Huawei HCIA-Storage cheat sheet

RAID 5 tolerates one drive failure; usable capacity = (N-1) × smallest drive capacity; minimum 3 drives.

RAID 6 tolerates two simultaneous drive failures; usable capacity = (N-2) × smallest drive capacity; minimum 4 drives.

RAID 10 (1+0) combines mirroring and striping: usable capacity = 50% of raw; tolerates one failure per mirrored pair; minimum 4 drives.

RAID 0 provides striping with no redundancy; usable capacity = 100% of raw; a single drive failure loses the entire array.

Fibre Channel (FC) SAN uses lossless, in-order delivery with buffer-to-buffer credits; iSCSI encapsulates SCSI over TCP/IP and is lossy, relying on TCP retransmission.

SAS and NL-SAS drives are dual-ported for HA; SATA drives are single-ported and unsuitable for multi-initiator HA storage.

SSD wear leveling and garbage collection cause write amplification; over-provisioning (typically 7–28%) preserves endurance and steady-state performance.

Erasure coding with k data and m parity fragments tolerates m fragment losses; storage overhead = (k+m)/k; reconstruction requires any k fragments.

Storage tiering places hot data on SSD and cold data on HDD; tiering granularity is typically a sub-LUN extent or chunk rather than a whole LUN.

A storage controller (engine) contains CPU, cache, and host/disk interfaces; dual-controller configurations provide active-active HA with transparent failover.

Write-back cache acknowledges writes before they reach disk; protected by a battery or supercapacitor (BBU) to prevent data loss on power failure.

Write-through cache writes to disk before acknowledging; lower risk but higher write latency than write-back.

Cache mirroring between two controllers ensures a write cached on one controller is also held on the peer, allowing takeover without data loss.

Multipathing (e.g., Huawei UltraPath) presents multiple physical paths to a LUN as one logical path; ALUA defines active/optimized and active/non-optimized target port states.

LUN masking maps a LUN to specific host initiators (WWN or IQN); without masking, a host may see and corrupt LUNs it does not own.

Thin provisioning allocates physical capacity on demand; over-subscription ratio = total provisioned capacity / physical capacity; monitor for out-of-space risk.

A snapshot preserves a point-in-time image; copy-on-write snapshots store only changed blocks, while full copies duplicate all data and consume full capacity.

Huawei OceanStor DeviceManager is the GUI for configuration, alarms, and performance monitoring; SmartKit provides host-side inspection and upgrade tools.

Huawei FusionSphere comprises FusionCompute (virtualization), FusionManager (management), and FusionStorage (distributed storage).

FusionCompute components: VRM (management), CNA (compute node agent), and UVP (hypervisor kernel); VRM manages clusters and CNA runs VMs.

FusionStorage uses a fully distributed architecture with DHT (distributed hash table) for data placement and no dedicated metadata node.

In FusionStorage, a VBS (virtual block store) handles I/O on compute nodes and OSDs (object storage devices) store data on disks; a VBS can access any OSD.

Huawei public cloud services include ECS (Elastic Cloud Server), EVS (Elastic Volume Service), OBS (Object Storage Service), and VPC (Virtual Private Cloud).

EVS volumes are block storage attached to ECS; OBS provides object storage with buckets and is accessed via REST/S3-compatible APIs.

Cloud service models: IaaS provides infrastructure, PaaS provides runtime and middleware, SaaS provides applications; the customer's management responsibility decreases from IaaS to SaaS.

A VPC isolates cloud network resources; subnets, security groups (stateful), and ACLs (stateless) control traffic; EIP provides public IP access to cloud resources.

Huawei backup and DR services include VBS (Volume Backup Service) for cloud disk backup and CSBS for cloud server backup; both support policy-based scheduling.

Project life cycle phases: initiation, planning, execution, monitoring and control, and closure; each phase produces deliverables and a phase gate review.

Triple constraint: scope, time, and cost are interdependent; a change in one requires adjustment in the others to maintain quality.

WBS (work breakdown structure) decomposes deliverables into work packages; the lowest level should be estimable and assignable.

Critical path is the longest duration path through the network diagram; it determines minimum project duration and has zero total float.

Earned value formulas: CV = EV − AC; SV = EV − PV; CPI = EV/AC; SPI = EV/PV; EAC = BAC/CPI (typical variance).

Risk response strategies: avoid, transfer, mitigate, accept for threats; exploit, share, enhance, accept for opportunities.

A RAID or storage migration project should include a rollback plan, a maintenance window, and a tested backup before cutover.

Change control: any scope change requires a formal change request, impact analysis, approval by the change control board, and updates to baseline documents.

Stakeholder register records interest, influence, and engagement strategy; high power/high interest stakeholders require active management.