ER is in Beta release and available only to selected users. If the daily peak bandwidth of your DCDN service exceeds 500 Mbit/s, you can submit a ticket to activate ER. We apologize for the inconvenience this may have caused. For more information about public previews, visit the official website.
Global deployment, near-site scheduling, and ultra-low latency: Compared with common cloud computing services, such as Function Compute and Elastic Compute Service (ECS), you do not need to deploy ER in a specific region. ER runs on points of presence (POPs) around the world. Client requests are automatically scheduled to the nearest POPs, which immediately run your code and process client requests. ER can significantly reduce the server response time, which helps you deliver low-latency services to your customers.
Automatic scale-out and pay-as-you-go billing: Alibaba Cloud has constructed an edge network that consists of more than 3,200 POPs around the world. These POPs provide unparalleled scalability. If the number of client requests in a region spikes, these requests are automatically scheduled to the nearest POPs that have sufficient compute resources. You are charged for ER based on the number of calls that are made to the service.
Easy-to-use serverless mode: ER allows you to focus on your business logic without having to manage underlying resources, such as CPU, memory, network, and operating system. You can upload your code by using the console or API to deploy applications. Serverless development can effectively reduce your development and O&M costs.
How it works
Scenarios in which ER is not used: 1. Clients initiate requests to the gateways of the POPs. → 2. The POPs search for the requested content in the cache. If a request is a cache hit, the requested content is returned to the client. → 3. If a request is a cache miss, the request is redirected to the origin server.
In the preceding figure, ER is executed behind the gateway. In this case, the configurations of DCDN-accelerated domain names that are configured in the console, such as HTTPS configurations, access control, performance optimization, and cache settings, will continue to take effect.
Environment: ER provides the staging environment, production environment, and canary release environment.
Staging environment: A staging environment is an independent POP that is used to test code. The staging environment simulates the production environment but is isolated from the production environment. You can modify the routine configuration or code in the staging environment. This does not affect the production environment. You can access the staging environment only after you add the IP address that is provided on the page to your hosts file.
Production environment: The production environment, also known as the deployment environment, consists of a large number of POPs that are distributed around the world. After you test your code in the staging environment, you can deploy the code to the production environment. The preceding figure is provided for reference only. ER automatically adds or removes POPs based on your request volume. The POP distribution may vary based on actual scenarios.
Canary release environment: The canary release environment allows you to gradually expand the scope of your deployment. It is risky to deploy code or configuration changes directly to the production environment. If errors exist in your code or configurations, the errors may affect your users. If the scale of your business is large or you want to control risks when you make code or configuration changes, we recommend that you use the canary release environment to manage the deployment scope. The canary release environment is complex. We recommend that you understand the basics of the environment before you use the canary release environment. For more information, see Canary release environment (optional).
Domain name: After you create a routine and deploy code, you need to use a domain name as the portal of the routine that clients can access. ER supports only DCDN-accelerated domain names as portals. You can associate ER with domain names in the DCDN console.
POP: POPs are used by DCDN for content delivery. Client requests are scheduled to the nearest POPs.
5 ms, 50 ms, and 100 ms
The CPU time slice that is allocated by ER for an execution. I/O wait is not included in the calculated CPU time. You can select one of the three options: 5 ms, 50 ms, and 100 ms. For more information, see Billing rules.
Real time (RT)
The response time of an execution for ER cannot exceed 120s. I/O wait is included in the calculated RT.
The time during which the gateway waits for ER. If ER does not return data within 10s, the gateway disconnects from the server and returns the HTTP 504 status code to the client.
Code package size
Number of fetch requests
The number of fetch requests that is allowed by ER in an execution.
Number of routines
You can create up to 50 routines for each Alibaba Cloud account.
Number of versions
You can save up to 10 versions for each routine.
WebSocket, TCP, and UDP are not supported.
You cannot initiate a fetch request in a routine to a domain name of another routine.