Method and apparatus for an internet protocol (IP) network clustering system

• US 6,006,259 A
• Filed: 11/20/1998
• Issued: 12/21/1999
• Est. Priority Date: 11/20/1998
• Status: Expired due to Term

• Current Assignee: Nokia Technologies Oy (Nokia Corporation)
• Original Assignee: Network Alchemy Incorporated (Nokia Corporation)
• Inventors: Kashtan, David Lyon, Palter, William L., Piper, Derrell D. II, Adelman, Kenneth Allen
• Primary Examiner(s): Meky, Moustafa M.
• Application Number:
  US09/196,941
• Time in Patent Office: 396 Days
• Field of Search: 709/200, 709/201, 709/202, 709/203, 709/206, 709/208, 709/209, 709/217, 709/218, 709/219, 709/220, 709/221, 709/223, 709/224, 709/225
• US Class Current: 709/223
• CPC Class Codes: H04L 61/00 Network arrangements, protocols or services for addressing or naming
• CPC Class Codes: H04L 61/10 of different types
• CPC Class Codes: H04L 61/35 involving non-standard use of addresses for implementing network functionalities, e.g. coding subscription information within the address or functional addressing, i.e. assigning an address to a function
• CPC Class Codes: H04L 67/1001 for accessing one among a plurality of replicated servers
• CPC Class Codes: H04L 67/1008 based on parameters of servers, e.g. available memory or workload monitoring of computer activity G06F11/30
• CPC Class Codes: H04L 67/1023 based on a hash applied to IP addresses or costs
• CPC Class Codes: H04L 67/1029 using data related to the state of servers by a load balancer
• CPC Class Codes: H04L 67/1034 Reaction to server failures by a load balancer
• CPC Class Codes: H04L 69/16 Implementation or adaptation of Internet protocol [IP], of transmission control protocol [TCP] or of user datagram protocol [UDP]
• CPC Class Codes: H04L 69/163 In-band adaptation of TCP data exchange; In-band control procedures
• CPC Class Codes: H04L 9/40 Network security protocols

Abstract

The present invention is an Internet Protocol (IP) network clustering system which can provide a highly scalable system which optimizes message throughput by adaptively load balancing its components, and which minimizes delay and packet loss especially in the TCP mode by a controlled fail-over process. An innovative process of defining the essential portion of TCP state required to maintain reliable message connections across the cluster is also disclosed.

Claims

1. An Internet Protocol (IP) Network cluster apparatus comprising:

a. a plurality of cluster members with all cluster members being addressable by a single dedicated Internet machine name and IP address for the cluster, each cluster member comprising a computer system having a processor, a memory, a program in said memory, a display screen and an input/output unit;

b. a filter mechanism in each cluster member, the filter mechanism using a hashing mechanism to generate an index number for each message session received by the cluster member, the index number being used to indicate to which workset a message belongs, worksets being assigned to cluster members to balance processing load, each cluster member checking whether the workset has been assigned to it in order to determine whether the cluster member must process the message received or ignore it.

2. The apparatus of claim 1 further comprising an assignment mechanism in each cluster member, for use by a cluster member designated as a master unit, the assignment mechanism used when a message of an unassigned message session is received by the master unit, the assignment mechanism using the index number calculated by the filter mechanism to assign sets of message sessions to cluster members for further processing in order to load balance processing of incoming messages.

3. The apparatus of claim 1 further comprising a first program code mechanism in each of the plurality of cluster members configured to save state for each message session including TCP state.

4. The apparatus of claim 3 further comprising a second program code mechanism in each of the plurality of cluster members configured to transfer an essential portion of the saved state for each message session to each of the other cluster members, whenever required.

5. The apparatus of claim 4 further comprising a third program code mechanism in each of the plurality of cluster members configured to permit a cluster member acting as a master unit to recognize an equipment failure in one of the other members in the cluster, to reassign the work of the failed cluster member to remaining members in the cluster thereby rebalancing the processing load and maintaining the message sessions.

