HCIE-Cloud Service flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Availability Zone

An isolated physical site within a region with independent power and network. Deploying tiers across multiple AZs removes a single-site failure domain and is the baseline high-availability pattern.

Enterprise Project

A resource-grouping and authorization dimension in Huawei Cloud that lets one account separate departments or projects, applying IAM policies and cost views per group of resources.

IAM Policy Structure

A JSON document with Effect (Allow or Deny), Action (service operations), Resource (target ARNs), and optional Condition blocks. System policies are predefined; custom policies are user-authored.

Security Group vs ACL

Security groups are stateful, instance-level, allow-only rules. Network ACLs are stateless, subnet-level rules that support both allow and deny and must permit return traffic explicitly.

EVS Disk Types

Ultra-high I/O and high I/O are SSD-backed for latency-sensitive databases; general purpose SSD balances cost; standard SATA HDD suits sequential, low-IOPS workloads such as logs and backups.

OBS Storage Classes

Standard for frequent access, Infrequent Access for rarely read data, Archive and Deep Archive for long-term retention. Lifecycle rules move objects between classes automatically based on age.

Shared EVS Disk

A block volume that can attach to multiple ECS instances concurrently, used for cluster quorum disks or shared file scenarios. It requires a cluster-aware file system to avoid corruption.

GaussDB(DWS)

Huawei Cloud's massively parallel processing data warehouse service, column-oriented and distributed, intended for analytical SQL over large datasets rather than high-concurrency OLTP transactions.

Multi-AZ Stateful Tier

Running database or stateful services with a standby in a second AZ so that a site failure triggers failover. Stateless tiers scale horizontally behind an ELB instead.

ELB Layer Choice

Layer 7 load balancers route on HTTP host and path and support sticky sessions; layer 4 balancers forward raw TCP or UDP for maximum throughput and non-HTTP protocols.

Decoupling with DMS

Placing a message queue such as Kafka or RabbitMQ between services so producers and consumers scale and fail independently, absorbing traffic spikes without synchronous back-pressure.

Storage Selection Rule

OBS for unstructured objects and static assets, SFS Turbo for shared POSIX file access across instances, EVS for block storage attached to a single instance or shared-disk cluster.

Cloud Connect vs VPN

VPN is fast to provision and inexpensive but shares internet paths. Cloud Connect is a dedicated line chosen when bandwidth, latency, jitter, or availability must meet a strict SLA.

WAF Placement

Web Application Firewall sits in front of public HTTP(S) endpoints to filter injection and bot traffic. It complements, not replaces, Anti-DDoS, security groups, and host security agents.

Sizing Headroom Rule

Size ECS flavors from measured peak vCPU and memory plus roughly 30 to 50 percent headroom, then confirm with a load test before committing to a flavor family.

Cost Baseline Pattern

Use yearly subscription for steady-state baseline capacity, pay-per-use for burst, and auto scaling to release idle resources. Non-production workloads can use scheduled shutdown.

Landing Zone Order

Provision account and IAM structure first, then networking and security groups, then shared logging and key management, then workloads. Management ports must never be open to all addresses.

RFS Templates

Resource Formation Service templates, or the Terraform provider, describe infrastructure as code so environments are reproducible. State should be stored remotely and parameterized per environment.

SMS Migration Flow

Install the agent on the source server, run a replication test, perform a final incremental sync during cutover, then validate boot and drivers on the target before switching DNS.

DRS Cutover

Database Migration Service performs full plus incremental synchronization so the target stays current; cutover downtime is limited to stopping writes and repointing the application.

OMS Object Migration

Object Migration Service copies buckets between sources and targets with throttling and checksum verification. Very large datasets may use physical data transfer appliances instead of network transfer.

Observability Trio

CES provides metrics and alarms, LTS collects and searches logs, and CTS records API audit events. Alarm on leading indicators such as latency and connection counts, not only on outages.

RPO and RTO

Recovery Point Objective is tolerable data loss; Recovery Time Objective is tolerable downtime. These values decide whether backup, pilot light, or active-active architecture is required.

CBR Backup Policy

Cloud Backup and Recovery applies scheduled, retained backups to ECS, EVS, and BMS resources. Retention and frequency should be derived from the tier's RPO.

Migration 6R Classification

Classify each source workload as rehost, replatform, refactor, repurchase, retire, or retain before mapping it to a target cloud service and scheduling the cutover window.

Instance Unreachable Order

Check security group rules, then subnet network ACLs in both directions, then EIP binding, then the OS firewall, then whether the process listens on all interfaces rather than loopback.

CPU Steal Time

Time the virtual CPU waits while the physical host serves other tenants. High steal indicates noisy-neighbor contention and is resolved by changing flavor or host, not by code tuning.

EVS IOPS Ceiling

Each EVS disk type has fixed IOPS and throughput limits. If measured values saturate the ceiling, move to a faster type, stripe disks, or cache hot reads in DCS.

RDS Slow SQL Log

A diagnostic log of statements exceeding a threshold. Review it for missing indexes and full table scans; note that read replicas offload reads but never fix write-bound bottlenecks.

ELB Health Check

A periodic probe of path, port, and timeout. A backend returning a non-success response on the probe path is removed from rotation even if the application otherwise works.

OBS Access Denied

Resolve by checking IAM policy action and resource match, then bucket policy, then object ACL precedence. Confirm the client uses the endpoint for the bucket's actual region.

Idle Resource Cleanup

Unattached EVS volumes, unbound EIPs, and orphaned snapshots still incur charges. Regular inventory and deletion, plus OBS lifecycle rules, are the fastest cost reductions.