Architecture
The Kunpeng BoostKit for CDN complies with open architecture standards and supports all open source CDN components. It is a great choice for a wide range of scenarios. Figure 1 shows the open source software supported by the Kunpeng BoostKit for CDN. For details about the components, see Table 1.
|
Component |
Description |
|---|---|
|
Layer 4 load balancing |
A common open source component is LVS. |
|
Layer 7 load balancing |
Common open source components are Nginx and Tengine. |
|
Cache management |
Common open source components are Varnish, Squid, and ATS. |
|
SSL offload |
Offloads RSA computation to the Kunpeng Accelerator Engine (KAE) provided by the Kunpeng server to release CPU computing power. |
|
Hardware platform |
Kunpeng server. |
Kunpeng BoostKit for CDN applies to the L1 edge node CDN and L2 central node or regional node CDN scenarios, as shown in Figure 2.
A CDN is usually deployed at multiple layers. The L1 edge nodes are close to subscribers and directly serve subscribers. If the content requested by subscribers is not on the L1 edge nodes, the system retrieves the content from the upper-layer nodes until the content retrieval is successful or the content is retrieved from the origin servers. To prevent excessive bandwidth consumed on content retrieval and reduce the stress on the origin servers, an L2 central node or regional node is used as the content retrieval buffer of L1 edge nodes. The global scheduler is used to allocate the best L1 edge node and the optimal content retrieval node for subscribers.
The content below describes the process. The request from a subscriber is first sent to the local DNS server, and then the optimal L1 edge node is selected by the global scheduler. The request is forwarded to the corresponding L1 edge node. If the content is hit in the local cache of the L1 edge node, data in the L1 edge node is directly read and sent to the subscriber. If the content is not hit, the upper-layer nodes are queried for the content. If the content is hit on an upper-layer node, the content is returned. If the content is not hit, the upper-layer nodes are queried for the content, until the content is retrieved from the origin servers.