HCIE-Storage key terms

A RAID level that stripes data and dual parity across member disks, so the array survives the failure of any two drives. Write performance is lower than RAID 5 because two parity blocks must be computed per stripe.

A RAID level that stripes data with single distributed parity, tolerating one disk failure. Read performance is strong, but small random writes incur a read-modify-write penalty and rebuilds stress surviving disks.

A nested configuration that mirrors pairs of disks and then stripes across the mirrors. It offers fast rebuilds and high random write throughput at the cost of 50 percent usable capacity.

The extra backend I/O generated when a host write must update parity. For RAID 5 a single logical write costs two reads and two writes; RAID 6 costs more, which is why parity RAID suits sequential workloads.

A standby disk kept powered and ready inside an array. When a member disk fails, the controller automatically rebuilds redundancy onto the spare, shortening the window of degraded protection.

The process of reconstructing lost data onto a replacement disk from parity or mirror copies. Rebuild time depends on disk size, load, and array type, and the array stays vulnerable until it completes.

The operating state of a redundant array after a member disk fails but before the rebuild finishes. Performance and protection are reduced, and a second failure may cause data loss.

A capacity model where volumes report a large logical size but consume physical space only as data is written. It improves utilization but requires monitoring and policies to avoid running out of real capacity.

A capacity model where the full logical size of a volume is reserved from the storage pool at creation time. Space is guaranteed but cannot be shared with other volumes, lowering overall utilization.

Deliberately leaving raw capacity unallocated, or presenting more logical capacity than physical space, to absorb growth and support snapshots, clones, and garbage collection headroom.

A logical aggregation of physical disks from which volumes or LUNs are carved. The pool abstracts RAID and disk placement so capacity can be flexibly allocated and expanded.

A logical unit number identifying an addressable block-storage device presented to a host over SAN protocols. A LUN maps to a volume or slice of a storage pool.

A protocol that carries SCSI commands over TCP/IP, letting hosts access block storage across standard Ethernet networks. It is common in cost-sensitive SANs and depends on careful network design.

A dedicated high-speed network standard for block storage, using HBAs, switches, and WWNs. It provides lossless, low-latency transport and is widely deployed in enterprise SANs.

World Wide Name, a unique 64-bit identifier burned into a Fibre Channel port or node. Zoning, masking, and multipath configuration all reference WWNs to control which hosts see which LUNs.

A Fibre Channel switch configuration that restricts which ports may communicate, isolating traffic and preventing unintended host access. Effective zoning is a prerequisite for secure LUN presentation.

A storage-array control that maps a LUN only to specified host initiators, so unauthorized hosts cannot see or mount it. It complements fabric zoning as a second access-control layer.

Host software that presents multiple physical paths to one LUN as a single device, providing failover and load balancing. It prevents duplicate devices and keeps I/O flowing when a path fails.

Asymmetric Logical Unit Access, a SCSI feature where a LUN reports preferred and non-preferred paths through array controllers. Multipath software uses these states to send I/O to the owning controller.

A file-sharing protocol common in UNIX and Linux environments, where a server exports directories and clients mount them. NFSv3 is stateless; NFSv4 adds locking, ACLs, and stronger security.

The native file-sharing protocol of Windows, also supported on other platforms. Modern SMB versions add signing, encryption, continuous availability, and multichannel, making them suitable for virtualization datastores.

Network-attached storage: a dedicated appliance or gateway that serves files over NFS, SMB, or similar protocols. Clients access shared directories rather than raw blocks, simplifying sharing across many hosts.

A storage model that manages data as objects in flat namespaces, each with a key, metadata, and payload, accessed via HTTP APIs. It scales horizontally and suits unstructured data and cloud workloads.

Amazon Simple Storage Service, whose API has become the de facto standard for object storage. Compatible platforms implement buckets, objects, versioning, and lifecycle policies for hybrid and private clouds.

A point-in-time logical image of a volume or file system, usually created copy-on-write or redirect-on-write. Snapshots enable fast rollback and backup with little initial space consumption.

A writable, independent copy of a volume or file system created from a snapshot. Space-efficient clones share unchanged blocks until either side is modified, making them fast for test and dev.

A technique where data blocks are shared until a write occurs, at which point the original block is copied elsewhere before modification. It underpins snapshots and space-efficient clones.

A set of volumes snapshotted or replicated together at the same point in time. It preserves write-order dependencies across related volumes, such as the data, log, and redo disks of one database.

Asynchronous or synchronous copying of volume data to a secondary array at another site. It supports disaster recovery by maintaining a recoverable replica with a defined RPO.

Recovery Point Objective: the maximum acceptable amount of data loss measured in time. It drives replication frequency and determines whether synchronous or asynchronous methods are required.

Recovery Time Objective: the target duration within which a service must be restored after a disruption. It shapes failover design, automation, and the choice of active-active versus active-passive configurations.

A redundancy scheme that splits data into fragments and computes parity fragments, so the object survives the loss of several fragments. It uses capacity more efficiently than replication at scale.

Fast memory or SSD space that absorbs reads and buffers writes to reduce latency to slower media. Cache hit ratio and write-back policy strongly influence measured array performance.

Automatically placing data across media of different performance and cost based on access patterns. Hot data migrates to SSDs while cold data moves to high-capacity disks, balancing cost and latency.

Quality of Service controls that cap or guarantee IOPS, bandwidth, or latency for specific volumes, hosts, or tenants. They prevent noisy-neighbor effects in shared and multi-tenant storage environments.

Eliminating duplicate data blocks or chunks by storing a single copy and referencing it multiple times. It is most effective on backup and virtual-desktop workloads and requires adequate compute and memory.

Encoding data so it occupies fewer physical blocks, reducing capacity consumption and sometimes improving read throughput. Inline compression acts before data lands on media; post-process compression runs later.

Solid-state drive, flash-based media with no moving parts, offering low latency and high random IOPS. Write endurance, garbage collection, and wear leveling are key operational considerations.