business running.

Bit A bit is the smallest unit of information in computing that can have one of two values. In the context of storage, a bit is represented as either a binary 1 or a binary 0.

Byte A byte (B) is a common unit of storage capacity equal to 8 bits. Larger units of bytes include kilobytes, kibibytes, megabytes, mebibytes, terabytes, and tebibytes. The difference between a kilobyte (KB) and a kibibyte (KiB) is that a KB is 1,000 bytes, whereas a KiB is 1,024 bytes. Likewise, a megabyte is 1,000,000 bytes (1,000 × 1,000), whereas a mebibyte is 1,048,576 bytes (1,024 × 1,024).

C

Cache Cache is a relatively small amount of fast storage (usually RAM or flash memory) that is used to temporarily hold read and write data to accelerate access.

Cache hit A cache hit occurs when an I/O operation is performed against cache rather than disk. Cache hits are extremely fast.

Cache miss A cache miss occurs when an I/O operation cannot be satisfied by cache and requires operations to disk. Cache-miss operations are slower than cache-hit operations.

CAS Content-addressable storage (CAS) is a type of object storage that fingerprints every object with a unique identifier based on the object's location in the object store and then uses that fingerprint when retrieving the object. CAS systems are popular as compliance archives, as they naturally provide tamper-proof storage.

Cascade topology In an FC SAN, a cascade topology is a simple row of switches that are each connected to one switch. It is rarely used.

CBT Changed block tracking (CBT) technology is used to keep track of changes to block devices so that snapshots and replication technologies can be more efficient.

CDMI Cloud Data Management Initiative (CDMI) is a SNIA-led initiative aimed at standardizing management of cloud storage services.

CDN A content delivery network (CDN) is a network of servers (usually web servers) that are located in strategic parts of the world to provide faster access to information by caching it closer to users.

CDP A continuous data protection (CDP) system provides very granular backups, and hence, extremely good RPO. CDP systems save changes to the data continuously (often via journaling) rather than on a schedule like a conventional backup system.

CEE Converged enhanced Ethernet (CEE) is a vendor-neutral term used to describe DCB Ethernet.

CHAP Challenge-Handshake Authentication Protocol (CHAP) is a mutual authentication protocol that can be used to identify the authenticity of devices on an IP network.

CIFS Common Internet Filesystem (CIFS) is a term commonly used to refer to the Microsoft Server Message Block (SMB) protocol that is the dominant file-sharing protocol used in Microsoft Windows networks. Most NAS arrays implement SMB. Strictly speaking, the term is now obsolete, as the protocol is now called SMB, but it remains in popular use.

Class of service Classes of service allows for prioritization of certain traffic types and workloads on a system.

CNA A converged network adapter (CNA) is a network card that is capable of Ethernet and FCoE traffic. CNAs can be either PCIe expansion cards or LAN on motherboard (LOM).

Compression A technology that identifies and exploits patterns in data sets to reduce the amount of space required to store data.

Core-edge topology A core-edge topology is the most common FC SAN fabric topology and is based on the hub-spoke, or star network, topology. It is common for a core-edge fabric to have a director class switch at the core and smaller workgroup-type switches at the edge. Core-edge topology is common in environments that have blade servers with embedded FC SAN switches in the blade enclosure.

COS Classes of service (COS) can be used to provide different levels of service, usually within a network.

CoW Copy-on-write (CoW) snapshots work on the principle of copying the original contents of a block to a snapshot area when new data is written in order to preserve the original contents.

CRC A cyclic redundancy check (CRC) is a common technique used in computing to ensure the integrity of data.

D

D2D2C Disk-to-disk-to-cloud (D2D2C) is a combination of backing up applications and data directly to disk, and then at a later date moving those backups to the cloud. Sometimes called backup-to-disk-to-cloud (B2D2C).

D2D2T Disk-to-disk-to-tape (D2D2T) is a way of combining backing up to disk and then eventually moving those backup jobs off to tape.

DAR Direct-access restore (DAR) is a feature of NDMP backups that allows for faster recovery of data as well as the ability to select individual files for restore.

DAS Direct-attached storage (DAS) refers to storage installed locally inside the server or externally as a JBOD. In both cases, the storage is accessible only by the single server it is attached to. A server that has two 500 GB drives installed locally is said to have 1 TB of DAS storage.

DBA Database administrators (DBAs) are the staff in a company who look after the company's databases.

DCB Data center bridging (DCB) is a term often used to refer to next-generation Ethernet technologies. DCB Ethernet is designed and developed under the watchful eye of the Data Center Bridging task group, which is part of the IEEE 802.1 working group. FCoE requires DCB Ethernet.

DCBX Data Center Bridging Capability Exchange (DCBX) Protocol is a link-layer protocol on CEE networks that allows connected devices to determine each other's capabilities as well as have configurations pushed to them.

Deduplication The process of removing duplicate data from a data set in order to store more data in less space. Deduplication, as opposed to compression, identifies larger chunks of duplicate or repeating data.

DID Destination ID (D_ID) is the N_Port ID of the intended recipient of an FC frame and is identified in every FC frame header. DID is used in FC frame routing.

