CDN:Website loads slowly after enabling Alibaba Cloud CDN acceleration

Analysis approach

Before you start troubleshooting, you must understand how CDN acceleration works. For more information, see What is Alibaba Cloud CDN. This helps you identify the potential causes of the issue. In summary, CDN adds a layer of cache nodes to the existing network. It distributes the resources from your origin server to the node that is closest to the user. When a client requests a resource, the client accesses the nearest CDN node to retrieve the cached content. This reduces the number of origin fetches and improves website speed. The possible causes of slow access include the following:

• Local network issues on the client, such as insufficient downstream bandwidth or an incorrect DNS configuration.

• Poor network quality between the client and the CDN node, which results in high latency.

• Abnormal CDN node behavior that causes slow responses.

• Large resource size, which leads to long download times.

• Poor network quality during CDN origin fetches.

• Slow response from the origin server.

To narrow down the scope of troubleshooting, you must gather information about the symptoms. This is a crucial step.

1. First, determine whether the slow access affects all users, only specific users, or users in a specific region or from a specific ISP. Alibaba Cloud provides products such as Application Real-Time Monitoring Service (ARMS) and Cloud Monitor. You can use these products to deploy probes in specific regions or for specific ISPs to obtain more accurate diagnostic results.

   • If only a few users experience slow access, the issue is likely caused by their local networks.

   • Check whether the affected users are located in a specific area. For example, if many China Mobile users in a city report slow access but China Unicom and China Telecom users in the same city do not, the issue may be related to the ISP in that region. You can use a diagnostic tool to run probes from that location.

   • If all users experience slow access, the issue is likely caused by the origin server or a configuration. It is highly unlikely that all CDN nodes or networks fail at the same time. For example, you can check whether the acceleration region is misconfigured, whether requests are for dynamic or uncacheable content, or whether the origin server responds slowly.

2. Determine whether slow or abnormal requests are served from the CDN cache.

   • If a request hits the CDN cache, an origin fetch is not required. The CDN node returns the cached data directly to the client. In this case, the origin server is not involved.

   • If a request misses the cache, you must check whether the slowdown occurs between the client and the CDN node or at the origin server.

Key metrics

In addition to standard monitoring metrics, CDN acceleration involves scenario-specific indicators. You can track these metrics to evaluate CDN performance and optimize your business configurations. For more information, see CDN key metrics.

Information gathering

A complete HTTP request follows this sequence: DNS resolution > TCP connection > SSL handshake (for HTTPS) > client request > server response. A deep understanding of this process helps you analyze issues. You must gather the following information from the client.

Obtain the client network status and the IP address of the CDN node

You can run the ping command to test the accelerated domain name. This lets you verify that the domain name is correctly resolved to a CDN node and check the network connectivity and latency between the client and the CDN node. If the ping command fails, you can perform link diagnostics. For more information, see Link diagnostics methods.

Obtain the client IP address and LocalDNS

CDN node scheduling relies on the client's LocalDNS. Make sure that the client uses the correct LocalDNS for its region and ISP. You can visit the Ali Kunlun User Diagnostic Tool to obtain the client's IP address and DNS settings.

Identify the URLs of slow-loading resources

You can open the developer tools of your browser, go to the Network tab, enter the URL, and then review all HTTP requests. Click the Time column to sort the requests by duration and identify slow-loading resources. In the Domain column, you can find the URLs of slow-loading resources under your CDN-accelerated domain name.

Obtain the HTTP request and response headers

In the Network tab, click the Name of the slow request. In the Headers section, you can view the General, Response Headers, and Request Headers sections. These headers indicate whether the request is for a static resource and whether the request hit the cache.

Obtain the HTTP request Timing data

The Timing tab shows the time spent in each phase of the request lifecycle.

Common cases

After you understand the principles of CDN acceleration and the HTTP request process, you can combine the observed symptoms with the client-side data to identify or locate issues. The following sections describe common scenarios.

Case 2: Low cache hit ratio or frequent origin fetches

In static resource acceleration scenarios, CDN caches static content on the node that is closest to the client. Users can retrieve resources directly from the cache. This avoids long-distance origin fetch paths. A low cache hit ratio increases the load on the origin server and reduces the acceleration efficiency for static resources. The cache hit ratio directly affects user experience. Therefore, achieving a high cache hit ratio is a core goal of CDN. You must optimize configurations based on the specific causes of a low cache hit ratio. For more information, see Improve the cache hit ratio of Alibaba Cloud CDN. You can check the X-Cache field in the CDN response header to verify whether a request hits the cache.

The following sections describe the symptoms and optimization strategies for low cache hit ratios or frequent origin fetches:

• The first-time access to a resource is slower than direct access to the origin server. This is because the CDN node does not have the resource and must first retrieve it from the origin server. You can use the URL prefetch feature to preload content onto CDN nodes. For more information, see Purge and prefetch resources. This ensures that first-time users can hit the cache, which improves the loading speed.

