- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Netflow
Hello,
I have 3 question regarding netflow/sflow and hope in this room there are anybody who can help me.
1. As i know fortinet have netflow and sflow feature, which one is more recommended to use?
2. If we apply netflow/sflow in outside interface it's true that captured traffic only showing conversation from NATed Public IP to the internet?
3. If we apply netflow/sflow in inside interface it's true that captured traffic only showing conversation from private ip to the internet?
- Labels:
-
FortiGate
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
Thank you for using the Community Forum. I will seek to get you an answer or help. We will reply to this thread with an update as soon as possible.
In the meantime, perhaps this document will help you decide which type is best for your network to use: https://docs.fortinet.com/document/fortigate/7.2.4/hardware-acceleration/631057/sflow-and-netflow-an...
Thanks,
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
I think these links will have most of the answers for your questions.
https://docs.fortinet.com/document/fortigate/7.4.0/administration-guide/998643/netflow
https://docs.fortinet.com/document/fortigate/7.4.0/administration-guide/505119/sflow
I am not sure which is better, probably sFlow as it has more information about the traffic, but personally I don't have experience with sFlow.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi, I did both, and from experience:
- sFlow can only send sampled information, so it provides statistical samples over period of time. It was done to save on agent resources back long time ago by HP, and it was the only protocol supported by Fortigate in the beginning. Today, I don't see value in using sFlow to save resources (unless you monitor 10s of Gigabit traffic, but I have no such set ups) as opposed to Netflow. Netflow can send info either on each packet passing the interface, or sampled over few packets (Netflow v9). So, Netflow gives you the choice - monitor each packet or a sample. And I haven't seen netflow daemon on FGT to load CPU more than 1-2% even on loaded firewalls. So, I'd recommend to use Netflow. Additionally, being invented by Cisco, Netflow has much more available collectors than sFlow to pick from.
- You will see IPs as they present in the packets passing the interface, just before the packet leaves the given interface. So, for WAN interface you will see IPs after Source/Hide NAT was done (i.e. you will see legal IPs).
- Not exactly - even in LAN each packet can have Source IP (private) and Destination IP (some host on the Internet).
Created on 02-18-2024 11:11 PM Edited on 02-18-2024 11:11 PM
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
I tried the sFlow and got the following result. If the sampler was set to 1/10, then when I sent 1GB of traffic, I saw 100MB in collector. If the sampler is 1/100, then I saw 10MB. If 1/1000, then 1MB. Is this the correct sFlow behavior?
Thanks.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Nobody knows. Misfortune.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Kuzma,
I recommend creating a new support forum thread to maximize the amount of views you get. Feel free to link to this related thread.
Kind regards,
Stephen
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi, sorry for the late reply.
The sampling is done over number of packets, not their rate or volume. That is - setting sample to 1/10 means sFlow will take into account every 10th packet. This means for the traffic with varying size your calculation wouldn't be 100% correct. It would be correct for uniform-sized traffic, say using iperf where each packet is more or less equal in size, 1/10 sample of 100 Mb will mean 1 Gb of real traffic.
In real life scenario, you can have 9 packets of 100 bytes and 10th packet (the one accounted for in the sample) of 1 Mbyte. But, given the high rates of traffic of today and nature of the traffic, the sampling will be actually quite close to the real traffic. To read more see https://sflow.org/packetSamplingBasics/