Differential backup A differential backup backs up all data that has changed since either the last full backup or the last differential backup.

Discovery domain Discovery domains are a way to partition an iSCSI SAN into smaller, more manageable and secure zones. Similar to zones in an FC SAN.

DL Data loss (DL) is one of the worst situations to be in as a storage administrator. DL can occur when a system fails and redundant components don't work as expected. In the event of data loss, you will more than likely be looking at recovering data from backups.

Domain ID ID to uniquely identify an FC switch in an FC fabric. Important to be aware of if employing port-based zoning in your fabric.

DR Disaster recovery (DR) is the process of recovering services, usually at another location, in the event that a disaster has taken out the service.

DSM In the Microsoft Windows world, a device-specific module (DSM) is an MPIO module that provides specialized integration with a particular model of storage array.

DU Data unavailable (DU) can occur when a system fails or connectivity to a system fails. In a DU situation, data has not been lost. It is still there on the system; you have just lost access to it.

Dual-controller architecture Storage array with two controllers (sometimes referred to as heads or nodes). These controllers can work in active/active configuration, active/passive, or ALUA.

E

E2E End-to-end (E2E) flow control is a flow control system on FC fabrics, transmitting COS 1 and COS 2 traffic.

EC2 Amazon's Elastic Compute Cloud (EC2) is the world's largest public cloud-based computing service. EC2 is part of Amazon Web Services.

EDA Equipment distribution areas (EDAs) in data centers are where servers and storage arrays reside.

eMLC Enterprise-grade MLC is a form of MLC flash technology that is more suitable than standard MLC for enterprise-class use cases.

Encapsulation The act of wrapping a frame from one protocol inside the frame of another. In FCoE networks, FC frames are encapsulated within Ethernet frames.

EOF End of frame (EOF) is an ordered set that appears at the end of every FC frame.

EPO The emergency power-off (EPO) switch is usually a big red button in the data center that is used to cut all power to the equipment in the room. Most data centers have more than one, and they are usually protected behind plastic covers to prevent accidental activation. Do not push this button unless you absolutely have to, as doing so might cost you your job.

ER The entrance room (ER) in a data center is the secure room that all of your external network connections come into, and it is the demarcation point between your equipment and your network provider's equipment.

ETS Enhanced transmission selection (ETS) is a DCB networking standard that attempts to standardize quality of service on CEE networks.

F

F_Port A fabric port (F_Port) is a physical port on an FC SAN switch that is connected to an N_Port.

Fabric Collection of connected FC switches that share a common set of services such as name server and zoning database.

FBA Fixed block architecture (FBA) is a type of disk drive (the only one now available) on which all sectors are preset to the same size and cannot be changed.

FC Fibre Channel (FC) is a high-speed, low-latency networking technology commonly used to connect computers with shared storage arrays over a dedicated network referred to as an FC SAN. Many high-end disk drives used to ship with FC interfaces, although these have been pretty much superseded by SAS.

FC-AL Fibre Channel arbitrated loop (FC-AL) is an older FC technology used on simple loop networks.

FC-P2P Fibre Channel point-to-point (FC-P2P) is a topology in which initiators are directly connected to targets without an intermediate switch.

FC-SW Fibre Channel switched fabric (FC-SW) is a topology in which initiators and targets are connected via a switched fabric.

FCID The 24-bit dynamically assigned address given to each node (end device) on an FC SAN fabric when the device performs a FLOGI operation. Sometimes call an N_Port ID.

FCoE Fibre Channel over Ethernet (FCoE) is the process of encapsulating FC frames within Ethernet frames. FCoE offers the potential to reduce the number of networking components required to deploy new services in a data center. FCoE networking equipment, such as CNAs and FCoE-capable switches, are required.

FEC Forward error correction (FEC) is a form of erasure code technology.

FFP Full-feature phase (FFP) is the final phase of an iSCSI session, where all previous phases have completed and data can now be sent between initiators and targets.

Fiber Fiber optic cable.

Firmware The software on a storage array or SAN switch that is the brains of the device. Firmware is the software that runs the specific storage-related hardware.

FLOGI A fabric login (FLOGI) is the process a device performs when it first logs in to a SAN fabric. As part of the FLOGI procedure, end devices are assigned their 24-bit N_Port IDs (used for routing) and are allocated their initial buffer credit stock.

FPGA A field-programmable gate array (FPGA) is somewhat similar to an ASIC but is programmable in the field rather than in the factory.

FQDN Fully qualified domain name (FQDN), also known as a DNS name.

Fragmentation Filesystem fragmentation occurs when files do not occupy contiguous space within a filesystem. Generally speaking, you should not defragment data that is on solid-state devices.

FRAM Ferroelectric RAM is a form of solid-state storage.

Front-end The front-end of a storage array is where the host-facing ports, cache memory, and main CPUs are.

FRU Field replaceable units (FRUs) are components that can be replaced in the field by customers. They are usually hot-swappable.

FSPF Fabric Shortest Path First (FSPF) is a Fibre Channel routing protocol that determines paths through the FC fabric and builds and maintains the routing table.

