Bonding performance

Bonding performance

Post by Lee Ziegenha » Sat, 19 Nov 1994 01:03:02



I have been running IP over ISDN for a couple of years using a single B
channel at 64 Kbps.  There is a Cisco router at each end, and my ping
response is about 30 ms.

Yesterday I switched to a pair of Motorola/UDS TA220k TAs and enabled
bonding (mode 1) using 2 B channels.  The throughput improved as
expected, but my ping response time degraded to 100 ms.

I'm a bit surprised by the increased delay.  I suppose the delay is
being introduced by buffering in the TA, but an increase of 70 ms seems
excessive for lines that should have essentially the same end-to-end
delay.  Both ISDN lines are served by the same CO, so it seems unlikely
the additionaly delay is being introduced by the phone company.  The
switch is a 5E9, with one line configured as NI-1 and the other as 5E
custom.

Is this reasonable?  Is there a significant difference between vendors
with respect to bonding performance?  Has anyone done any bonding
performance comparisons on various TAs?

Thanks!

 
 
 

Bonding performance

Post by sohl,william » Tue, 22 Nov 1994 10:33:34



>I have been running IP over ISDN for a couple of years using a single B
>channel at 64 Kbps.  There is a Cisco router at each end, and my ping
>response is about 30 ms.
>Yesterday I switched to a pair of Motorola/UDS TA220k TAs and enabled
>bonding (mode 1) using 2 B channels.  The throughput improved as
>expected, but my ping response time degraded to 100 ms.
>I'm a bit surprised by the increased delay.  I suppose the delay is
>being introduced by buffering in the TA, but an increase of 70 ms seems
>excessive for lines that should have essentially the same end-to-end
>delay.  Both ISDN lines are served by the same CO, so it seems unlikely
>the additionaly delay is being introduced by the phone company.  The
>switch is a 5E9, with one line configured as NI-1 and the other as 5E
>custom.
>Is this reasonable?  Is there a significant difference between vendors
>with respect to bonding performance?  Has anyone done any bonding
>performance comparisons on various TAs?

Consider the following with regard to additional buffering
as a reesult of the bonding now going on.  When you ping
someplace you are encountering 4 additional points where
additinal buffering can add delay:

   1. The outgoing bonding process in your terminal equipment,
   2. The receive bonding process at the distant terminal,
   3. The outgoing ping message in the distant terminal equipment, and
   4. The recieve ping in your terminal equipment.

Now I'm not an expert as to how much delay may be introduced at
any one of those points, but I think it would account for a fair amount
of the additional delay.

Maybe someone more familiar with the bonding circuitry can shed
additional details.

Bill Sohl, Bellcore NISDN Hotline Technical Consultant (1-800-992-ISDN)
-----------------------------------------------------------------------
Bill Sohl (K2UNK) BELLCORE (Bell Communications Research, Inc.)
Morristown, NJ          

--
Bill Sohl, Bellcore NISDN Hotline Technical Consultant (1-800-992-ISDN)
-----------------------------------------------------------------------
Bill Sohl (K2UNK) BELLCORE (Bell Communications Research, Inc.)
Morristown, NJ             email via UUCP      bcr!cc!whs70


 
 
 

Bonding performance

Post by Hascall Shar » Wed, 23 Nov 1994 05:28:15


Quote:>Is this reasonable?  Is there a significant difference between
vendors
>with respect to bonding performance?  Has anyone done any
bonding
>performance comparisons on various TAs?

If you were placing a call outside your LATA then the increased
delay could be traceable to variable routes taken in the
interexchange network.  If you are within your local area, then
the delays are most likely introduced in the TA equipment.
There are various ways of implementing BONDING that affect
performance, but still maintain interoperability with the
specification.  I have seen implementations that introduce
as little as 2 msec delay (Teleos) and others that inject up to
100 msec.  This parameter is measured as the delay injected into
the data path at the transmit and receive side in addition to
the transit network delay.  I do not know of any third party
performance comparisons of BONDING equipment, but it is an
important issue, especially when running TCP/IP over it.  
Therefore, it is not correct to say that BONDING has poor
performance characteristics (in this regard), but that different
implementations will have different characteristics.
Good Luck.
Chip Sharp

--
---------------------------------------------------------------------
Hascall ("Chip") Sharp                        Teleos Communications, Inc.
Sr. System Engineer                     2 Meridian Road
                                        Eatontown, NJ  07724 USA

 
 
 

1. bonding..bonding..

A couple of bonding related questions...

Regarding bonding and I mean BONDING (not Multilink PPP):--

1)  does bonding rely on the B channels being on the same basic rate
ISDN connection or can you bond two B channels which are on different
connections?

2) does bonding require any special equipment/performance requirements
within the ISDN network or is it reliant entirely on Q.931 signalling
and end user equipment?

3) is there any requirement on maximum/minimum synchronisation and
transmission delay between the two B channels?

Many thanks for your thoughts.

RS.

2. Microsoft Money 2002 Deluxe

3. NEW ! Performance tool to improve your network performance

4. Ant on Himalaya

5. Performance Test Metrics for dns server performance.

6. GP1U52X IR recvr spurious pulses?

7. Bonded T1s from different carriers?

8. Completing Dynamite Dan - help required URGENTLY!!!

9. Problems with bonding 2 x SHDSL lines using MLPPP and CISCO1721

10. Bonding serial leased lines on 2600 router?

11. Bonding over mulitple cisco 3750?

12. Bonding 2 T1's

13. Bonding two T1s