System and method for continuous in-line monitoring of data-center traffic
Disclosed is a method for continuous in-line monitoring of data-centric traffic to guarantee application performance. The method includes, in each switch of a plurality of switches in a network fabric, grouping all packets entering each respective switch of the plurality of switches based on either 5-tuple applications or EPG based applications, collecting performance statistics at every hop in the network fabric across all flows in-line in a flow table maintained in each respective switch and periodically exporting the performance statistics to analysis module.
1. A method comprising:
grouping all packets entering each switch of a plurality of switches in a network fabric, the grouping based on end point group (EPG) based applications;
collecting performance statistics at every hop in the network fabric across all flows in-line in a flow table maintained at each switch; and
periodically exporting the performance statistics to an analysis module.
2. The method of claim 1 , wherein the performance statistics include one or more of latency, jitter, or microburst statistics.
3. The method of claim 1 , further comprising:
computing a maximum latency and an average latency experienced by the packets at every hop.
4. The method of claim 1 , further comprising:
isolating one of the plurality of switches based on the performance statistics,
wherein,
the performance statistics identify a spike experienced by an individual frame, and
the spike identify a problem associated with the one of the plurality of switches.
5. The method of claim 1 , wherein the EPG based applications comprise one or more of Src EPG, Dst EPG, or Protocol.
6. The method of claim 1 , further comprising:
sending the packets entering each switch to a local central processing unit at each switch, and
analyzing each of the flows via a light weight flow analyzer module at each switch.
7. The method of claim 6 , further comprising:
storing, via the light weight flow analyzer module at each switch, traffic information in a time series database.
8. A system comprising:
a processor; and
a computer-readable storage device storing instructions which, when executed by the processor, cause the processor to perform operations comprising:
grouping all packets entering each switch of a plurality of switches in a network fabric, the grouping based on end point group (EPG) based applications;
collecting performance statistics at every hop in the network fabric across all flows in-line in a flow table maintained at each switch; and
periodically exporting the performance statistics to an analysis module.
9. The system of claim 8 , wherein the performance statistics comprise one or more of latency, jitter, or microburst statistics.
10. The system of claim 8 , wherein the instructions, when executed by the processor, cause the processor to perform further operations comprising:
computing a maximum latency and an average latency experienced by the packets at every hop.
11. The system of claim 8 , wherein the instructions, when executed by the processor, cause the processor to perform further operations comprising:
isolating one of the plurality of switches based on the performance statistics,
wherein,
the performance statistics identify a spike experienced by an individual frame, and
the spike identify a problem associated with the one of the plurality of switches.
12. The system of claim 8 , wherein the EPG based applications comprise one or more of Src EPG, Dst EPG, or Protocol.
13. The system of claim 8 , further comprising:
a local central processing unit at each switch, the local central processing unit configured to receive the packets entering a respective one of the plurality of switches; and
a light weight flow analyzer module at each switch, the light weight flow analyzer module configured to analyze a respective one of the flows.
14. The system of claim 13 , wherein the light weight flow analyzer module is configured to store traffic information in a time series database.
15. A non-transitory computer-readable storage device storing instructions which, when executed by a processor, cause the processor to perform operations comprising:
grouping all packets entering each switch of a plurality of switches in a network fabric, the grouping based on end point group (EPG) based applications;
collecting performance statistics at every hop in the network fabric across all flows in-line in a flow table maintained at each switch; and
periodically exporting the performance statistics to an analysis module.
16. The non-transitory computer-readable storage device of claim 15 , wherein the performance statistics comprise one or more of latency, jitter, or microburst statistics.
17. The non-transitory computer-readable storage device of claim 15 , wherein the instructions, when executed by the processor, cause the processor to perform further operations comprising:
computing a maximum latency and an average latency experienced by the packets at every hop.
18. The non-transitory computer-readable storage device of claim 15 , wherein the instructions, when executed by the processor, cause the processor to perform further operations comprising:
isolating one of the plurality of switches based on the performance statistics,
wherein,
the performance statistics identify a spike experienced by an individual frame, and
the spike identify a problem associated with the one of the plurality of switches.
19. The method of claim 1 , wherein the performance statistics include latency, jitter, and microburst statistics.
20. The system of claim 8 , wherein the performance statistics include latency, jitter, and microburst statistics.