FSW A full-stripe write (FSW) occurs when a system can write a full stripe (RAID stripe) of data to a RAID volume. FSWs protect RAID systems from overhead associated with having to perform in-place updates.

FTP File Transfer Protocol (FTP) is an old UNIX-based protocol used for uploading and downloading files. It is still in use today but is being superseded by more-modern file-sharing protocols and services.

FUBAR Fouled up beyond all repair (FUBAR) is a term used to indicate that something is truly broken.

G

Garbage collection In the context of flash memory, garbage collection is a set of background processes that perform tasks such as pre-erasing blocks that have been marked for deletion and reorganizing blocks that are only sparsely filled. Garbage collection runs in the background and speeds up write operations to flash media as well as lengthens the life expectancy of the flash media.

GB A gigabyte (GB) is a common unit of storage capacity equal to 1,000 megabytes (MB). The GB unit is a decimal rather than binary value. The corresponding binary equivalent is the gibibyte (GiB).

GFS Grandfather-Father-Son (GFS) is the most popular and widely used backup-retention and tape-rotation scheme.

GiB A gibibyte (GiB) is a multiple of a byte, equal to 1,024 mebibytes (MiB). See also byte.

Grid storage architecture A storage array that has more than two controllers. These controllers usually work in active/active mode. Sometimes referred to as cluster architecture.

H

HA High availability (HA) is the ability for a service to be able to sustain component-level outages. HA is often associated with cluster designs.

Hardware offloads The process of offloading storage-related tasks from the host to the storage array. Sometimes referred to as hardware assists or hardware acceleration. VMware VAAI and Microsoft ODX are examples of hardware offloads.

HBA A host bus adapter (HBA) is a Fibre Channel network card.

HDA The horizontal distribution area (HDA) in a data center is where distribution-layer switches typically reside, as well as main patch panels.

HDD Hard disk drive (HDD). A term commonly used to refer to mechanical disk drives. Sometimes called hard disk or hard drive.

Hot backup A backup that occurs while the system being backed up remains online. Sometimes called an online backup.

Hot-spare A drive in a RAID group that is automatically used to rebuild data to in the event that another drive in the system fails.

HVAC An acronym used to refer to the heating, ventilation, and air conditioning systems used in a data center.

Hybrid cloud A combination of private cloud and public cloud used by an application or service.

Hybrid drive A form of storage media that combines a mechanical disk drive with solid-state memory. They come in standard drive form factors of 2.5 inches and 3.5 inches.

I

ILM Information Lifecycle Management (ILM) is a complex process of classifying data according to its business value and placing it on the most appropriate infrastructure.

Incremental backup An incremental backup backs up all data that has changed since the last full backup.

Initiator A process running on a server that initiates traffic to targets. Quite often we refer to hosts or HBAs as initiators as they often run the initiator process. Initiators are somewhat analogous to clients in the client-server computing model.

Inline Inline operation, in the context of storage, performs actions against I/O as the I/O is being processed and before it is written to disk. In the example of deduplication, inline deduplication is where the data is deduplicated in-flight, before it is written to disk.

IOPS Input/output operations per second (IOPS) is a common storage metric useful for measuring random workloads.

IQN An iSCSI qualified name (IQN) is a hierarchical name used to identify iSCSI devices on an iSCSI SAN.

iSCSI Internet SCSI is a SAN technology that allows SCSI data to be sent over an IP network. The IP network is usually dedicated to iSCSI traffic and a layer 2 LAN.

ISL Inter-switch links (ISLs) are used to connect two FC switches in the same fabric. ISLs are formed between two E_Ports on connected switches in the same fabric.

ISL trunk In terms of an FC fabric, an ISL trunk is a bonding of multiple ISLs into a single logical ISL. Some vendors refer to this as a PortChannel.

iSNS Internet Storage Name Service (iSNS) is the equivalent of the SNS on an FC SAN and is a simple database that contains the configuration of the iSCSI SAN and attached devices.

IVR Inter-VSAN routing (IVR) allows frames to be transmitted between devices on separate VSANs.

J

JBOD JBOD is an acronym for just a bunch of disks and refers to a bunch of disks grouped together with no intelligence layered on top of them. So no RAID, no snapshot capabilities, no replication, no deduplication. Literally just a bunch of disks providing IOPS and capacity. Generally used to refer to an expansion shelf of disks that is connected to a server as direct-attached storage.

L

LACP Link Aggregation Control Protocol (LACP) is a networking technology that allows multiple network connections to be bonded.

LAN A local area network (LAN) is a network that spans a local campus and does not cover a larger area such as metropolitan or global areas.

Latency Usually refers to any form of delay introduced by network equipment. Spinning disk drives suffer from mechanical latencies such as rotational latency and seek time.

LBA Logical block addressing (LBA) is a layer of abstraction used to hide the complexities of lower layers of disk drive technologies. Instead of an operating system having to understand the complex workings of every disk drive ever made, every disk drive represents its capacity as a simplified logical address space known as LBA.

LC A Lucent connector(LC) is a standard small-form-factor male connector for a pair of optical fibers. LC connectors are half the size of SC connectors.

