Application High-Availability Service (AHAS) Application Protection is the commercial version of the open source framework Sentinel and a core component of the technology stack for Alibaba's Double 11. It uses traffic as an entry point to ensure business stability through throttling, circuit breaking, and system load protection. It provides advanced traffic protection and analyzes traffic distribution within seconds.
Scenarios
AHAS Application Protection is widely used in scenarios such as flash sales, message-based peak-load shifting, cluster throttling, and real-time circuit breaking. It ensures business stability across multiple dimensions.
In a typical distributed application, as shown in the following figure, a request from a terminal first reaches the gateway. It then passes through a firewall and a network load balancer. The request also involves calls to other downstream services and third-party applications before it reaches the frontend web service. At different layers, AHAS Application Protection provides real-time traffic analysis with second-level granularity. For example, at the client layer, it provides real-time traffic monitoring and distribution analysis. This helps O&M engineers take targeted protective measures to ensure application stability.
Features
Traffic analysis with second-level granularity and real-time push of dynamic rules.
Advanced and diverse protection methods:
Inbound traffic throttling: Throttles traffic based on service capacity. This is often used for application entry points, such as gateways, frontend applications, and service providers.
Hot spot fencing: Isolates hot spot traffic from normal traffic. This prevents hot spots from consuming capacity intended for normal traffic.
Dependency fencing or degradation: Applies fencing or degradation to dependencies between and within applications to minimize the impact of unstable dependencies and ensure application stability.
System protection: AHAS Application Protection dynamically adjusts inbound traffic based on system capacity, such as load and CPU utilization, to ensure system stability.
Real-time, per-machine monitoring with support for aggregation and historical monitoring queries.
References
For more information about Sentinel, see About Sentinel.