System and method for regulating incoming traffic to a server farm

• US 6,883,033 B2
• Filed: 02/20/2001
• Issued: 04/19/2005
• Est. Priority Date: 02/20/2001
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Maruyama, Kiyoshi, Chan, Shun Shing, Zohar, Miram, Chaar, Jarir Kamel, George, David Alson, Lorrain, Jean A.
• Primary Examiner(s): Eng, David Y.
• Application Number:
  US09/921,989
• Publication Number:
  US 20020124103A1
• Time in Patent Office: 1,519 Days
• Field of Search: 709/229, 709/223, 709/238
• US Class Current: 709/229
• CPC Class Codes: H04L 47/10 Flow control; Congestion control
• CPC Class Codes: H04L 47/193 at the transport layer, e.g. TCP related
• CPC Class Codes: H04L 47/20 Traffic policing
• CPC Class Codes: H04L 47/22 Traffic shaping
• 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 67/1012 based on compliance of requirements or conditions with available server resources

Abstract

A system and method for controlling the rates at which application workload, e.g., TCP connection requests, is admitted to a collection of servers, such as a server farm of an application service provider (ASP) that hosts Internet World Wide Web (WWW) sites of various owners. The system and method are intended to operate in an environment in which each customer has a workload-based SLA for each type of application hosted by the provider and used by the customer. The system and method achieve support (minimum, maximum) TCP connection requests for multiple customers and applications. According to one aspect, the system and method guarantee, control and deliver TCP connection-based workload SLA'"'"'s to customers whose applications are hosted by the server farm with the use of a workload regulator that operates by regulating only new TCP connection request packets while transparently passing existing TCP connection packets and other request packets received for customers. The regulator further operates by regulating the flow of incoming TCP connection requests to each customer business activity application so as to guarantee a level of service previously agreed to each customer (per their respective SLA'"'"'s) by applying rate admittance to requests and by dropping (or rejecting) requests to guarantee the agreed service levels to the customer'"'"'s application.

Claims

1. A method of regulating admittance of a number of TCP connection requests to a plurality of servers that support to a plurality of applications and a plurality of customers, each of the customers having electronic business activity hosted by the plurality of servers on the basis of a service level agreement that specifies a minimum TCP connection rate that the customer is guaranteed, the method comprising the steps of:

receiving incoming workload for the plurality of customers for the plurality of applications into a common buffer, the incoming workload comprising IP packets associated with existing TCP connections and TCP connection requests;

taking an IP packet from the common buffer;

detecting whether the IP packet taken from the common buffer is a TCP connection request packet or is associated with an existing TCP connection;

immediately sending the IP packet to the servers if the IP packet is associated with an existing TCP connection;

rejecting the IP packet if associated with a TCP connection request unless the IP packet is necessary to meet the minimum TCP connection rate for the customer associated with the TCP connection request or if sending the IP packet to the servers does not prevent meeting the minimum TCP connection rate for another of the customers.

2. The method according to claim 1, wherein all of the incoming workload is received into the common buffer and the TCP connection request packets are processed on a first in, first out basis.

3. The method according to claim 1, wherein the method does not directly control outbound traffic from the plurality of servers.

4. The method according to claim 1, wherein the detecting step is performed by determining whether a SYN bit of the IP packet is ON or OFF.

5. The method according to claim 1, wherein the rejected IP packet is dropped.

6. The method according to claim 1, wherein the rejected IP packet is returned with a reset code bit ON.

7. The method according to claim 1, further comprising the step of calculating whether TCP connection request packets of one or more of the customers can be admitted for one or more of the applications when the one or more customers already exceed the minimum TCP connection rate for the one or more applications.

8. The method according to claim 1, further comprising the step of communicating with an external means to receive revisions to one or more of the service level agreements.

9. The method according to claim 1, further comprising the step of communicating with an external means to send collected statistics data to the external means.

10. The method according to claim 1, wherein the detecting, sending and rejecting steps are performed with a real-time admittance/rejection algorithm wherein the flow of the TCP connection request packets is regulated to the plurality of servers on a per-packet basis.

11. The method according to claim 1, wherein the detecting, sending and rejecting steps are performed with a target-rate-based admittance/rejection algorithm wherein the flow of the TCP connection request packets is regulated to the plurality of servers on a per-unit-time basis.

12. A method of controlling and guaranteeing a service level agreement based on a number of TCP connection requests supported to a plurality of applications and a plurality of customers having electronic business activity hosted by a server farm, each of the customers having a service level agreement that specifies a minimum TCP connection rate that the customer is guaranteed, the method comprising the steps of:

receiving all incoming workload for the plurality of customers for the plurality of applications into a common buffer, the incoming workload comprising IP packets associated with existing TCP connections and TCP connection requests;

taking TCP connection request packets from the common buffer on a first in, first out basis; and

controlling flow of the IP packets from the common buffer to the server farm for each application and each customer so as to provide at least the minimum TCP connection rate for each customer and each application by regulating flow of the TCP connection request packets to the server farm, the controlling step comprising;

detecting whether an IP packet is a TCP connection request packet or associated with an existing TCP connection;

immediately admitting IP packets associated with existing TCP connections;

admitting TCP connection request packets if necessary to meet the minimum TCP connection rate for the customer associated with the TCP connection request; and

admitting a TCP connection request packet of one of the customers if doing so does not prevent meeting the minimum TCP connection rate for another of the plurality of customers.

13. The method according to claim 12, wherein the method does not directly control outbound traffic from the serve farm.

14. The method according to claim 12, wherein the detecting step is performed by determining whether a SYN bit of the IP packet is ON or OFF.