LOM LAN on motherboard (LOM) is the process of embedding a network interface card directly onto a server motherboard.

Lossless network A network that does not drop packets when congestion arises.

Lossy network A network that drops packets when congestion arises.

LRU Least recently used (LRU) algorithms work on a queue principle, in which the queue elements that have been used most recently stay in the queue and those that have been used (referenced) for the longest drop out of the queue. LRU algorithms are common in cache memory systems.

LSAN A logical SAN (LSAN) allows frames to be transmitted between logical fabrics (in a Brocade environment). It also allows frames to be transmitted between discrete physical fabrics.

LTO Linear Tape-Open (LTO) is the most popular magnetic tape format in use.

LUN A LUN is usually equivalent to a volume on a SAN. Technically speaking, LUN is an acronym for logical unit number. All devices on a SCSI bus require a LUN to be addressable. This includes volumes presented on a SCSI bus.

LUN masking A way of selectively assigning access to LUNs to specified hosts, usually based on the WWPN of the HBA ports (or IP addresses in an iSCSI SAN) installed on the connected hosts.

LVM LVM (Logical Volume Manager) is the default volume manager in Linux distributions.

LZ Lempel-Ziv compression is a family of data compression algorithms.

M

MAID A massive array of idle disks (MAID) is a group of disks that spin down unused disks to conserve power.

MAN A metropolitan area network (MAN) is a network that spans metropolitan areas.

MB A megabyte (MB) is a common unit of storage capacity equal to 1,000 kilobytes (KB). The MB unit is a decimal rather than binary value. The corresponding binary equivalent is the mebibyte (MiB).

MBps Megabytes per second (MBps) is a common storage performance metric that is useful for measuring sequential performance (throughput).

MDA The Main Distribution Area (MDA) in a data center is where the network core typically resides, as well as other systems such as the PBX.

Memristor Memristor technology is a form of solid-state storage.

Mesh topology In a fabric that uses a mesh topology, all FC SAN switches are connected to all other FC SAN switches.

Metadata Metadata is data about data, sometimes referred to as control data. As an example, the text of this manuscript is the actual data of the data; some of the metadata associated with this manuscript might include the date the files were last saved, the size of the files, the file types, the file ownership permissions, and so on.

MiB A mebibyte (MiB) is a multiple of a byte, equal to 1,024 kibibytes (KiB). See also byte.

In the context of SNMP, MIB is an acronym for management information base, which defines what settings on a particular device can be managed and monitored.

Microsecond One millionth of a second. Expressed as either μs or simply us.

Millisecond One thousandth of a second. Expressed as ms.

MLC Multi-level cell (MLC) is a type of flash memory cell that can store 2 bits of data per cell. MLC has double the storage density of SLC but can sustain a significantly lower number of P/E cycles than SLC. This means that MLC cells cannot be written to as many times as SLC cells.

MMF Multimode fiber is the most popular form of fiber used in the data center.

MPIO Multipath I/O (MPIO) is host-based software that allows storage volumes (LUNs) to be accessed over multiple paths. This allows for increased resiliency as well as increased performance.

MTP Multi-fiber termination push-on (MTP) is a ribbon fiber cable that carries multiple fibers and terminates them at a single connector.

Multiplexing The act of sending streams from multiple backup jobs to a single tape in order to keep a tape operating at maximum speed.

Multitenancy A single system that is shared by more than one customer is said to be multitenant. Public cloud services such as AWS are hugely multitenant.

MW Megawatt. Unit of power equal to a million watts.

N

N_Port An N_Port is a node port on an FC SAN and corresponds to end-point devices on the SAN.

N_Port ID A 24-bit dynamically assigned address given to each node (end device) on an FC SAN fabric when the device performs a FLOGI operation. Sometimes call an FCID.

Nanosecond One billionth of a second. Expressed as ns.

NAS Network-attached storage (NAS) is shared storage that is accessed over an IP network.

NBD Next Business Day (NBD) is a service level often associated with hardware maintenance contracts. In an NBD contract, the third party agrees to replace failed components during the next business day. NBD contracts are usually cheap contracts and represent a low level of service.

NDMP Network Data Management Protocol (NDMP) is a standards-based protocol designed specifically for NAS backups.

NDU Nondisruptive upgrade (NDU) is a form of upgrade that can be performed while the system is online. The upgrade operation causes no disruption to the running system.

NFS Network File System (NFS) is the predominant file-sharing protocol used in Linux and UNIX environments and is commonly used for VMware deployments. Most NAS arrays implement the NFS protocol.

NL-SAS Nearline-SAS is a form of mechanical disk drive that combines a capacity-oriented head-disk assembly used in SATA drives with a SAS interface and supports the SAS protocol.

Nonpersistent storage See volatile storage.

Nonvolatile storage See persistent storage.

NPIV N_Port ID virtualization (NPIV) technology allows a single physical N_Port to register for multiple N_Port IDs.

NVRAM Nonvolatile RAM is a form of cache memory that is protected from losing its contents (volatizing). This protection usually comes in the form of a battery backup unit or a super capacitor.

