Method for determining a networking capability index for each of a plurality of networked computers and load balancing the computer network using the networking capability indices

• US 5,819,045 A
• Filed: 12/29/1995
• Issued: 10/06/1998
• Est. Priority Date: 12/29/1995
• Status: Expired due to Fees

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Rungta, Sanjay, Raman, Anant
• Primary Examiner(s): Rinehart, Mark H.
• Application Number:
  US08/581,041
• Time in Patent Office: 1,012 Days
• Field of Search: 395/200.01, 395/200.05, 395/200.09, 395/200.11, 395/200.15, 395/200.31, 395/200.38, 395/200.47, 395/200.63, 395/200.56 , 395/200.54, 395/200.71, 395/675, 395/674, 364/281, 364/281.6, 364/975.5, 370/232, 370/233, 370/234
• US Class Current: 709/226
• CPC Class Codes: H04L 43/00 Arrangements for monitoring or testing data switching networks
• CPC Class Codes: H04L 43/0888 Throughput
• CPC Class Codes: H04L 47/10 Flow control; Congestion control
• CPC Class Codes: H04L 47/125 by balancing the load, e.g. traffic engineering
• CPC Class Codes: H04L 47/13 in a LAN segment, e.g. ring or bus

Abstract

A method for determining a networking capability index (NCI), which is representative of the load a computer applies to a computer network. The method includes selecting a group of packet sizes, measuring the capability of the computer to transmit packets of each of the selected sizes from the process application to the input/output kernel as a function of time, and determining a probability distribution for the selected packet sizes, wherein the probability distribution for a selected packet size is representative of the percentage of total traffic corresponding to the selected packet size. The method then determines the NCI by summing the product of each selected packet size times each corresponding throughput value times each corresponding probability distribution value. Also a method for load balancing a plurality of networked computers arranged in segments by determining the NCI for each of the networked computers, and selectively connecting the computers to have approximately equal combined NCI values for all segments.

Claims

1. A method for determining a value representative of the load a computer applies to a computer network, said method comprising the steps of:

selecting a group of packet sizes;

measuring the capability of the computer to transmit packets of each selected packet size from the process application to the input/output kernel as a function of time;

determining a probability distribution for the selected packet sizes, said probability distribution representative of a percentage of total traffic from the computer corresponding to each of the selected packet sizes; and

determining a networking capability index (NCI) for the computer based at least in part on said measured capabilities as a function of time for the selected packet sizes, said probability distribution for the selected packet sizes, and the selected packet sizes.

2. The method of claim 1 wherein the step of determining the networking capability index for the computer comprises the step of calculating:

##EQU3## wherein N is a number of selected packet sizes, P_i is a traffic probability distribution value for the ith selected packet size, X_i is the measured packet rate value for the ith selected packet size, and S_i is the packet size for the ith selected packet size.

3. The method of claim 2 wherein the probability distribution values are determined by:

measuring the distribution of traffic corresponding to the selected packet sizes within the computer network.

4. The method of claim 1 wherein the probability distribution values are determined by:

setting the traffic probability distribution values to a set of default values.

5. The method of claim 4 wherein five packet sizes are selected, having packet sizes of 96, 256, 512, 1024 and 1460 bytes per packet, respectively, and wherein the corresponding probability distribution values are about 0.5, 0.3, 0.02, 0.05, and 0.13, respectively.

6. A method for load balancing a plurality of networked computers arrangeable in separate segments, said method comprising the steps of:

determining a networking capability index (NCI) for each of said computers said NCI representative of a maximum expected throughput capability for a computer for transmitting from an application process running within the computer to an input/output kernel also running within the computer; and

selectively connecting said computers within said segments yielding segments having approximately equal combined NCIs for all computers connected thereto.

7. The method of claim 6 wherein the step of determining the NCI comprises the steps of:

selecting a group of packet sizes;

measuring the capability of the computer to transmit packets of each selected packet size from the process application to the input/output kernel as a function of time;

determining a probability distribution for the selected packet sizes, said probability distribution representative of a percentage of total traffic from the computer corresponding to each of the selected packet sizes; and

determining the networking capability index (NCI) for the computer based at least in part on said measured capabilities as a function of time for the selected packet sizes, said probability distribution for the selected packet sizes, and the selected packet sizes.

8. The method of claim 6 wherein the step of determining the NCI for the computer comprises the step of calculating:

##EQU4## wherein N is a number of selected packet sizes, P_i is a traffic probability distribution value for the ith selected packet size, X_i is the measured packet rate value for the ith selected packet size, and S_i is the packet size for the ith selected packet size.

9. The method of claim 7 wherein the probability distribution values are determined by:

measuring the distribution of traffic corresponding to the selected packet sizes within the computer network to be load balanced.

10. The method of claim 7 wherein the probability distribution values are determined by:

setting the traffic probability distribution values to a set of default values.

11. The method of claim 10 wherein five packet sizes are selected, having packet sizes of 96, 256, 512, 1024 and 1460 bytes per packet, respectively, and wherein the corresponding probability distribution values are about 0.5, 0.3, 0.02, 0.05, and 0.13, respectively.

12. A method for load balancing a computer network containing a plurality of multiport bridges and a group of computers, said method comprising the steps of:

determining a networking capability index (NCI) for each of a plurality of types of computers to be connected to the computer network, said NCI representative of a maximum expected throughput capability for a type of computer for transmitting from an application process running within the type of computer to an input/output kernel also running within the type of computer;

determining the number of ports of the plurality of multiport bridges;

determining a target group NCI for all of a plurality of subnetworks connected to the multiport bridges;

dividing the group of computers into subnetworks with the computers divided to achieve a group NCI of each subnetwork substantially equal to the target group NCI; and

determining whether the maximum number of workstations over all of the subnetworks is acceptable and, if not, decreasing the target group NCI until the maximum number of workstations over all of the subnetworks is acceptable.

13. The method of claim 12 wherein the step of determining whether the maximum number of workstations over all of the subnetworks is acceptable is achieved by determining whether there is sufficient hardware to connect all of the computers to the network.

14. The method of claim 12 wherein the step of determining whether the maximum number of workstations over all of the subnetworks is acceptable is achieved by determining whether there is sufficient hardware to connect all of the computers while also allowing for any expected future expansion requiring additional computers.