Huawei HCIA-Storage key terms

A technique that combines several physical disks into one logical unit to improve performance, capacity, or fault tolerance. Common levels such as RAID 5 and RAID 6 use parity so data survives a disk failure.

A virtualized RAID architecture that carves physical disks into small chunks and pools them, then builds logical volumes from those chunks. Rebuilds are faster and affect only the disks holding the failed disk's chunks.

An idle disk kept in reserve inside a storage system. When an active disk fails, the hot spare is automatically substituted so the array can rebuild without waiting for a human to insert a replacement.

The process of reconstructing lost data onto a replacement disk using parity or mirrored copies. Rebuild time depends on disk size, array type, and load; a long rebuild leaves the array degraded and more vulnerable.

The state of a redundant array after one member disk has failed but the data is still readable from parity or mirrors. Performance and protection are reduced until the rebuild finishes.

A collection of physical disks whose capacity is aggregated and from which logical volumes or file systems are provisioned. Thin provisioning and quality-of-service policies are typically applied at this layer.

A logical unit number identifies a block-storage volume presented to a host over SAN protocols. The host sees it as a raw disk it can partition and format with its own file system.

Allocating logical capacity to volumes on demand rather than reserving it upfront. Space is consumed only as data is written, which raises utilization but requires monitoring so the pool does not run out.

A point-in-time copy of a volume or file system that usually shares unchanged data blocks with the source. It enables quick rollback after corruption or a failed upgrade without duplicating the full data set.

A fully independent copy of a source volume or file system, often created instantly from a snapshot and then separated. Unlike a snapshot, a clone can be modified without affecting the original.

Copying data continuously or on a schedule from one storage system to another for disaster recovery or distribution. Synchronous replication keeps sites identical but adds latency; asynchronous replication tolerates distance.

A protocol that carries SCSI block commands over TCP/IP, letting hosts use standard Ethernet to reach SAN storage. It is cheaper to deploy than Fibre Channel but depends on a well-designed network.

A dedicated high-speed network technology for block storage, using HBAs, switches, and worldwide names for addressing. It offers predictable latency and is common in enterprise SAN environments.

Presenting several physical paths between a host and a storage volume and load-balancing or failing over between them. It prevents a single failed cable or switch port from cutting off access to data.

Network-attached storage exposes shared file systems over protocols such as NFS and SMB/CIFS. Clients access files by name, making NAS well suited to shared home directories and unstructured data.

A storage model that keeps data as objects in flat buckets, each with metadata and a unique key, accessed through HTTP APIs. It scales horizontally and suits backups, archives, and cloud-native applications.

A data protection method that splits data into fragments and adds parity fragments so the original can be rebuilt from a subset. It uses less capacity than full mirroring, especially at large scale.

Solid-state storage uses flash memory with no moving parts, giving much lower latency and higher random-read performance than spinning disks. Write endurance and garbage collection are key operational considerations.

A hard disk drive stores data on rotating magnetic platters read by moving heads. It offers low cost per gigabyte for capacity tiers but far lower random IOPS than flash media.

Input/output operations per second measures how many read or write requests a device or system completes in a second. It is the usual metric for transactional workloads, while throughput suits large sequential transfers.

The time a single I/O request takes to complete, usually measured in milliseconds or microseconds. Array cache, network hops, and disk media all contribute to the end-to-end figure an application experiences.

Fast memory in a storage controller that buffers writes and retains recently read data. A battery or capacitor protects cached writes so they are not lost during a power failure.

A caching policy that acknowledges a write once it reaches controller cache, flushing to disk later. It improves write latency but requires protected cache; write-through is safer but slower.

The processor module in a storage system that runs the operating software, manages RAID and caching, and serves host I/O. Enterprise arrays use paired controllers so one can take over if the other fails.

A configuration where two nodes or controllers serve workload simultaneously and can take over each other's work on failure. It uses hardware better than active-passive but needs consistent shared state.

Designing a system so that no single component failure causes an outage, through redundancy, failover, and health monitoring. The goal is expressed as availability targets such as four or five nines.

Recovery point objective is the maximum acceptable amount of data, measured in time, that a recovery may lose. It drives how frequently replication or backups must run.

Recovery time objective is the maximum acceptable time to restore a service after a disruption. It determines the architecture of failover, standby systems, and restore procedures.

A copy of data kept separately so it can be restored after deletion, corruption, or disaster. Strategies include full, incremental, and differential schedules, plus the rule of keeping copies offline or immutable.

Running multiple isolated virtual machines or containers on shared physical hardware through a hypervisor. It improves utilization and enables live migration, snapshots, and rapid provisioning.

The software layer that creates and runs virtual machines, allocating CPU, memory, storage, and network to each. Type 1 runs directly on hardware; Type 2 runs on a host operating system.

A software-defined computer with virtual CPU, memory, disks, and network interfaces, isolated from other guests. Its disk images typically live on shared storage so it can be migrated between hosts.

Moving a running virtual machine from one host to another with little or no downtime, usually while its disk stays on shared storage. It supports maintenance and load balancing without service interruption.

Delivering computing resources such as servers, storage, and applications over a network on demand, with pay-per-use or subscription billing. Service models include IaaS, PaaS, and SaaS.

Infrastructure as a service provides virtual machines, networks, and block or object storage that customers configure themselves. The provider manages hardware and virtualization; the customer manages the guest systems.

Storage capacity delivered as a service and accessed over a network, typically billed by consumed capacity and operations. Tiers range from hot, low-latency classes to archive classes with retrieval delays.

Shrinking the physical capacity needed for data through deduplication, compression, and sometimes pattern detection. Savings vary by workload; deduplication works best on repetitive data such as backups and virtual desktops.

Placing data automatically on different media classes, such as flash for hot data and disk for cold data, based on access patterns. It balances performance against cost per gigabyte.