O

ODX Offloaded Data Transfer (ODX) is a data copy offload technology used by Microsoft Windows servers. ODX is similar to some features of VMware VAAI.

OM Optical multi-mode (OM) is used to categorize multimode fiber cables. OM3 and OM4 are currently the most modern form of MMF cable.

OOB An out-of-band (OOB) network is physically separated from the network that sends data. Management networks are often OOB networks, as this prevents management traffic from interfering with production network traffic.

OSD Object storage devices (OSDs) are storage systems designed to work with objects rather than files or blocks. Sometimes referred to as object stores.

Overprovisioning The act of provisioning more storage from a thinly provisioned device, usually a storage array, than the device actually has. For example, provisioning 200 TB of storage from a storage array that has only 150 TB of physically installed usable capacity. In the context of SSD, overprovisioning is the application of more physical flash to an SSD than the user actually gets to see. This extra space is used by the SSD (actually the controller) to extend the life of the device and improve performance via things such as garbage collection and wear leveling.

Oversubscription Occurs when multiple devices use a single switch port and the combined potential throughput of all devices exceeds the potential throughput of the switch port. For example, four hosts connected at 8 Gbps using a single 16 Gbps ISL.

P

Parity A technique used by RAID algorithms to increase resiliency and enable RAID groups to recover from failed drives.

PB A petabyte (PB) is a common unit of storage capacity equal to 1,000 terabytes (TB). The PB unit is a decimal rather than binary value. The corresponding binary equivalent is the pebibyte (PiB).

PCM Phase-change memory (PCM) is a form of solid-state storage.

PDU In the context of data centers and mechanical and electrical systems, PDU is an acronym for power distribution unit.

In the context of iSCSI SAN technology, PDU is an acronym for protocol data unit.

Persistent storage Storage that does not lose its contents when the power is removed. Examples of persistent storage include disk drives, tapes, and flash memory. Sometimes referred to as nonvolatile storage.

PFC Priority-based flow control (PFC) allows a PAUSE condition to be selectively enforced on any of the eight lanes defined on a CEE network.

PiB A pebibyte (PiB) is a multiple of a byte, equal to 1,024 tebibytes (TiB). See also byte.

PIT copy A point-in-time copy is a copy of an entire volume taken at a particular point in time, such as 5 p.m. PIT copies are often snapshots.

Platter The flat, circular recording surfaces of a disk.

PLOGI Port login (PLOGI) is the process of an end device on a SAN registering its capabilities with the SNS. End devices also perform PLOGI operations with devices they will be communicating with.

pNFS Parallel NFS is an implementation of NFS that was introduced with NFS version 4.1.

Portal In an iSCSI SAN, a portal is effectively a network interface port on the iSCSI target device (combination of IP address and TCP port).

Portal group In an iSCSI SAN, a portal group is a collection of portals that reside within a single iSCSI target device. Somewhat similar to a NIC team.

PortChannel In terms of an FC fabric, a PortChannel is a bonding of multiple ISLs into a single logical ISL. Some vendors refer to this as an ISL trunk.

Post-processing The act of performing an action on a data set after the data set has been written. For example, post-process deduplication saves the data to disk in the first place and then at a later date will run a deduplication operation against the data.

Primary storage Storage that is used to store data that is in active use.

Principal switch Every FC fabric requires a single principal switch. The principal switch is responsible for many things within the fabric, one of which is the allocation of domain IDs to other switches in the fabric.

Private cloud Normally services hosted by a third party, off premise, but offering customers their own dedicated infrastructure without multitenancy. It is also possible to build your own on-premise private cloud, using technologies such as OpenStack.

Public cloud Services hosted by a third party, off premise, on a large shared multitenant platform. Services in the public cloud are accessed over the Internet.

PUE Power usage effectiveness (PUE) is an industry-accepted way to measure the power efficiency of a data center. It is a simple calculation of how much of the total power supplied to the data center is actually used by IT equipment, rather than things like HVAC and other M&E systems.

Purge In the context of backups, the act of removing backup jobs from tape (or other media) after the retention period for the backup expires. Archives and other types of data store can also be purged to free up space or prevent data from being accessed in the future.

R

R_RDY A Receiver Ready (R_RDY) is a special control frame used on FC SAN fabrics to indicate to a sender that the receiver is ready to receive more frames. In B2B flow-control scenarios (COS 3), R_RDY frames are used to replenish a sender's stock of buffer credits.

RAID Redundant array of independent drives (RAID) is a vital technology in the storage industry used to increase availability and performance of storage systems. In the past, RAID was referred to as redundant array of inexpensive disks. However, because these drives are no longer always inexpensive and because solid-state technologies are technically drives and not disks, the term is now more accurately referred to as redundant array of independent drives.

RAID 0 Striping without parity.

RAID 1 Mirroring.

RAID 5 Block-level striping with distributed parity.

RAID 6 Block-level striping with two sets of distributed parity.

RAID 10 Hybrid of RAID 1 and RAID 0, mirroring and striping. The mirror sets are created first, and then a layer of striping is added across the mirrors.

RAID group A group of physical drives that are combined and configured to work together as a RAID group. RAID groups can be configured with any RAID level so long as there is a sufficient number of drives in the RAID group. Also known as RAID set or RAID array.

