The document introduces the main performance indicators for measuring the website acceleration conditions from front and back of Baidu AI Cloud CDN.
You can observe your website before and after using CDN according to the following main performance metrics. The indicators include but not limited to:
Delay: It means the time cost by a data packet sent to the website server from the user’s computer and rapidly returns to the user’s computer from the website server. The lower the delay is, the better the performance will be.
Download speed: It is the transmission rate when a user downloads data from the network or a network server. The quicker the download rate, the better the performance.
Opening speed: It is the speed at which a user opens a website. The faster the open speed, the better the performance.
Packet loss rate: It refers to the ratio of the number of user’s data packets lost in network transmission to the number of data sets sent.
Origin rate: Origin rate is divided into forwarded request proportion and origin traffic proportion.
Number of forwarded requests to origin: It is the proportion of the requests from edge nodes not cached, expired in cache (cacheable), and non-cacheable in total requests. The lower, the better the performance will be.
Origin traffic ratio: Origin traffic is the traffic generated by the file size of the origin request and the traffic generated by the request itself. So, the ratio of traffic forwarding to origin = Traffic forwarding to origin (traffic forwarding to origin + traffic that the user requests for access), the lower the ratio, the better the performance will be.
Cache hit probability: It is the proportion of the times of access to the data already cached by the node in the total access times when a terminal user accesses the acceleration node. The higher the cache hit rate, the better the performance.
The various application scenarios of CDN all have some specific metrics. You can view further based on your business scenarios:
Small files, mainly html, js, jpg and css images and webpage materials. Such kind of acceleration is most sensitive to delay. Generally speaking, longer page load time may result in great user losses.
Time of building the connection: Meaning the process of finishing DNS analysis and finding the corresponding IP address to create the TCP connection. Since the process established for TCP is a fixed three-way handshake, the time it takes to establish a connection can basically reflect the node resources and scheduling capabilities of the CDN service. The closer the node is, the shorter time it takes to establish a connection.
Time of the first package: It is the time from the client sending the request to receiving the first packet from the server. This metric reflects the overall performance of the CDN service node program, that is, how long it takes for the server program to schedule and process after the connection.
Large file downloads, including downloads of files in various types such as apk, rar, zip and ipa, with a single file larger than 10M. This type of scenario usually has no strict requirements for CDN service quality, and the core indexes are mainly Download speed and Total download time.
AV-on-demand scenario, the types of which include the audio files of mp4, flv, mkv and wmv. Lag rate is more noted in video on-demand scenarios. More accurately, video on-demand can be regarded as a branch of large files, but the CDN service quality requirements in video on-demand scenarios are stricter than large file downloads.
Premiere time: Premiere time is the time from opening to seeing the video screen. It is usually affected by domain name resolution, connection and first packet time.
Lag rate: Lag refers to the frame hysteresis under the scenarios of video/audio playback, resource loading, etc. Lag rate mainly refers to the proportion of lag time of video playback for every 100 users when the lag time of video playback for all users is reported. The lower the lag rate, the better the performance.
For the CDN service, the core metrics of this type of scenarios are similar to those of audio and video on-demand, and it has two metrics: premiere time and lag rate. In the industry, if the duration of a premiere live reaches 300ms and the lag rate is less than 15%, it can be seen as a superior CDN service.