Replies: 2 comments 2 replies
-
Hey @jhavinit, You would need to identify the bottleneck and there are a few options:
There is for sure still some capacity to optimise nfdump, but without a detailed analysis on the object itself, it's rather guessing. |
Beta Was this translation helpful? Give feedback.
-
More Workers only help, if compressing and storing away the data is the bottleneck. If v9 decoding is the bottleneck, it does not help. Decoding v9 is single threaded. The performance depends on how many elements are collected per flow. |
Beta Was this translation helpful? Give feedback.
-
Hi folks,
We are planning to export 10Million per second netflow from router. Can a single instance of nfcapd can capture 10million per seconds netflows data if so what configs we can do. Can we do some sort of load balancing to split load to virtual interfaces and create a listener using each nfcapd, if so how? So far I have tried in ubuntu 64 core machine with 300gb ram to achieve ~1.5million per second. Please advise. Thanks in advance.
Beta Was this translation helpful? Give feedback.
All reactions