Switch adapter testing

• US 6,601,195 B1
• Filed: 09/09/1999
• Issued: 07/29/2003
• Est. Priority Date: 09/09/1999
• Status: Expired due to Fees

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: International Business Machines Corporation
• Inventors: Sostheim, Tal, Chirashnya, Igor
• Primary Examiner(s): Beausoliel, Robert
• Assistant Examiner(s): CHU, GABRIEL L
• Application Number:
  US09/393,582
• Time in Patent Office: 1,419 Days
• Field of Search: 714/43, 714/4, 714/48, 714/56, 370/230, 370/235
• US Class Current: 714/43
• CPC Class Codes: G06F 11/26 Functional testing

Abstract

In a computer network system that includes a multiplicity of nodes interconnected by a network of switches, wherein the nodes are linked to the network by respective data link adapters, a method for testing the adapters. One of the nodes is selected to serve as a destination node, and data are conveyed at a controlled rate from a plurality of the nodes, other than the destination node, through the respective adapters to the destination node. An error is detected in the data conveyed from one of the nodes so as to identify a fault in the adapter of that node.

Claims

1. In a computer network system that includes a multiplicity of nodes interconnected by a network of switches, wherein the nodes are linked to the network by respective data link adapters, a method for testing the adapters, comprising:

selecting one of the nodes to serve as a destination node;

conveying data at a controlled rate from a plurality of the nodes, other than the destination node, through the respective adapters to the destination node; and

detecting an error in the data conveyed from one of the nodes so as to identify a fault in the adapter of that node,

wherein conveying the data at the controlled rate comprises transmitting data from the plurality of the nodes at a substantially maximal transmission rate that the transmitting nodes can achieve.

2. A method according to claim 1, wherein transmitting the data comprises sending data from the plurality of the nodes at an aggregate rate greater than a data throughput capacity of one of the switches in the network through which the data are conveyed.

3. A method according to claim 2, wherein sending the data comprises sending data packets, which are queued in the data link adapters of the nodes sending the packets when the aggregate rate is greater than the data throughput capacity of the one of the switches.

4. In a computer network system that includes a multiplicity of nodes interconnected by a network of switches, wherein the nodes are linked to the network by respective data link adapters, a method for testing the adapters, comprising:

selecting one of the nodes to serve as a destination node;

conveying data at a controlled rate from a plurality of the nodes, other than the destination node, through the respective adapters to the destination node; and

detecting an error in the data conveyed from one of the nodes so as to identify a fault in the adapter of that node,

wherein conveying the data comprises conveying data packets, and wherein detecting the error comprises detecting a corrupted packet at the destination node.

5. A method according to claim 4, wherein conveying the data packets comprises conveying packets including redundant identification information regarding a source node sending the packets, whereby the source node is identified at the destination node despite the corruption of the packet.

6. In a computer network system that includes a multiplicity of nodes interconnected by a network of switches, wherein the nodes are linked to the network by respective data link adapters, a method for testing the adapters, comprising:

selecting one of the nodes to serve as a destination node;

conveying data at a controlled rate from a plurality of the nodes, other than the destination node, through the respective adapters to the destination node; and

detecting an error in the data conveyed from one of the nodes so as to identify a fault in the adapter of that node,

wherein conveying the data comprises conveying data packets, and wherein detecting the error comprises finding a discrepancy between a number of packets sent by one of the plurality of the nodes to a number of packets received therefrom by the destination node.

7. In a computer network system that includes a multiplicity of nodes interconnected by a network of switches, wherein the nodes are linked to the network by respective data link adapters, a method for testing the adapters, comprising:

selecting one of the nodes to serve as a destination node;

conveying data at a controlled rate from a plurality of the nodes, other than the destination node, through the respective adapters to the destination node; and

detecting an error in the data conveyed from one of the nodes so as to identify a fault in the adapter of that node,

wherein conveying the data comprises selecting groups of a predetermined number of the nodes and sending data from the nodes in a given one of the groups simultaneously through a selected one of the switches to the destination node.

8. A method according to claim 7, wherein the switches have multiple ports, and wherein sending the data comprises sending data simultaneously from each of the nodes in the given group through a respective one of the ports of the selected switch.

9. A method according to claim 7, wherein the switches have multiple ports, and wherein sending the data comprises sending data from one of the nodes in the given group through one of the ports of the selected switch while sending data from the other nodes in the given group through another one of the ports of the selected switch.

10. In a computer network system that includes a multiplicity of nodes interconnected by a network of switches, wherein the nodes are linked to the network by respective data link adapters, a method for testing the adapters, comprising:

selecting one of the nodes to serve as a destination node;

conveying data at a controlled rate from a plurality of the nodes, other than the destination node, through the respective adapters to the destination node; and

detecting an error in the data conveyed from one of the nodes so as to identify a fault in the adapter of that node,

wherein conveying the data comprises sending data packets, which in normal operation of the system are routed between any pair of the nodes over a plurality of different routes in alternation, and wherein sending the data packets comprises routing substantially all of the packets conveyed from at least one of the plurality of nodes to the destination node over at least one respectively-assigned route.

11. A method according to claim 10, wherein each of the data link adapters routes data from the respective node through the network in accordance with a routing table stored in a memory, and wherein routing substantially all of the packets comprises downloading a test routing table containing the respectively-assigned route to the adapter of the at least one of the plurality of nodes.

12. A manageable computer network system, comprising:

a multiplicity of nodes, including a management node;

a network of switches, each switch having multiple ports; and

a multiplicity of data link adapters, each linking a respective one of the nodes to one of the ports of one of the switches,

wherein the management node carries out a diagnostic test of the switch adapters by selecting one of the nodes to serve as a destination node and commanding a plurality of the other nodes to send data at a controlled rate through the respective adapters to the destination node, and wherein the destination node detects an error in the data conveyed from one of the sending nodes so as to identify a fault in the adapter of that node.

13. A system according to claim 12, wherein the data comprise data packets, and wherein the data link adapters comprise respective queues, in which the data packets accumulate during the diagnostic test.

14. A system according to claim 13, wherein the error detected by the destination node comprises corruption of a packet.

15. A system according to claim 12, wherein the plurality of the other nodes commanded to send data comprises a group of a predetermined number of the nodes, which send data simultaneously through a single switch to the destination node.

16. A computer software product for testing data link adapters respectively linking a multiplicity of processor nodes, one of which nodes is designated a primary node, to switches in a computer network system, the product comprising a computer-readable medium, in which program code is stored, which code, when read by the primary node, causes the primary node to select one of the nodes to serve as a destination node, and to command a plurality of the nodes, other than the destination node, to convey data through the respective adapters to the destination node and to detect an error in the data conveyed from one of the nodes so as to identify a fault in the adapter of that one of the nodes.