Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Does Ten-Gigabit Ethernet need fault tolerance? (nonredundant NICs)




Joe,

You wrote "There is no problem with parts of the segment having non-redundant
NICs,".  With a full duplex, point to point link, there are only two interfaces, one
at each end.  It was my understanding that 10GbE was to be a full duplex, point to
point implementation only.  Within a data switch, at layer 2 or layer 3, multiple
links to multiple systems can be implemented, creating virtual segments; but each
10GbE link is itself, full duplex, point to point.  It is the single 10GbE link that
I am writing about when I refer to fault tolerance.  I am not writing about fault
allowance at the virtual segment level, as in the case of allowing for nodal data
switch failures within the virtual segment.  Are we talking about two different
things?

Thank you,
Roy Bynum
MCI WorldCom

Joe Gwinn wrote:

> Roy,
>
> At 9:12 PM 99/7/24, Roy Bynum wrote:
> >
> >Does RTFC allow a minimally trained individual to simply plug two fiber T/R
> >pairs into the 10GbE interface to implement fault tolerance and if a second T/R
> >pair, parallel to the first, is not plugged in the fault tolerance is not
> >implemented?  This will be the simplest and most common implementation process.
>
> Yes, this will work, by design.  The rostering algorithm will just treat
> the missing path as broken, and press on.  There is no problem with parts
> of the segment having non-redundant NICs, although those NICs will be cut
> out of the segment if those NICs or their links fail.
>
> Joe
>
> >Joe Gwinn wrote:
> >
> >> Those people reading the RTFC Principles of Operation may wish to keep a
> >> few points in mind:
> >>
> >> RTFC assumes smart NICs and dumb hubs.  However, for 10GbE we probably
> >> would want dumber NICs and smarter hubs, both to allow a single NIC design
> >> to be used for both FT and non-FT systems, and to allow the hub to assign
> >> NodeIDs, rather than having to depend on humans to manually set these in
> >> strictly increasing numerical order on the NIC cards.  Letting users out of
> >> handling such a critical step is probably essential.  So, don't feel that
> >> the partitioning set forth in the RTFC Principles of Operation is in any
> >> way required; I would expect GbE to repartition the rostering algorithm
> >> between NICs and hubs.
> >>
> >> The NICs could be designed such that up to four standard-issue NIC cards
> >> could be installed side-by-side in a single PCI bus, strapped to
> >> communicate with one another via the PCI bus they share, together
> >> implementing a RTFC quad node.  Or, one could implement dual-ring NIC
> >> cards, with two dual cards side by side to implement a quad redundant
> >> system.  My guess would be that most commercial users will find that dual
> >> suffices, while there are military applications for both dual and quad
> >> redundant systems.  Basically, dual redundant suffices for ordinary fault
> >> tolerance, but quad redundant is often needed for battle damage tolerance.
> >>
> >> Joe
> >>
> >> The basic technical document, the RTFC Principles of Operation, is on the
> >> GbE website as:
> >>
> >> http://grouper.ieee.org/
> >> groups/802/3/10G_study/public/email_attach/gwinn_1_0699.pdf
> >>
> >> http://grouper.ieee.org/groups/802/3/10G_study/public/email_attach/gwinn_2_0
> >> 699.pdf
> >>
> >> The first document is the text, and the second contains the drawings.