RAIN Redundant/reliable array of independent nodes (RAIN) is a form of network-level RAID in which redundancy and fault-tolerance are achieved at the node level. Common in scale-out architectures, where each node is loosely coupled to the rest of the cluster and the system can continue to function in the event of a node failure.

RAM Random access memory (RAM) is a form of silicon-based volatile storage.

Read/write head Each platter surface of a disk drive has a read/write head (R/W head) that flies above the surface of the platter and reads and writes data to the platter.

Rebuild The act of rebuilding a RAID group to full redundancy after a drive has failed. Data is usually rebuilt to a spare drive.

Redundancy The practice of having multiple components that do the same job so that in the event of component-level failure, the entire system does not fail. Instead of the entire system failing, the spare component takes over.

ReFS Resilient Filesystem (ReFS) is a next-generation Microsoft filesystem.

Replication Making a remote copy of a volume that can be used if access to the source volume fails. Replication can be synchronous or asynchronous. Replication can also occur at various levels in the stack, including at the storage array or at several layers in the host.

Response time See latency.

RHEL Red Hat Enterprise Linux.

Ring topology In an FC SAN, a ring topology is where all switches are connected in a simple ring/loop.

Rotational latency A form of mechanical latency (delay). The time taken for the correct sector to arrive below the read/write heads after the heads have been positioned over the correct track. The faster a drive spins (RPM), the lower its rotational latency. Solid-state storage does not suffer from rotational latency.

RoW Redirect-on-write (RoW) snapshots work on the principle that the original blocks composing a volume are not changed, and when changes are made to the original contents, the new data is written to new blocks elsewhere. The original blocks are marked as being part of the snapshot.

RPM Revolutions per minute (RPM) is the unit of measure describing how fast a disk drive spins. The most common RPM values for modern disk drives are 15,000 RPM, 10,000 RPM, 7,200 RPM, and 5,400 RPM.

RPO Recovery point objective (RPO) refers to the point in time at which an application or service can be returned to in the event of a disaster or failure.

RSCN Registered state change notification (RSCN) is an FC SAN technology used to inform devices on the SAN of changes in the fabric that affect them.

RTO Recovery time objective (RTO) refers to the amount of time it will take to recover an application or a service after a disaster occurs.

RTT Round-trip time (RTT) is the time taken (usually in milliseconds) for an I/O or packet of data to be sent from the source and returned by the target.

RU An industry standard rack unit (RU) is 1.75 inches. Sometimes shortened to U.

Rx Short for receive.

S

S3 Amazon's Simple Storage Solution (S3) is the largest public cloud object store on the planet. Part of AWS.

SAN A storage area network (SAN) is a network dedicated to block-level storage traffic. The major types of SAN are iSCSI and FC. And iSCSI SAN is based on IP technology, and an FC SAN is based on Fibre Channel technology.

SAS Serial Attached SCSI (SAS) is a serial point-to-point implementation of the SCSI protocol. Many modern disk drives (spinning and solid state) support the SAS protocol and ship with physical SAS interfaces. SAS is considered higher performance and better than SATA.

SATA Serial Advanced Technology Attachment (SATA) is a serial implementation of the ATA protocol. Many modern disk drives (spinning and solid state) support the SATA protocol and ship with physical SATA interfaces. SATA is commonly associated with low-end disk drives and applications.

SC A standard connector (SC) is a standard male connector for a pair of optical fibers that was commonly used in the days of 1 Gbps FC. SC connectors are twice the size of LC connectors and rarely used nowadays.

Scale-out Scaling out is the process of growing a system by adding more computing, network, and storage capacity.

SCSI Small Computer System Interface. SCSI is a set of standards for physical connections, as well as a protocol and command set for transferring data.

SDN Software-defined networking is essentially the separation of the control plane and the data plane, allowing the control plane to be moved into software outside the switch hardware.

SDS Software-defined storage is the decoupling of the intelligence of a storage system from the underlying hardware. Traditional storage arrays have been assembled in a factory where the firmware was designed to run on specific hardware, and you had to buy the whole package—software and hardware. In an SDS architecture, you can run the storage controller (the intelligence) on top of any hardware of your choice. Common implementations of SDS implement the software controller as a VM on top of a hypervisor such as ESXi.

Sector The smallest addressable unit of a disk drive, usually 512 bytes, 520 bytes, or 4 KB. Each track on a disk drive is divided into multiple sectors.

Seek time In disk drive technology, seek time is the time it takes to position the read/write heads over the correct position of a disk drive's platter. Seek time is a form of mechanical latency and has a significant impact on random workloads. There is no concept of seek time in solid-state storage technologies.

SendTargets A technology used to identify devices on small to medium-sized iSCSI SANs.

Serverless backup A form of LAN-free backup that uses the SCSI EXTENDED COPY command to copy blocks directly from a source to the target LUN without passing through the server.

SFTP A secure version of the FTP protocol.

SID Source ID (S_ID) is the N_Port ID of the sender of a frame and is identified in every FC frame header.

