[Oisf-users] EXTERNAL: Re: autofp vs workers - updated comparison?

[Eric Leblond]

Hello,
Le 22 juil. 2015 1:59 AM, "Cooper F. Nelson" <cnelson at ucsd.edu> a écrit :
>
> -----BEGIN PGP SIGNED MESSAGE----- 
> Hash: SHA1 
>
> Details are in Documentation/networking/packet_mmap.txt 
>
> > In the AF_PACKET fanout mode, packet reception can be load balanced among 
> > processes. This also works in combination with mmap(2) on packet sockets. 
> > 
> > Currently implemented fanout policies are: 
> > 
> >   - PACKET_FANOUT_HASH: schedule to socket by skb's packet hash 
> >   - PACKET_FANOUT_LB: schedule to socket by round-robin 
> >   - PACKET_FANOUT_CPU: schedule to socket by CPU packet arrives on 
> >   - PACKET_FANOUT_RND: schedule to socket by random selection 
> >   - PACKET_FANOUT_ROLLOVER: if one socket is full, rollover to another 
> >   - PACKET_FANOUT_QM: schedule to socket by skbs recorded queue_mapping 
>
> - From what I can tell, if a packet socket is full you will still drop 
> packets from the last flow that was pinned to that socket.  

No there is no flow based load balancing in rolloverode. So no drop on the flow.

However, new 
> flows will not be assigned to that socket, rather they are "rolled over" 
> to another socket. 

Yes but that is not the interesting part. What is more interesting is the 'rollover' socket option: You keep your usual load balancing and if this is activated then the kernel will send packet to a new socket.

>
> - -Coop 
>
> On 7/21/2015 3:25 PM, Rasmor, Zachary R wrote: 
> > Hello Eric, 
> > 
> > Thanks for the reply, this definitely sounds interesting. One 
> > question...My understanding was that it is necessary for all packets 
> > belonging to a given flow to be processed by the same worker thread. 
> > Am I mistaken? 

No there is no need to have a perfect pinning but pinning will help a lot on performance side.

And if my understanding is correct, how does this work 
> > with the rollover option when a packet is sent to a different 
> > socket? 
> > 
> > Regards, Zach 
>
>
> - -- 
> Cooper Nelson 
> Network Security Analyst 
> UCSD ACT Security Team 
> cnelson at ucsd.edu x41042 
> -----BEGIN PGP SIGNATURE----- 
> Version: GnuPG v2.0.17 (MingW32) 
>
> iQEcBAEBAgAGBQJVrtzcAAoJEKIFRYQsa8FWJ0sH/jvV4DsiRE17eETP4dpZBKDQ 
> lXEQLZy8saUKMxxWANc3oT/NApMCHOwIp2ZWzESWIJNnYapaNfodwT3Kzhfkx3uY 
> 2FZ2NocoOELYPzDVuzDRTGhN6OcrcqLgerFxImVttnWd47NZ0sF61MRqxZ87LgfB 
> ShkpNQH5bY2H+kORSTOw8Dkr1f680kzC3pBfFZrf32dwRARYNMhUzdf3bIpJMcI9 
> p6wGSz8yz83A96pRZ95YOr5henAAURQV4sFpjRVgH6t7Ojo1pqSEH6RgJMpIl5yq 
> R8Yvc1B5n/ntrlvuwtaxUqu+j6bzkKldLzUflcMm/nPoPJX3SMEQKopaqGOyp+w= 
> =XWkM 
> -----END PGP SIGNATURE-----