010-151 DCTECH key terms

A vertical measurement of 1.75 inches (44.45 mm) used to describe equipment height in a standard 19-inch rack. A 2U server occupies two rack units of vertical space.

A standardized metal frame with 19-inch (482.6 mm) horizontal mounting rails used to hold servers, switches, and storage. Equipment mounts via front flanges at fixed hole spacing.

An access-layer switch placed at the top of each rack that aggregates server connections in that rack and uplinks to aggregation or spine switches. It shortens cabling runs and simplifies moves and changes.

A switch located in a dedicated row-end cabinet that serves many racks in the row. It centralizes switching but requires longer horizontal cable runs from each server rack.

A standards-based hierarchical cabling design using defined subsystems such as horizontal cabling, backbone cabling, and cross-connects. It separates cabling from equipment so changes are predictable and documented.

A passive panel of ports that terminates permanent horizontal cabling and provides a connection point for patch cords. It protects fixed cabling from wear and simplifies re-patching.

A short, flexible cable with connectors on both ends used to link a patch panel port to a switch or server port. Its length and category must match the link requirements.

Twisted-pair copper cabling rated for higher bandwidth and reduced alien crosstalk, commonly supporting 10GBASE-T up to 100 meters. It is thicker and less flexible than Category 6.

A shielded copper cable carrying a single differential pair, commonly used for short 10G, 25G, or 40G direct-attach links. Assemblies are sold as DAC cables with fixed transceivers.

A factory-terminated twinax cable assembly with transceiver ends, used for short in-rack or adjacent-rack links. Passive DACs suit short runs; active DACs extend distance and reduce host power draw.

Cabling that carries data as light pulses through glass or plastic strands, immune to electromagnetic interference. Multimode suits building runs; single-mode supports long distances and high speeds.

A small form-factor fiber connector with a 1.25 mm ferrule, often used in duplex pairs on transceivers and patch panels. It is denser than SC and common in data center fiber links.

An articulated arm mounted at the rear of a rack-mounted server that routes and supports power and data cables while the server slides in and out. It prevents strain and cable pinching.

A shared enclosure that houses multiple hot-swappable server blades and provides common power, cooling, and management. Blades share midplane connectivity to network and storage modules.

A passive circuit board inside a blade chassis that connects blades to shared power, cooling, and I/O modules. It carries signals and power but typically no active logic.

A part that can be removed or installed while the system remains powered and operating, such as a drive, fan, or power supply. It requires correct redundancy and support.

Two or more power supplies that share load or provide failover so the server stays powered if one fails. Each supply should connect to a separate power feed or PDU.

A technique combining multiple physical drives into a logical unit for performance, capacity, or fault tolerance. Levels such as RAID 1, 5, and 10 differ in parity, mirroring, and rebuild behavior.

A dedicated high-speed network that gives servers block-level access to shared storage arrays. It is typically built with Fibre Channel or iSCSI and isolated from the general LAN.

A high-speed serial protocol and network used mainly for block storage traffic in SANs. It uses dedicated adapters, switches, and zoning rather than standard Ethernet addressing.

A protocol that carries SCSI storage commands over TCP/IP networks. It lets servers use Ethernet infrastructure for block storage but requires careful network isolation and tuning.

A unique 64-bit identifier assigned to Fibre Channel devices such as host bus adapters and storage ports. Zoning and LUN masking use WWNs to control access.

An adapter that connects a server to a network and handles packet transmission and reception. Modern servers often use multi-port or converged adapters for data and management traffic.

A single adapter that carries both Ethernet and storage traffic, such as FCoE, over one physical link. It reduces cabling and adapter count but adds configuration complexity.

The interconnected set of access and aggregation switches that carries traffic between racks and to the core. Redundant paths and loop prevention are essential design requirements.

A logically segmented network within a physical switch infrastructure that isolates broadcast domains. VLAN tagging lets one link carry traffic for multiple networks.

Software that provides Layer 2 switching between virtual machines and physical uplinks on a hypervisor host. It applies policies for VLANs, security, and traffic shaping.

Software that creates and runs virtual machines by abstracting physical CPU, memory, storage, and network resources. It manages VM scheduling, isolation, and virtual device emulation.

An isolated software-defined computer with its own virtual hardware that runs on a hypervisor. Multiple VMs share a physical host while appearing as separate systems to the network.

A Layer 2 protocol that prevents bridging loops by blocking redundant paths and electing a root bridge. It reconverges after topology changes but can cause temporary outages.

Combining multiple physical links into one logical link to increase bandwidth and provide redundancy. Member links must match in speed and configuration, and the switch must support the standard.

A separate management path, such as a dedicated Ethernet port or serial console, used to access servers and switches when production networks are down. It supports remote recovery and monitoring.

A small independent controller on a server motherboard that provides remote power control, console access, and hardware monitoring. It operates even when the main system is powered off.

A standardized set of commands for monitoring and controlling server hardware through a baseboard management controller. It supports remote power, sensor readings, and event logging.

A device that distributes electrical power to rack equipment, often with metering or per-outlet switching. Redundant PDUs on separate feeds support fault-tolerant server power.

A battery-backed power system that conditions incoming power and supplies temporary runtime during outages. It bridges short interruptions and supports orderly shutdown of equipment.

A device that lets one keyboard, monitor, and mouse control multiple servers. In data centers, KVM-over-IP extends this access remotely through the network.

A layout that separates server intake and exhaust air using aisles and physical barriers to prevent mixing. It improves cooling efficiency and allows higher rack power density.

The ratio of total facility power to the power delivered to IT equipment, used to measure data center efficiency. A value closer to 1.0 indicates less overhead consumption.

The amount of power, heat, and equipment concentrated in a rack, often expressed in kilowatts per rack. It drives cooling, power distribution, and cabling design decisions.