• Resources with low traffic have low popularity. CDN nodes are shared among all users. Cache rules define the maximum cache duration. If your accelerated domain name has low traffic, files may be evicted early based on their popularity. Popularity is defined by access frequency. Files with low popularity are removed from the cache sooner.

• A poor cache configuration with a short cache duration leads to frequent origin fetches.

  • If you do not configure custom cache rules, static files that do not have ETag or Last-Modified response headers cannot be cached. To resolve this issue, you can add these headers at the origin server or configure CDN cache rules. For more information, see Configure cache rules.

  • If you do not configure custom cache rules, CDN uses the default policies. For more information, see Default cache rules of Alibaba Cloud CDN and their priorities. The default cache durations are short, for example, 3,600 seconds or less. This causes frequent expiration and origin fetches. You must set appropriate cache durations based on your business needs.

  • If the origin server sends Cache-Control headers that disable caching, such as s-maxage=0, max-age=0, no-cache, no-store, or private, or a Pragma: no-cache header, CDN ignores your cache rules because the headers have a higher priority. You must modify these headers at the origin server to cacheable values, such as Public. For more information, see Configure an NGINX HTTP caching policy and Configure an Apache caching policy.

• URLs with variable parameters: If resource URLs include changing parameters, CDN treats each unique URL as a new request, even if the URLs point to the same file. This triggers unnecessary origin fetches. You can enable parameter filtering. For more information, see Ignore parameters.

• Range origin fetch for large files: For large files, you can enable range origin fetch to optimize origin retrieval. For more information, see Configure range origin fetch.

Case 3: Slow dynamic requests

If slow requests are for dynamic content, each client request must be sent to your origin server to generate real-time data. CDN cannot cache dynamic content. Therefore, CDN does not provide acceleration benefits and only forwards requests to the origin server. If your website or application has many dynamic elements, such as API operations, you can consider the following solutions:

1. Separate static and dynamic content. You can use a CDN-accelerated domain name for static resources and a separate domain name that points directly to the origin server for dynamic requests.

2. Use Dynamic Route for CDN (DCDN) to accelerate dynamic requests. For more information, see What is DCDN?. Note: DCDN accelerates dynamic requests using Layer 4 routing and transmission optimizations to reach your origin server faster. However, if your origin server responds slowly, you must optimize the origin server.

Case 4: Slow cross-border origin fetches

If your origin server and primary users are in different countries, CDN origin fetches may be affected by undersea cable capacity limits and public network congestion. This can cause unstable origin fetch quality, slow content retrieval, and high failure rates, which affects CDN caching and delivery. To resolve this issue, you can integrate Global Accelerator (GA) to accelerate your origin server for specific countries and then configure region-specific CDN back-to-origin policies.

Depending on your origin server type, you can see the following topics to configure GA acceleration:

• Accelerate ECS origin servers. For more information, see Use GA with CDN to accelerate origin fetches.

• Accelerate ALB origin servers. For more information, see Use GA to accelerate ALB backend applications.

• Accelerate OSS origin servers. For more information, see Access OSS using transfer acceleration.

Case 5: Slow homepage loading

When a client accesses http://www.example.com/, the browser requests the homepage. After the request is successful, the server returns HTML code. The browser then requests additional resources, such as images, JS files, and CSS files, that are referenced in the HTML code. If the homepage is dynamic or uncacheable, every request triggers an origin fetch. A slow origin server causes slow homepage loading, which is displayed as a long Pending state in the Network tab. To verify whether a request hits the cache, see Case 2: Low cache hit ratio or frequent origin fetches.

In this scenario, the homepage request remains in the Pending state until data is returned. After the homepage is loaded, static resources are loaded quickly.

Case 6: Large website resources

If your website loads large resources, you can enable performance optimization for your accelerated domain name. For more information, see Performance optimization. This feature can reduce file sizes and improve acceleration efficiency and page readability. The following file formats are supported for intelligent compression: text/html, text/xml, text/plain, text/css, application/javascript, application/x-javascript, application/rss+xml, text/javascript, image/tiff, image/svg+xml, application/json, and application/xmltext.

Case 7: Slow access for users of a specific ISP in a region

Sometimes, affected clients have common characteristics. For example, many China Mobile users in a city report slow access during a specific period, whereas China Unicom and China Telecom users do not. This issue is likely related to the local ISP network or the CDN node that serves that region. You can start troubleshooting by obtaining ping data from affected users to check the latency between the client and the CDN node.

In addition, you can bind your test machine to the IP address of the CDN node that is reported by users and run tests. This is similar to binding the test machine directly to the origin server, but you use the IP address of the CDN node instead. Before you run tests, verify whether the node itself responds slowly. If the node responds slowly, check whether requests hit the cache and check the resource size. Then, analyze the issue based on the preceding cases. If you cannot identify the issue, you can submit a ticket to Alibaba Cloud.