Alibaba ACA Cloud Computing key terms

A geographic area where Alibaba Cloud places data centers, chosen so that resources and data reside close to users and within a required legal jurisdiction.

An isolated physical location within a region, with independent power and network, used to spread resources so a single facility failure does not take an application down.

The split of duties where the cloud provider secures the physical infrastructure and platform, while the customer secures their own data, accounts, configurations, and access policies.

The ability to add or remove computing resources automatically as demand changes, so capacity tracks workload instead of being fixed for peak usage.

A billing model where usage is charged by actual consumption, such as per second or per GB, with no upfront commitment and no long-term contract.

Alibaba Cloud's virtual server product, giving you configurable CPU, memory, image, disk, and network settings that you manage like a physical machine.

A named hardware profile for an ECS instance that fixes the CPU-to-memory ratio and performance class, such as general purpose, compute optimized, or memory optimized.

A template containing an operating system and optional preinstalled software used to create ECS instances consistently and to launch replacements quickly.

A service that automatically adds or removes ECS instances based on rules or schedules, keeping capacity matched to load and replacing unhealthy instances.

A logical collection of ECS instances managed together by Auto Scaling, sharing minimum, maximum, and desired capacity settings and a common configuration template.

A discounted ECS instance that uses spare capacity and can be reclaimed by the provider with short notice, suited to interruptible batch or stateless workloads.

Alibaba Cloud's highly durable object store that holds unstructured data as objects inside buckets, accessed over HTTP through APIs, SDKs, or URLs.

The top-level container in OSS that holds objects, has a globally unique name, lives in one region, and carries its own access and redundancy settings.

A durability and cost tier for stored objects, such as Standard, Infrequent Access, Archive, and Cold Archive, each trading retrieval speed for lower storage price.

An OSS policy that automatically transitions objects to cheaper storage classes or deletes them after set ages, reducing manual cleanup and storage cost.

The regional service address, internal or public, that clients use to reach a bucket; internal endpoints keep traffic inside the VPC without public exposure.

A distributed cache layer that serves content from edge nodes near users, cutting latency and origin load by caching files close to the request source.

A cache server in a CDN point of presence that stores copies of content and answers user requests locally instead of forwarding every request to the origin.

The share of CDN requests answered from edge caches rather than fetched from origin; a low rate signals poor caching rules or content that changes too often.

The authoritative source of content, such as an OSS bucket, ECS instance, or load balancer, that a CDN pulls from when a cache miss occurs.

A logically isolated private network you define with your own IP address range, subnets, route tables, and gateways, separating your resources from other tenants.

A subnet inside a VPC that is bound to one availability zone and supplies the IP range from which ECS and other resources receive private addresses.

A set of rules inside a VPC that decides where outbound traffic for a subnet goes, such as to a gateway, NAT device, or peered network.

A managed VPC component that lets instances without public IP addresses reach the internet for outbound traffic while blocking unsolicited inbound connections.

A managed traffic distributor that spreads incoming requests across multiple backend servers, performs health checks, and removes failed servers from rotation.

The group of ECS instances or endpoints behind a load balancer that receive forwarded traffic, configured with weights and ports for each member.

A periodic probe a load balancer sends to backends to confirm they respond correctly, so unhealthy members are skipped and traffic goes to working servers.

A managed service that connects VPCs and on-premises networks across regions through a global backbone, simplifying routing compared with building many individual peering links.

A stateful virtual firewall attached to ECS instances that allows or denies inbound and outbound traffic by protocol, port, and source or destination address.

Alibaba Cloud's identity service for creating users, groups, and roles and granting them scoped permissions through policies instead of sharing the root account.

A JSON permission document that states which actions are allowed or denied on which resources, attached to users, groups, or roles to enforce least privilege.

An identity with permissions but no permanent credentials, assumed temporarily by trusted users, services, or accounts to obtain short-lived access tokens.

A temporary credential issued when a role is assumed, carrying limited permissions and an expiry time, safer than embedding long-term keys in code.

A long-term credential pair of an ID and secret used for programmatic API calls; it must be stored securely, rotated, and never committed to source code.

Alibaba's family of large language and multimodal models, offered in different sizes and tuned variants for chat, coding, and reasoning tasks.

Alibaba Cloud's platform for building generative AI applications, offering hosted model APIs, prompt tooling, retrieval augmentation, and agent orchestration.

Alibaba Cloud's machine learning platform providing notebooks, training jobs, model deployment, and pipeline tools for building and running custom models.

A pattern where relevant documents are retrieved and supplied to a language model as context, grounding answers in private data and reducing unsupported output.