United States Patent
US Patent 7894372: Topology-centric resource management for large…
US 7894372 · granted 2011-02-22
Abstract
Topology-centric resource management for large scale service clusters is described herein. According to certain embodiments of the invention, techniques include 1) creating optimized topology with network switches to connect service modules based on application flows and bandwidth requirements, 2) providing centralized or decentralized monitoring schemes to maintain the topology view of a service cluster, and 3) using the topology information for optimizing load balancing and service information dissemination. Other methods and apparatuses are also described.
| Patent Number | 7894372 |
|---|---|
| Title | Topology-centric resource management for large scale service clusters |
| Filed | 2005-05-31 |
| Granted | 2011-02-22 |
| Inventor(s) | Chu; Lingkun, Yang; Tao, Zhou; Jingyu |
| Assignee | IAC Search & Media, Inc. |
| CPC Classification | H04L 12/28, H04J 3/26, H04J 3/24 |
| Number of Claims | 28 |
Abstract
Topology-centric resource management for large scale service clusters is described herein. According to certain embodiments of the invention, techniques include 1) creating optimized topology with network switches to connect service modules based on application flows and bandwidth requirements, 2) providing centralized or decentralized monitoring schemes to maintain the topology view of a service cluster, and 3) using the topology information for optimizing load balancing and service information dissemination. Other methods and apparatuses are also described.
Claim 1
A computer implemented method, comprising: dividing a plurality of nodes of a service cluster into a plurality of groups each having a plurality of members, each grouphaving a group leader node to communicate with other groups of the service cluster; allowing a node to join a group; (i) determining whether the node joining the group is a leader of the group; and (ii) if the node joining the group is the leader ofthe group, increasing a time-to-live (TTL) value of the node joining the group, whereupon the node joins another group at a next higher level of a hierarchy of groups, each group comprising nodes having a same TTL value; wherein steps (i) and (ii) arerepeated until the node is not elected as a leader of a current group or the TTL value of the node reaches a maximum value; and in response to a message received from a member of a respective group, the group leader node of the respective groupdistributing the message to the other groups of the service cluster and a remainder of the members of the respective group.
Claims
28 totalA computer implemented method, comprising: dividing a plurality of nodes of a service cluster into a plurality of groups each having a plurality of members, each grouphaving a group leader node to communicate with other groups of the service cluster; allowing a node to join a group; (i) determining whether the node joining the group is a leader of the group; and (ii) if the node joining the group is the leader ofthe group, increasing a time-to-live (TTL) value of the node joining the group, whereupon the node joins another group at a next higher level of a hierarchy of groups, each group comprising nodes having a same TTL value; wherein steps (i) and (ii) arerepeated until the node is not elected as a leader of a current group or the TTL value of the node reaches a maximum value; and in response to a message received from a member of a respective group, the group leader node of the respective groupdistributing the message to the other groups of the service cluster and a remainder of the members of the respective group.