Active-active data center using RHI, BGP, and IGP anycast for disaster recovery and load distribution

• US 20060193252A1
• Filed: 02/25/2005
• Published: 08/31/2006
• Est. Priority Date: 02/25/2005
• Status: Active Grant

• Current Assignee: Cisco Technology, Inc. (Cisco Systems, Inc.)
• Original Assignee: Cisco Technology, Inc. (Cisco Systems, Inc.)
• Inventors: Naseh, Zeeshan, Gundi, Vinay
• Granted Patent: US 7,609,619 B2
• Time in Patent Office: Days
• US Class Current: 370/225
• CPC Class Codes: H04L 45/04 Interdomain routing, e.g. hierarchical routing
• CPC Class Codes: H04L 45/28 using route fault recovery

Abstract

A distributed data center topology having at least a pair of active data centers that can recover from a disaster at one of the data centers and achieves load balancing using IGP (Interior Gateway Protocol) between data centers. The distributed data centers use virtual IP addresses, route health injection and Border Gateway Protocol (BGP) for business continuance, disaster recovery and load balancing. The active/active topology supports load balancing where each site concurrently hosts active applications or applications can be hosted in a logical active/standby mode. IGP and RHI (Route Health Injection) are used to propagate routes to an edge router and BGP (Border Gateway Protocol) and IP Anycast are used for site-to-site recovery and load balancing between data center sites.

Claims

1. A data center topology coupled to the internet comprising:

at least two active data centers each coupled to the internet by at least one internet service provider, each data center having an edge router broadcasting a common IP address and linked to at least one other data center;

means for advertising a selected one of said active data centers such that traffic is load balanced between said data centers;

means for detecting a disaster at one of said data centers; and

means for removing a route from the routing tables of routers in said internet so that traffic is routed away from said data center having the detected disaster.

2. The data center topology of claim 1 wherein said traffic attracting means comprises a load balancing means that routes traffic based on the shortest distance from a client to one of said at least two data centers.

3. The data center topology of claim 1 wherein said means for detecting a disaster at one of said data centers uses route health injection to remove routes from internet routing tables of a router.

4. The data center topology of claim 1 further comprises an IGP link between said data centers.

5. The data center topology of claim 4 wherein said IGP link comprises a high bandwidth link.

6. The data center topology of claim 1 wherein each data center further comprises said edge router coupled by an internal network to a router and content switching module, said content switching module and said router sharing a common chassis for supporting virtual IP addresses.

7. The data center topology of claim 6 wherein the edge router of at least one of said data centers is coupled to multiple internet service providers.

8. In an autonomous system, a method for load balancing and recovering from a disaster comprising the steps of:

hosting active applications and serving clients from at least two active data centers in said autonomous system;

advertising the same route from said at least two active data centers;

routing client traffic to one of said plurality of data centers using IP Anycast mechanism;

using “

advertise active”

command in route health injection (RHI) to install a host route in an edge router if a virtual server in said data center is in an operational state;

advertising the availability of a virtual IP (VIP) address throughout the network;

injecting, from an internal routing protocol into an external routing protocol, a static route by redistribution;

configuring route summarization in OSPF;

configuring BGP route redistribution and route preference to update neighboring routers; and

using Interior Gateway Protocol and sub-net based load balancing on a direct link between said at least two active data centers.

9. The method of claim 8 further comprising the step of using Exterior Border Gateway Protocol (E-BGP) for interautonomous system routing and to exchange network routing information with other networks.

10. The method of claim 8 further comprising the step of controlling incoming and outgoing routes in BGP by the using weights and the MED attribute to prefer routes during steady state operation.

11. The method of claim 8 further comprising the step of probing the health of the servers in each data center and the appropriate application daemon that runs on the servers.

12. In an active/active data center topology having a plurality of data centers with each data center comprising at least a server cluster, a method for load balancing and managing traffic flow between at least two active data centers comprising the steps of:

operating a plurality of application services at a server cluster at each data center, of which application services are functional in both data center;

redistributing IGP routes into BGP at both data centers;

directing traffic to said data centers using an IP Anycast mechanism to achieve load balancing based on directing traffic to the closest data center;

monitoring the health of each data center;

in the event, one of said data centers is unavailable, removing a route from BGP so that future traffic is directed to another data center.

13. The method of claim 12 further comprising the step of activating route health injection (RHI) at each active data center using the “

advertise active”

command.

14. The method of claim 12 further comprising the step using the same virtual IP address (VIP) to represent the server cluster at each data center for each application.

15. The method of claim 14 wherein the same sub-net that comprises all the VIPs is advertised to internet with BGP.

16. The method of claim 15 further comprises the steps of redistributing RHI static routes into an IGP-specific protocol at said router and using an OSPF metric-type 1 for adding the internal cost to the external cost of each route.

17. In an autonomous system, a method for load balancing and recovering from a disaster comprising the steps of:

hosting active applications and serving clients from at least two active data centers in said autonomous system;

advertising the same route from said at least two active data centers;

routing client traffic to one of said plurality of data centers using IP Anycast mechanism;

using “

advertise active”

command in route health injection (RHI) to install a host route in an edge router if a virtual server in said data center is in an operational state;

advertising the availability of a virtual IP (VIP) address throughout the network;

injecting, from an internal routing protocol into an external routing protocol, a static route by redistribution;

configuring route summarization in OSPF; and

configuring BGP route redistribution and route preference to update neighboring routers.

18. The method of claim 17 further comprising the step of using Exterior Border Gateway Protocol (E-BGP) for interautonomous system routing and to exchange network routing information with other networks.

19. An autonomous system, comprising:

a first data center having an edge router that couples said first data center to an internet service provider;

a second data center having an edge router that couples said second data center to the internet by at least one internet service provider and coupled to said first data center by an IGP link

a load balancer for balancing traffic between said data centers; and

said edge router in said first data center and said edge router in said second data center broadcasting a common IP address and adapted to update neighboring routers in response to a health check that indicates the occurrence of a disaster at the other data center.

20. The autonomous system of claim 19 wherein said edge routers exchange health check information over said IGP link.