My group is seriously looking at the issues surrounding load
balancing, fault-tolerance and failover with Linux based server farms.
These features along with filesystem issues that Stephen Tweedie is
addressing still separate Linux from the 'big iron' server market.
With this interest I have been watching the work that Steven Clarke
has been doing on the IP port forwarding variant of the masquerade
code.
A cursory analysis of this seems to indicate that this strategy only
works when the machines being forwarded to are sitting on a private
subnet with the masquerading machine servering as the router. More
useful to our group is a packet rewriting solution where the load
balancer as well as other machines sit on a common network.
I have been experimenting with rinetd which is a very nice userland
solution that can redirect connections in this type of environment.
The difficulty with this is the fact that the originating IP address
is lost in the forwarding process and the overhead experienced with
the necessary transitions/copies between kernel and user space.
I would be interested in any commentary on this issue. Any pointers
to related work would also be of interest.
Have a pleasant end of the week.
Greg
As always,
Dr. G.W. Wettstein Enjellic Systems Development - Specialists in
4206 N. 19th Ave. intranet based enterprise information solutions.
Fargo, ND 58102 WWW: http://www.enjellic.com
Phone: 701-281-1686 EMAIL: greg@wind.enjellic.com
------------------------------------------------------------------------------
"It's a bug, but it has nothing to do with eval. You can get the same
effect with @tmp = *abcd. The problem is that the assignment of undef to
$d is not clobbering $d sufficiently to keep it from reconstituting itself
magically when evaluated in a boolean context."
-- Larry Wall
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.rutgers.edu