System and method for continuous measurement of transit latency in individual data switches and multi-device topologies

Patent metadata

Field	Value
Publication number	US-9847922-B2
Application number	US-201514662055-A
Country	US
Kind code	B2
Filing date	Mar 18, 2015
Priority date	Mar 18, 2015
Publication date	Dec 19, 2017
Grant date	Dec 19, 2017

Abstract

A system, method, and associated apparatus for continuous measurement of transit latency in individual data switches and multi-device topologies is disclosed, which enables an accurate measure of real-word latency on production networks by applying a time stamp to production network data on ingress to a network element and determining a traversal time by capturing the time stamp on egress from the network element. In one embodiment, time stamps applied to network data are removed on egress to avoid undesirable packet growth across multiple hops of the network. Timing information on data flow through the data/forwarding plane of each network node of a production network can be gathered without the use of test harnesses, injected network data, or dedicated ports.

First claim

What is claimed is:

1. A network element comprising:
a first port to receive production network data, wherein production network data is data that is used for purposes other than testing the performance of a network;
a second port to transmit the production network data;
a data plane coupled to the first port and the second port, the data plane to forward the production network data to the second port, wherein the data plane couples to a timing module to generate ingress timing data for the network data received at the first port and generate egress timing data for the network data transmitted via the second port, the data plane to further add a marker to the production network data that indicates to another network element to collect timing information about the production network data, and the data plane transmits the marked production network data out the second port; and
a control plane coupled to the data plane to receive the ingress timing data and the egress timing data from the data plane and to compute a network element latency for the production network data.

2. The network element as in claim 1 , wherein the data plane is further to append the ingress timing data to the production network data after the production network data is received from the first port.

3. The network element as in claim 2 , wherein the data plane is further to capture and remove the ingress timing data from the production network data before the second port transmits the production network data.

4. The network element as in claim 1 , wherein the control plane is to further to transmit the ingress and egress timing data to an external device for correlation.

5. The network element as in claim 4 , wherein the production network data includes a network data header and the control plane is to further to transmit the network data header to the external device with the ingress and egress timing data.

6. The network element as in claim 1 , wherein the control plane is to further to correlate the timing data to determine latency for the production network data within the network element and transmit the latency to an external device for consolidation.

7. The network element as in claim 1 , wherein the data plane is further to generate metadata to identify the production network data.

8. The network element as in claim 7 , wherein the metadata is further to identify a source and destination interface for the production network data, the source interface including the first port and the destination interface including the second port.

9. The network element as in claim 8 , wherein the data plane is further to append at least a portion of the metadata to the production network data after the production network data is received from the first port.

10. The network element as in claim 9 , wherein the data plane is further to remove the metadata from the production network data before the second port transmits the production network data and transmit the metadata, ingress timing data and egress timing data to the control plane.

11. A non-transitory machine-readable medium storing instructions to cause one or more processing units to perform operations to determine real-time latency within a network element, the operations comprising:
receiving production network data at a first port of the network element, wherein production network data is data that is used for purposes other than testing the performance of a network;
applying an ingress timestamp to the production network data;
forwarding the production network data to a second port of the network element;
after forwarding the production network data to the second port, capturing the ingress timestamp on the production network data and determining an egress timestamp for the network data, wherein a network element latency is computed from the ingress timestamp and egress timestamp;
removing the ingress timestamp from the production network data before transmitting the network data;
adding a marker to the production network data that indicates to another network element to collect timing information about the production network data; and
transmitting the marked production network data.

12. The medium as in claim 11 , the operations further comprising transmitting the ingress and egress timestamp to a data collector.

13. The medium as in claim 12 , the operations further comprising generating metadata for the production network data to uniquely identify the network data, wherein the metadata includes a first identifier to identify the first port and second identifier to identify the second port.

14. The medium as in claim 13 , the operations further comprising:
determining a first portion of the metadata after receiving the production network data at the first port;
appending at least the first portion of the metadata to the production network data before forwarding the production network data to the second port; and
before transmitting the network data, determining at least a second portion of the metadata and removing the first portion of the metadata from the production network data.

15. The medium as in claim 14 , the operations further comprising transmitting the metadata to the data collector with the ingress and egress timestamp.

16. A non-transitory machine-readable medium having executable instructions to cause one or more processing units to perform a method of determine a latency of a path between a first and second node in a multi-path network, where the path includes an intermediate node in-between the first and second nodes, the method comprising:
receiving intra-node timing data from each of the first, second, and intermediate nodes, the intra-node timing data including a latency time for production network data within that node, wherein the intra-node timing data includes an ingress timestamp and an egress timestamp and production network data is data that is used for purposes other than testing the performance of the network, wherein multiple elements of the production network data have differing characteristics between the multiple elements and differing elements having differing characteristics have differing latency over at least one link of the multi-path network;
determining link timing data including a traversal time for each link in a path between the first, second, and intermediate node;
determining a path latency for the production network data using the link timing data and the intra-node timing data;
wherein each of the first, second, and intermediate nodes determine intra-node timing data based on a selection of the same element of the production network data to sample, each selection performed using a sampling algorithm to select the element based on one or more characteristics of the element; and
wherein the sampling algorithm is configured to sample elements based on representative characteristics of the production network data.

17. The medium as in claim 16 , the method further comprising computing intra-node latency for each node based on the intra-node timing data for that node.

18. The medium as in claim 17 , wherein computing intra-node latency includes subtracting a time indicated by the ingress time from the time indicated by the egress time.

19. The medium as in claim 18 , wherein determining path latency for the network data includes summing the link timing data for each link in the path and the intra node latency for each node in the path.

20. The medium as in claim 18 , wherein the ingress time and egress time are absolute timestamps and det

Assignees

• Arista Networks Inc

Inventors

• Peach John
• Chandra Subhachandra

Classifications

• H04L43/0858 Primary

  One way delays · CPC title

• H04L43/106

  using time related information in packets, e.g. by adding timestamps · CPC title