SIS Single-instance storage (SIS) is deduplication at the file level—meaning that if there are duplicates of an entire file, they can be deduplicated. SIS is not a true form of deduplication.

SLA A service-level agreement (SLA) is a formal agreement between two parties that outlines certain levels of service that will be provided. Examples include the amount of uptime a service will have, and the RPO and RTO of a particular service.

SLC Single-level cell (SLC) is a type of flash memory cell that can store a single bit of data per cell. SLC is high performance but low density. Other technologies such as MLC and TLC store more data per cell.

SLDC Streaming Lossless Data Compression (SLDC) is the compression technology native to LTO backup takes.

SLP Service Location Protocol (SLP) is a technology intended for automatic device discovery on medium to large iSCSI SANs.

SMART Self-Monitoring, Analysis, and Reporting Technology (SMART) is a form of drive-monitoring technology that is used to track errors on drives and predict failures.

SMB Server Message Block is the de facto file-sharing protocol in use in Microsoft Windows–based networks. Most NAS arrays implement SMB. SMB is sometimes referred to as CIFS, because earlier implementations were called CIFS.

SMF Single-mode fiber (SMF) is used in conjunction with longwave lasers (optics) for long cable runs such as those required by telecommunications companies.

SMI-S Storage Management Initiative Specification (SMI-S) is a SNIA-led initiative aimed at providing a common web-based standard for managing heterogeneous storage devices.

Snapshot A snapshot is a point-in-time space-efficient local copy of a volume.

SNIA The Storage Networking Industry Association (SNIA) is a group of storage vendors and customers that aims to advance storage technologies, standards, and education.

SNMP Simple Network Management Protocol (SNMP) is an IP-based protocol for monitoring and managing network-attached devices.

SNS The Simple Name Server (SNS), sometimes called simply the name server, is where all devices on a SAN register themselves and their capabilities. The SNS can also be queried for information about other devices on the SAN.

SOF The start of frame (SOF) delimiter is the first 4 bytes of all FC frames and specifies things such as class of service.

Spindle Each mechanical disk drive has a single spindle and spindle motor at the center of every platter. It is the spindle and motor that are responsible for spinning the disk platters. Sometimes people use the term spindle to refer to the disk drive itself. For example, a storage array with 100 disk drives might be said to have 100 spindles.

SPOF A single point of failure (SPOF) is anything in a design that, if it fails, will bring the entire system down.

SRA A Storage Replication Adapter (SRA) is a VMware SRM plug-in that allows vCenter to integrate with the replication features of a storage array.

SRM In the context of VMware, SRM is an acronym for Site Recovery Manager.

In the context of storage management applications, SRM is an acronym for Storage Resource Management tools.

SSC A solid-state card (SSC) is a form of solid-state storage that comes in a PCI card form factor. As such, they can plug into most PCIe expansion slots and provide extremely fast storage (faster than SSD).

SSD The solid-state drive (SSD) is a form of solid-state storage (SSS) that is packaged in a standard disk drive form factor such as the common 2.5-inch and older 3.5-inch disk drive form factors. It is usually plug-compatible with traditional spinning disk drives so that it can easily be used instead of a traditional spinning disk drive.

SSS Solid-state storage (SSS) refers to all forms of semiconductor-based, nonvolatile storage devices. A common example is flash memory.

STP Spanning Tree Protocol (STP) is an Ethernet technology that prunes paths in an Ethernet network that could cause loops in the network to appear.

In the context of network cabling, STP is an acronym for shielded twisted-pair copper cabling.

STR Sustained transfer rate (STR) refers to how fast a disk drive can read or write data over multiple tracks. STR is lower than maximum transfer rate (MTR), which often refers to burst rates.

Striping The process of laying out a volume over multiple spindles so that the volume has access to the performance of multiple spindles. RAID 0 is a form of striping.

Synchronous replication A replication technology that offers zero data loss but requires good network links for replication and can add latency to the overall solution.

Synthetic full backup The process of creating a full backup image by combining multiple incremental backups.

T

Tape Tape is a plastic film with magnetic medium to provide a sequentially accessed storage medium. Although tape was used as primary storage in the Jurassic period of computing, it's now used as a backup or archival storage medium.

Target A process, usually on a disk array, that accepts I/O and commands from an initiator. We often refer to disk arrays as targets, as they run the target processes. Targets are somewhat analogous to servers in the client-server computing model.

TB A terabyte (TB) is a common unit of storage capacity equal to about 1,000,000,000,000 (one trillion) bytes. This is usually easier to understand as 1,000 gigabytes (GB). The TB unit is a decimal rather than binary value. The corresponding binary equivalent is the tebibyte (TiB).

Thick volume A non-thinly provisioned volume. A volume that consumes all of its allocated space as soon as it is created. Sometimes referred to as a traditional volume.

TiB A tebibyte (TiB) is a multiple of a byte, equal to 1,024 gibibytes (GiB). See also byte.

Tiering The process of moving data (volume extents) to the most appropriate tier of storage. Usually, frequently accessed data (extents) is placed on high-performance storage.

