Invention Grant
- Patent Title: Techniques for determining client-side effects of server-side behavior using canary analysis
-
Application No.: US14922101Application Date: 2015-10-23
-
Publication No.: US11616719B2Publication Date: 2023-03-28
- Inventor: Michael Lloyd Cohen
- Applicant: NETFLIX, INC.
- Applicant Address: US CA Los Gatos
- Assignee: NETFLIX, INC.
- Current Assignee: NETFLIX, INC.
- Current Assignee Address: US CA Los Gatos
- Agency: Artegis Law Group, LLP
- Main IPC: G06F8/65
- IPC: G06F8/65 ; H04L45/42 ; G06F11/36 ; H04L67/01

Abstract:
In one embodiment of the present invention, a sticky canary router routes each request associated with a service to either a canary cluster of servers that implement a modification to the service or a baseline cluster of servers that do not implement the modification. The sticky canary router implements a mapping algorithm that determines the routing of each request based on a current time, a time window for the routing, and a characteristic of the request. Notably, the mapping algorithm may be implemented such that, for time segments with duration equal to the time window, the sticky canary router routes all requests received from a particular device in a consistent fashion—either to the canary cluster or to a baseline cluster. Configured thusly, the sticky canary router enables the analysis of approximately full sections of client interactions with the canary servers, thereby facilitating identification of client-side effects of the changes.
Public/Granted literature
- US20170118110A1 TECHNIQUES FOR DETERMINING CLIENT-SIDE EFFECTS OF SERVER-SIDE BEHAVIOR USING CANARY ANALYSIS Public/Granted day:2017-04-27
Information query