Evaluating suitable cloud services

When designing a high-performance architecture, start by selecting the right compute, storage, and network services. Understanding your business characteristics and metrics, along with the cloud products available for each architecture component, helps you choose solutions that optimize system performance over time.

Database selection is a specialized task with many considerations beyond performance and is not covered in detail here. Instead, this topic focuses on the effective use of database product portfolios in a high-performance architecture. For example, many scenarios use caching databases such as Redis to significantly improve performance.

Compute

Compute services

Familiarize yourself with the main compute solutions to evaluate and select the right one for each business system and workload.

Alibaba Cloud service	Type	Business scenario	Key features
Elastic Compute Service (ECS)	Cloud server	Server migration, monolithic application environments, and custom images	Runs on physical servers. Meets system setting requirements at the VM level. Offers a rich set of instance families and specifications.
Container Service for Kubernetes (ACK) and Elastic Container Instance (ECI)	Container	Microservices and hybrid cloud deployments	Runs on cloud servers. Lightweight, fast deployment, portable, and scalable.
Function Compute (FC)	Function	Event-driven applications	Strong concurrency. Weaker computing power and more limitations for a single function.

Compute metrics

Compute performance is quantified by specific metrics, such as a disk handling 100,000 IOPS or a server providing 10 Gbps network bandwidth. Linking these cloud product metrics to your business system metrics is key to identifying performance bottlenecks and resource utilization. Use Cloud Monitor to continuously track these metrics and drive ongoing architecture optimization.

Compute specifications

For non-Serverless compute services, selecting the right instance family and specification is critical. An incorrect choice can lead to performance bottlenecks, unnecessary costs, and more complex future optimizations. Analyze tracked metrics to understand the relationship between your workload and CPU, memory, disk, and network usage, then select an instance family and memory-to-vCPU ratio that delivers optimal performance and cost-effectiveness.

For I/O-intensive workloads, such as medium-to-large OLTP core databases, medium-to-large NoSQL databases, real-time log analysis, and large-scale enterprise commercial software, select a storage-enhanced instance family for better storage performance. For network-intensive workloads, such as NFV/SD-WAN, audio and video, telecom service forwarding, caching, big data, and machine learning, select a network-enhanced instance family to meet higher requirements for bandwidth, packets per second (PPS), concurrent connections, and multiple NICs. For workloads that benefit from GPU or FPGA acceleration, select the appropriate heterogeneous computing specifications.

Storage

Storage services

The cloud offers a wider variety of storage services than traditional data centers. To fully leverage this advantage, adapt these services to your business needs rather than relying on a single storage type. During architecture design, identify the storage requirements for each business scenario and workload, and define core storage metrics to evaluate and select a suitable solution.

Alibaba Cloud service	Key features	Common business scenarios
Elastic Block Storage (EBS)	High performance, low latency, and 99.9999999% data reliability	I/O-intensive, high-performance, and low-latency businesses, such as applications, OLTP databases, and NoSQL databases
Object Storage Service (OSS)	Massive capacity, high throughput, secure, low cost, multiple access methods, and 99.9999999999% data reliability	Large-scale data tiering, data backup, website hosting, separation of dynamic and static resources, and CDN acceleration
Apsara File Storage NAS/CPFS/HDFS	High throughput, low latency, and 99.999999999% data reliability	Scientific computing, shared access and data persistence for containers, and data storage for online enterprise production applications

Storage solutions are typically evaluated across three categories: block storage, object storage, and file storage. Start by clarifying the access methods and paths for your business scenario. For example, an internet application that requires public network access through an object-based API is best served by OSS. Containerized applications that require shared read/write access in a private network with fully elastic persistent storage benefit more from file storage.

From the perspectives of functionality, compatibility, and high availability architecture, Alibaba Cloud provides additional storage service options. These include Tablestore for massive structured data, CSG, which is compatible with multiple storage services, a data migration service for online and offline migration, Hybrid Disaster Recovery (HDR) for disaster recovery, and Cloud Backup for backups.

Storage metrics

Beyond business scenarios and architecture requirements, convert your business system data into storage performance metrics to evaluate and select storage products. Key metrics include throughput, IOPS, I/O latency, access frequency, data scale, data growth rate, and data reliability. Benchmark tests and continuous performance data collection help you obtain and analyze these metrics.

Alibaba Cloud also offers different service types within each product to meet varying performance requirements. For example, Elastic Block Storage includes basic disks, ultra disks, standard SSDs, and various levels of enterprise SSDs (ESSDs). Object Storage Service offers multiple storage classes, such as Archive Storage, Infrequent Access, and Standard, allowing you to match storage tiers to specific business demands.

Network

Network services

Networking connects Alibaba Cloud services, data centers, and application components, and significantly impacts workload performance. Selecting the right network product portfolio can often optimize performance without requiring changes to your business logic.

Alibaba Cloud service	Key features	Common business scenarios
Cloud Enterprise Network (CEN)	Ultra-large scale, flexible networking, and intelligent O&M	Hybrid cloud networking
Global Accelerator (GA)	High availability, high security, easy deployment, and high-quality acceleration	Game acceleration, enterprise application acceleration, and internet application acceleration
Server Load Balancer (CLB/NLB/ALB)	Ultra-high performance, auto scaling, 99.995% high reliability, and rich forwarding features	On-demand service forwarding and native applications
NAT Gateway	High performance, high reliability, and easy O&M	Centralized egress
Anycast Elastic IP Address	Wide acceleration range, low latency, easy deployment, and high availability	Website acceleration and internet application acceleration

These networking products enable high-performance hybrid cloud networking, global nearest-point access for acceleration, cross-zone high availability, and public-to-private address translation. Combine them with services such as CDN to further increase concurrent performance for your compute resources.

Network metrics

Network performance is constrained by more factors than compute or storage. Focus on key metrics such as latency, packet loss rate, bandwidth, throughput, requests per second, concurrent connections, and new connections. Analyzing network performance requires examining the entire request path. Stress testing and continuous monitoring help you identify core workload metrics, locate bottlenecks, and dynamically optimize specifications and solutions.