[Oisf-users] [EXT] Re: Need help with Suricata conf

[Cloherty, Sean E]

Although this is and old post, I read that MPP had little/no performance impact in this thread on https://home.regit.org/2012/07/suricata-to-10gbps-and-beyond/  and these guys seem to be familiar with Suricata…. 😉


•  [cid:image001.jpg at 01D5CD3F.27981650] Victor Julien<http://blog.inliniac.net/> says:
2012/08/10 at 14:29 <https://home.regit.org/2012/07/suricata-to-10gbps-and-beyond/#comment-71478>
Does the max-pending-packets setting really affect performance here? In workers mode the number of packets simultaneously processed should be the same as the number of threads, plus a few for tunnel decoding and stream handling.
•  [cid:image002.jpg at 01D5CD3F.27981650] Regit<http://home.regit.org> says:
2012/08/12 at 08:20 <https://home.regit.org/2012/07/suricata-to-10gbps-and-beyond/#comment-71706>
You’re right indeed. I don’t remember seeing any difference here.
•  [cid:image003.png at 01D5CD3F.27981650] Peter Manev says:
2012/08/18 at 13:06 <https://home.regit.org/2012/07/suricata-to-10gbps-and-beyond/#comment-72161>
no effect in my tests…
From: Oisf-users <oisf-users-bounces at lists.openinfosecfoundation.org> On Behalf Of Michal Purzynski
Sent: Thursday, January 16, 2020 4:51 PM
To: Eric Urban <eurban at umn.edu>
Cc: oisf-users at lists.openinfosecfoundation.org
Subject: [EXT] Re: [Oisf-users] Need help with Suricata conf

We wrote guides a while back, that might help answering some of your questions. Let me address what I can, right here.

https://github.com/pevma/SEPTun
https://github.com/pevma/SEPTun-Mark-II


We are testing Suricata in af-packet IDS mode and we ran into a couple of issues with configuring.

We have various HW setups, and therefore Suricata runs in different runmodes (either workers or autofp) depending on a specific platform. Currently I'm trying to configure Suricata to get the best performance as possible, but some settings are ambiguous and even the documentation didn't help a lot.

The best performance setup (for the IDS) comes from the workers runmode.


Here are some of the things which I do not understand about configuration:

1) ring-size: <number of packets> - Ring size will be computed with respect to max_pending_packets and number of threads. You can set manually the ring size in number of packets by setting the following value. So as I understand this value defines a cache size of each thread when running in workers mode, but when running in autofp there may be different numbers of packet capture and packet processing threads. To which type of thread does the ring-size refer in autofp mode? And when this value is not set - what is the default value?

For workers mode - each threat will have its own ring, so also watch out for the memory usage. For autofp, no idea.


2) tpacket_v3 has such properties as block-size and block-timeout which look a bit complicated. What should I take into consideration when trying to tune those values?

Some good and proven in the battle numbers to start from are in above SepTun guides. Feel free to start somewhere from there and modify as you see fit.
As a general rule, I would first modify the ring-size before touching the block-size.

More details and diagrams here
https://www.kernel.org/doc/Documentation/networking/packet_mmap.txt


3) buffer-size: <number of bytes?> - what is this buffer?

It shouldn't be there - it's not used by the AF_Packet processing pipeline. Looks like something we could remove from this section of the config file.


4) max-pending-packets: <number of packets> - is this a number of packets which can be simultaneously processed by each of packet processing thread?

That's what I believe the documentation says, but the code makes me think it's the global per-Suricata instance limit of the maximum numbers of packets that can be queued by all of your threads, i.e.

ring_size * threads

Just make it generous.

Here is how I see it: say we set the ring-size to 100k packets, set m-p-p to 1k and run 8 packet processing threads. This setup means that each of 8 threads can analyze 1k packets at once, while other packets have to wait in its 100k packets buffer - is it correct? And again, if this setting is commented out - what is the default value?

default is 1024 (looking at the suricata.c)

I believe you should just make it larger than numbers of threads * packets per thread (so ring size).

Victor / Eric could you either confirm or correct me?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openinfosecfoundation.org/pipermail/oisf-users/attachments/20200117/a6794cc1/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.jpg
Type: image/jpeg
Size: 1443 bytes
Desc: image001.jpg
URL: <http://lists.openinfosecfoundation.org/pipermail/oisf-users/attachments/20200117/a6794cc1/attachment-0002.jpg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image002.jpg
Type: image/jpeg
Size: 1538 bytes
Desc: image002.jpg
URL: <http://lists.openinfosecfoundation.org/pipermail/oisf-users/attachments/20200117/a6794cc1/attachment-0003.jpg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image003.png
Type: image/png
Size: 3416 bytes
Desc: image003.png
URL: <http://lists.openinfosecfoundation.org/pipermail/oisf-users/attachments/20200117/a6794cc1/attachment-0001.png>