6. The apparatus of claim 5 further comprising a fourth program code mechanism in each of the plurality of cluster members configured to permit units which are not acting as the master unit to recognize an equipment failure in the master unit, to immediately and cooperatively designate one of the remaining cluster members as a new master unit, the new master unit to reassign the work of the failed cluster member to remaining cluster members thereby rebalancing the processing load and maintaining the message sessions.

7. The apparatus of claim 1 wherein the memory of each of the cluster members includes a flash memory card containing a program code mechanism which describes the personality of the cluster member including its cluster address.

8. A method for operating a plurality of computers in an Internet Protocol (IP) Network cluster, the cluster providing a single-system-image to network users, the method comprising the steps of;

a. providing a plurality of cluster members, each cluster member comprising a computer system having a processor, a memory, a program in said memory, a display screen and an input/output unit;

b. interconnecting the cluster members together, and assigning all cluster members a same internet machine name and a same IP address whereby a message arriving at the cluster will be recognized by the appropriate member in the cluster and an output from any cluster member will be recognized as coming from the cluster, and whereby the cluster members can communicate with each other; and

c. providing a filter mechanism in each cluster member, the filter mechanism using a hashing mechanism to generate an index number for each message session received by the cluster member, the index number being used to indicate to which workset a message belongs, worksets being assigned to cluster members to balance processing load, each cluster member checking whether the workset has been assigned to it in order to determine whether the cluster member must process the message received or ignore it.

9. The method of claim 8 further comprising an assignment mechanism in each cluster member, for use by a cluster member designated as a master unit, the assignment mechanism used when a message of an unassigned message session is received by the master unit, the assignment mechanism using the index number calculated by the filter mechanism to assign sets of message sessions to cluster members for further processing in order to load balance processing of incoming messages.

10. The method of claim 8 comprising the additional step of each cluster member saving state for each message session connection including TCP state, and for segregating this state into an essential state portion and a non-essential state portion.

11. The method of claim 10 comprising the additional step of each cluster member transferring to each other cluster member the saved essential state portion for message sessions for which that cluster member is responsible, such transfer to be made whenever the essential portion of the state changes, whereby all cluster members maintain essential state for all message session connections.

12. The method of claim 11 comprising the additional step of each cluster member recognizing the equipment failure of one of the cluster members, immediately reassigning a task of being the master if it is the master unit that failed, the master unit reassigning the work which was assigned to the failed cluster member, rebalancing the load on the remaining tunnel-servers.

13. An Internet Protocol (IP) network cluster apparatus comprising:

a. a plurality of interconnected cluster members, each cluster member comprising a computer system having a processor, a memory, a program in said memory, a display screen and an input/output unit;

b. means in each of the plurality of cluster members for recognizing other members of the plurality of cluster members which are connected together and cooperating with the other members to adaptively designate a master unit; and

c. means for generating an index number for each message session received by a cluster member, the index number being used to indicate whether the cluster member must process the message received or ignore it.

14. The apparatus of claim 13 further comprising means in each of the plurality of cluster members for saving essential state for each message session.

15. The apparatus in claim 14 further comprising means in each of the plurality of cluster members for periodically transferring the saved essential state for each message session to each of the other members in the cluster.

16. The apparatus of claim 15 further comprising means in each of the plurality of cluster members for permitting a cluster member acting as a master unit to recognize an equipment failure in one of the other cluster members, and for reassigning work of the failed cluster member to remaining members in the cluster thereby rebalancing the processing load and maintaining message session connections, and for permitting cluster members which are not acting as a master unit to recognize an equipment failure in the master unit, to immediately and cooperatively designate one of the remaining cluster members as a new master unit, the new master unit to reassign work of the failed cluster member to remaining members in the cluster thereby rebalancing the processing load and maintaining message session connections.