15. The method according to claim 12, further comprising the step of dropping IP packets that are not admitted during the controlling step.

16. The method according to claim 12, further comprising the step of returning TCP connection request packets that are not admitted during the controlling step, the IP packets not admitted during the controlling step being returned with a reset code bit ON.

17. The method according to claim 12, further comprising the step of calculating whether a TCP connection request packet of one or more of the customers can be admitted for one or more of the applications when the one or more customers already exceed the minimum TCP connection rate for the one or more applications.

18. The method according to claim 12, further comprising the step of communicating with an external means to receive revisions to one or more of the service level agreements, and to send collected statistics data to the external means.

19. The method according to claim 12, wherein the controlling step is performed with a real-time admittance/rejection algorithm wherein the flow of the TCP connection request packets is regulated to the server farm on a per-packet basis.

20. The method according to claim 12, wherein the controlling step is performed with a target-rate-based admittance/rejection algorithm wherein the flow of the TCP connection request packets is regulated to the server farm on a per-unit-time basis.

21. A system for regulating admittance of a number of TCP connection requests to a plurality of servers that support to a plurality of applications and a plurality of customers, each of the customers having electronic business activity hosted by the plurality of servers on the basis of a service level agreement that specifies a minimum TCP connection rate that the customer is guaranteed, the system comprising:

a common buffer for receiving incoming workload for the plurality of customers for the plurality of applications, the incoming workload comprising IP packets associated with existing TCP connections and TCP connection requests; and

regulator means for taking an IP packet from the common buffer, detecting whether the IP packet is a TCP connection request packet or is associated with an existing TCP connection, immediately sending the IP packet to the servers if the IP packet is associated with an existing TCP connection, and reiecting the IP packet if associated with a TCP connection request unless the IP packet is necessary to meet the minimum TCP connection rate for the customer associated with the TCP connection request or if sending the IP packet to the servers does not prevent meeting the minimum TCP connection rate for another of the customer.

22. The system according to claim 21, wherein the common buffer is configured to receive all of the incoming workload, and the controlling means is configured to process the TCP connection request packets on a first in, first out basis.

23. The system according to claim 21, wherein the system is not configured to directly control outbound traffic from the plurality of servers.

24. The system according to claim 21, wherein the regulator means is configured to determine whether a SYN bit of an IP packet is ON or OFF.

25. The system according to claim 21, wherein the regulator means further comprises means for dropping rejected IP packets.

26. The system according to claim 21, wherein the regulator means further comprises means for returning rejected IP packet with a reset code bit ON.

27. The system according to claim 21, further comprising means for calculating whether TCP connection request packets of one or more of the customers can be admitted for one or more of the applications when the one or more customers already exceed the minimum TCP connection rate for the one or more applications.

28. The system according to claim 21, further comprising means for communicating with an external means to receive revisions to one or more of the service level agreements.

29. The system according to claim 21, further comprising means for communicating with an external means to send collected statistics data to the external means.

30. The system according to claim 21, wherein the regulator means operates with a real-time admittance/rejection algorithm wherein the flow of the TCP connection request packets is regulated to the plurality of servers on a per-packet basis.

31. The system according to claim 21, wherein the regulator means operates with a target-rate-based admittance/rejection algorithm wherein the flow of the TCP connection request packets is regulated to the plurality of servers on a per-unit-time basis.

32. A system of controlling and guaranteeing a service level agreement based on a number of TCP connection requests supported to a plurality of applications and a plurality of customers having electronic business activity hosted by a server farm, each of the customers having a service level agreement that specifies a minimum TCP connection rate that the customer is guaranteed, the system comprising:

a common buffer into which is received all incoming workload for the plurality of customers for the plurality of applications, the incoming workload comprising IP packets associated with existing TCP connections and TCP connection requests;

a gatekeeper that takes the IP packets from the common buffer on a first in, first out basis and controls flow of the IP packets from the common buffer to the server farm for each application and each customer so as to provide at least the minimum TCP connection rate for each customer and each application by regulating flow of the IP packets to the server farm, the gatekeeper being operable to detect whether an IP packet is a TCP connection request packet or associated with an existing TCP connection, immediately admit IP packets associated with existing TCP connections, admit a TCP connection request packet if necessary to meet the minimum TCP connection rate for the customer associated with the TCP connection request, and admit a TCP connection request packet of one of the customers if doing so does not prevent meeting the minimum TCP connection rate for another of the plurality of customers.

33. The system according to claim 32, wherein the system does not directly control outbound traffic from the serve farm.

34. The system according to claim 32, wherein the gatekeeper is configured to determine whether a SYN bit of the IP packet is ON or OFF.

35. The system according to claim 32, further comprising a guide means for dropping TCP connection request packets that are not admitted by the gatekeeper.

36. The system according to claim 32, further comprising a guide means for returning TCP connection request packets that are not admitted with a reset code bit ON.

37. The system according to claim 32, further comprising an assistant means for calculating whether TCP connection request packets of one or more of the customers can be admitted for one or more of the applications when the one or more customers already exceed the minimum TCP connection rate for the one or more applications.

38. The system according to claim 32, further comprising a manager means for communicating with an external means to receive revisions to one or more of the service level agreements, and to send collected statistics data to the external means.

39. The system according to claim 32, wherein the gatekeeper operates with a real-time admittance/rejection algorithm wherein the flow of the TCP connection request packets is regulated to the server farm on a per-packet basis.

40. The system according to claim 32, wherein the gatekeeper operates with a target-rate-based admittance/rejection algorithm wherein the flow of the TCP connection request packets is regulated to the server farm on a per-unit-time basis.