The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:1998-Sep> msg00247



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Colored thread loop prevention

  • From: Yoshihiro Ohba <yoshihiro.ohba@toshiba.co.jp>
  • Date: Fri, 18 Sep 1998 10:17:06 +0900
  • Cc: yoshihiro.ohba@toshiba.co.jp, erosen@cisco.com, mpls@external.cisco.com
  • X-SMAP-Received-From: outside

Indra,

 > Yes, I understand that the algorithm will work with the hop
 > count limit. What I'd like to know is if the algorithm would
 > still work without the hop count limit.
 > I think it may be possible if a node remembers the thread color that
 > it just withdrew. In that case, it would stall the withdrawn thread
 > when it comes back to the node.
 > Would this mechanism work? does it introduce another unwanted bahavior?

There are heuristics to reduce the possibility of the occurrence of
thread chasing without introducing a new unwanted bahavior.

The above method (remembering the withdrawn thread color at each node)
is one of them.  Another method is to delay withdrawing at the trigger
node.  

However, these are only heuristics and do not guarantee chasing-free.
For example, the heuristics depend on timers, a timer to forget the
thread color or a timer to start withdrawing.  If there is a highly
congested link, there is a (small) possiblity that the thread may
revisit a node after the timer expires.  I think the thread TTL
mechanism is needed to stop any kind of thread chasing.


Regards,
Ohba