Performance advantage for loopback interface?

Performance advantage for loopback interface?

Post by Alan L. Stang » Tue, 23 Feb 1999 04:00:00




> We have a config file that contains an IP address.  In this configuration the
> application will be using the local machine, so I'd like to use the loopback
> interface IP, but another engineer is claiming that, in general, there is a
> performance advantage, for any application, to using the IP associated with
> the 1 (and only) hme interface.

> Which interface is going to be the better choice, performance-wise?

This sort of thing is short circuited in the kernel.  A quick netperf run will
demonstrate this...

I don't have an idle enough Sparc system, so here's a dual Pentium II running
Solaris 2.7:

pro6$ uname -a
SunOS pro6 5.7 Beta_Update i86pc i386 i86pc

/usr/local/netperf/netperf
TCP STREAM TEST to localhost
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

  8192   8192   8192    10.08     368.65

pro6$ !! -H pro6
/usr/local/netperf/netperf -H pro6.bnl.gov
TCP STREAM TEST to pro6.bnl.gov
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

  8192   8192   8192    10.00     367.23

We haven't had time up upgrade to 2.7 FCS...Similar results are achieved with
2.6

In any case, use localhost (127.0.0.1) if that's what you really mean.

--

 
 
 

Performance advantage for loopback interface?

Post by Rick Jon » Thu, 25 Feb 1999 04:00:00


While everyone is in a "my loopback is bigger than your's" mood, I'll
point-out that interface MTU's can mask a range of path-length
sins. In general, it is good to include a -v2 to the output of netperf
to show the TCP MSS of the connection. That will help when folks
(inevitably) start comparing loopback figures between OSes running on
the same hardware.

Also, the netperf TCP_RR test with a single-byte request/response is a
good measure that remains blithely oblivious to MTU/MSS values.

rick "still behind on the netperf DB chores :(" jones

--
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to email, or post, but please do not do both...
my email address is raj in the cup.hp.com domain...

 
 
 

1. Performance advantages of VLB SCSI?

There was a thread in pc.hardware a while back where it was generally
recognised that under DOS there is *no* speed advantage of VLB SCSI over
ISA SCSI.

Does anyone know whether there is any advantage of using VLB SCSI under
Linux?  If not, is a SCSI disk going to be as fast as a VLB IDE drive?
I know that SCSI loads the processor less, but how do they compare in
Bytes per sec?

Your help much appreciated,

Tim

--
Tim Monks ________ Vision, Speech and Signal Proc.,  Southampton University _

2. The Vacation Program

3. SIOCGIFFLAGS fails loopback interface

4. problem with ppp & dialup

5. "Bandwidth" of loopback interface?

6. Linux Fdisk and VFAT32

7. Snoop - loopback interface?

8. Bus Error running Netscape?

9. Network Loopback Interface

10. volatile loopback interface

11. new kernel loses lo loopback interface

12. snoop on loopback interfaces?

13. activate software loopback interface