TLC Triple-level cell (TLC) is a type of flash memory cell that can store 3 bits of data per flash cell. TLC is relatively low performance but very high-density flash. Of the three common types of flash memory (SLC, MLC, and TLC), TLC supports the lowest number of P/E cycles, meaning that TLC flash cells can be written to far less than MLC and SLC.

TOE A network interface card that has a processor that performs TCP-related functions is said to have a TCP Offload Engine (TOE).

ToR A Top of Rack (ToR) switch is a LAN or SAN switch located at the top of a data center rack that acts as the access-layer device.

TP Thin provisioning. The act of allocating space to a volume on demand (as areas of the volume are written to).

TP stun Thin provisioning stun (TP stun) is an attempt to gracefully deal with out-of-space conditions on TP volumes.

Track A concentric ring running all the way around the surface of a disk drive platter.

Tx Short for transmit.

U

U An industry-standard rack unit that is 1.75 inches.

Unified storage A storage array that supports both block and file protocols (SAN and NAS). Sometimes referred to as a unified array.

UPS An uninterruptable power supply (UPS) is typically a battery-based device that supplies power to a connected system, for a relatively short period of time, in the event that utility power is lost. Most good UPS systems are constantly online and provide instantaneous power to systems so that the power loss is not noticeable. The relatively short periods of time that UPS systems can provide power to equipment is usually enough to allow you to gracefully shut down the system or wait for the on-site generator to start.

UTP Unshielded twisted-pair (UTP) cabling is a popular form of copper cabling used in the data center that consists of a pair of copper wires twisted around each other to reduce interference.

V

VAAI VMware's vStorage API for Array Integration (VAAI) is a set of technologies designed at offloading storage-related functions from the hypervisor to the storage array.

VHD Virtual hard drive (VHD) is the block-based virtual hard drive form used by Microsoft Hyper-V implementations.

VMDK A VMDK file is a VMware virtual machine disk file.

VMFS VMFS is the Virtual Machine Filesystem used in most versions of VMware.

Volatile storage Storage that loses its contents when the power is removed. An example of volatile storage is random access memory (RAM).

VSA A virtual storage appliance (VSA) is a storage controller that runs as a virtual machine on top of a hypervisor (usually VMware ESXi). A VSA is a form of software-defined storage.

VSAN In the context of an FC SAN, a virtual SAN (VSAN) is a standards-based technology designed and implemented by Cisco that allows physical switches to be partitioned into virtual switches and virtual fabrics.

In the context of VMware, VSAN is a software-defined storage system that runs in the vSphere hypervisor.

VSS Microsoft Volume Shadow Copy Service (VSS) is a framework that allows for application-consistent copies (snapshots and so on) of volumes to be taken.

VTL A virtual tape library (VTL) is a disk-based storage system that emulates a tape library.

W

WAFL WAFL is NetApp's proprietary Write Anywhere File Layout filesystem that works on a redirect-on-write basis.

WAN A wide area network (WAN) is a network that spans extremely large areas such as cross-continent.

Wear leveling Processes to ensure even distribution of write workload (P/E cycles) across all flash cells in a flash memory device. There are two types of wear leveling: background wear leveling and in-flight wear leveling.

Wide-striping The process of laying out a volume's address space (storage) over a large number of physical drives.

WORM Write once, read many (WORM) technologies allow data to be written once and then never overwritten again. However, they allow data to be read as often as necessary. WORM technologies are useful as compliance archives where data must be securely written to the archive and then be tamper-proof.

Write amplification Refers to a situation where the number of writes to physical media is more than the number of writes issued by the host. Write amplification is especially important in the world of flash memory, where each flash cell has a limited number of times that it can be erased and reprogrammed. Reducing the amount of write amplification to flash memory is extremely important from both a performance and a flash durability perspective.

Write coalescing The process of holding write data in cache until enough write data exists to perform an optimized write to the persistent storage media. Write coalescing helps avoid things like write amplification.

Write endurance Write endurance refers to the number of program/erase (P/E) cycles a flash memory cell is rated to be able to sustain by the manufacturer.

WRITE_SAME WRITE_SAME is a T10 SCSI standard that allows more-efficient writing of long identical streams of data. A common implementation is in hardware-accelerated zeroing operations, which write long streams of zero data to volumes and areas of volumes.

WWPN A world wide port name (WWPN) is 64-bit worldwide unique address given to every HBA port in a SAN. A WWPN is heavily used in SAN environments for security-related tasks such as zoning and LUN masking.

X

XOR An eXclusive OR (XOR) operation is used when calculating parity in RAID arrays.

Z

ZDR Zoned data recording (ZDR) is a common technique in hard disk technology that sees the outer tracks of a platter have more sectors than the inner tracks. Each platter surface is divided into multiple zones, with all tracks in the same zone having the same number of sectors.

Zero space reclamation Identifying deleted data (zeros) and releasing it from a TP volume so it can be used for the growth of other volumes.

Zone set Multiple zones' definitions are grouped together into zone sets. It is zone sets that are applied to the fabric.

Zoning Zoning is the act of partitioning a SAN into smaller, more manageable units. When zoning is in place on a SAN, only devices that exist in the same zone can communicate.