RE: Does Ten-Gigabit Ethernet need fault tolerance?
Roy,
No (see your assertion below), it is going to add very little cost indeed
and we know this because we (several vendors) have already implemented
prestandard versions of link aggregation on low cost switches. P802.3ad
takes the results of this field experience (many thousands of systems
shipped) and provides a solution that is plug and play factory default
maximal link aggregation (at the switch vendor's option) across a very wide
range of system implementations. The cost is the memory cost of the
implementation, processing is close to negligible.
I suggest you look at P802.3ad, this will be sent for 802.3 working group
ballot very shortly.
I have put together a simulation of this (so you can see how big an
implementation would be, albeit a model implementation) with a number of
test cases, this is also publicly available.
802.3ad can hold one link as a standby for the other. The protocol
information exchanged and the rules for using it allow both systems to
select the same link to use (rather than each making a separate transmit
decision), however I believe that the most likely configuration uses
admission control at the edge of the switched network to ensure that the
capacity after failure is sufficient to handle the applied load.
Mick
-----Original Message-----
From: Roy Bynum [mailto:rabynum@xxxxxxxxxxx]
Sent: Monday, July 19, 1999 9:52 PM
To: Mick Seaman
Cc: stds-802-3-hssg@xxxxxxxx
Subject: Re: Does Ten-Gigabit Ethernet need fault tolerance?
Mick,
What you are describing is similar to the N+1 architecture of SONET/SDH
linear
spans, with the subscription rate monitored to the maximum of N. The
monitoring
of this is going to a "woolly booger". The system processing to control
this is
going to add a lot of cost to the systems. This is one reason that carriers
are
going to simpler architectures.
Will P802.3ad work with only two links and not loose any transport bandwidth
in
the event of the failure of one link? If it does not, it is not fault
tolerant,
it is fault allowant.
Thank you,
Roy Bynum
MCI WorldCom
Mick Seaman wrote:
> The traffic restoration time of P802.3ad will be as fast as:
>
> 1) Failure of an individual link (in a set of aggregated links) can be
> detected.
>
> 2) Each of the connected systems (routers, switches, or end stations) can
> move 'flows' from that link to another link.
>
> The second of these, moving flows, is entirely under the control of the
> system that is transmitting the frames (distribution of flows is not
> symmetric). That is to say each transmitter at each end of the link can
> independently start use another link other than the failed link.
>
> The constraints on moving the flows are therefore:
>
> (a) how fast the transmitting system can react to link failure
>
> (b) any deliberate allowance that that system may make to preserve
ordering
> because the remote system takes time to 'collect' the frames from each
> individual link - see the P802.3ad draft for a detailed discussion of
this,
> which makes it easier for links to be terminated on separate physical
pieces
> of hardware (cards etc.).
>
> Disregarding (b) for the present - see more below - recovery from a failed
> link takes place without any protocol being exchanged between the
> communicating systems so that (2) can be as short a time as the
implementor
> cares to make it. Given the simplicity of the decision to be made - it can
> be precomputed, if this link fails I do X - it is not unreasonable to do
> this at interrupt time in a software system. So it happens that failover
has
> been implemented in very low cost (as cheap as managed Ethernet gets)
volume
> products within 20 milliseconds. More attention to the subsystem design
> could naturally achieve failover rather more quickly though I don't know
> anyone is bothered to strive for sub 10 millisecond yet.
>
> How much allowance is made for (b) is an implementation issue, but can be
> minimized by receive side switch design. If this became a hot area I am
sure
> that vendors would start quoting it.In the IP world the right answer is to
> make no allowance at all.
>
> Going on from P802.3ad, a similar approach can be used to dramatically
> improve 802.1D spanning tree reconfiguration. Again a choice of the link
(or
> set of aggregated links) to failover to can be preselected for dual
> redundant tree topologies - failover can be initiated without the
> transmission of additional protocol messages - and again 20 milliseconds
> failover is achieved. Achieving these failover times for failure of any
> single link in the network requires careful consideration of the topolgy
to
> be used, however work underway in 802.1 is designed to free spanning tree
> reconfiguration times from all worse case end to end timer estimates, so
> where messages need to be exchanged the 'old' very slow spanning tree
> reconfiguration times no longer apply. This work is underway as P802.1w
> Rapid Reconfiguration. It will result in protocol enhancements that will
> plug and play with existing switches - though those switches won't achieve
> the rapid reconfiguration times (new switches attached to them may well
do,
> depends on topology). Rapid reconfiguration at this level protects a level
2
> network of switches.
>
> It should be noted that P802.3ad may well reconfigure from link failure on
> an aggregated link basis. This is the best approach for aggregated links
> since it does not disrupt topology at all (to first order, capacity
changes
> may well require traffic redistribution at layer 3 or more probably at
layer
> 2.5 (MPLS)). Above this spanning tree can be used for rapid local
> reconfiguration to avoid disrupting routing. Above that routing may have
to
> kick in to rediscover best paths etc. Again it is possible for routing to
> precompute alternates on a local level and failover to them rapidly as a
> first response to failure, good network topology is however required for
> this to be effective and is may be impossible on a long haul basis. MPLS
> failover may well be more challenging and require setting up the MPLS
label
> switching path again - I think this is the only recovery strategy where
> RSVP-like MPLS signalling is used. However these problems, which are very
> real, will not be helped at all by local link recovery except as in as
much
> as that completely masks failure as in the P802.3ad Link Aggregation case.
>
> Note that in the case where protocol messages are sent to aid rapid
> reconfiguration the design challenge is to minimize the number of messages
> sent to achieve a given timing. Almost any protocol (well any correct
> protocol of which there are rather fewer) can achieve rapid
reconfiguration
> if the designer is allowed to send many messages, just depends on how much
> processing is tolerated.
>
> All of the above, IMHO, indicates to me that Ethernet should stick to
> providing timely indication of link failure, and making sure that both
> communicating systems see the failure. Timely indication of link recovery
is
> also desirable but since this will undoubtedly be confirmed by a protocol
> excahnge before the link is brought into service, it is less critical.
>
> Mick
>
> -----Original Message-----
> From: owner-stds-802-3-hssg@xxxxxxxxxxxxxxxxxx
> [mailto:owner-stds-802-3-hssg@xxxxxxxxxxxxxxxxxx]On Behalf Of Roy Bynum
> Sent: Sunday, July 18, 1999 5:33 PM
> To: mick@xxxxxxxxxxx
> Cc: stds-802-3-hssg@xxxxxxxx
> Subject: Re: Does Ten-Gigabit Ethernet need fault tolerance?
>
> Mick,
>
> When implemented how fast is the FT traffic restoration of P802.3ad
supposed
> to
> work? From that restoration time, calculate how much data got lost. One
of
> the
> major features of tightly coupled error detection is that the traffic
> restoration times are greatly reduced. At present, the fastest that L3
> Matrix
> (I can not use any vendor implementation names.), MPLS, or L1/REI/L3 is
> proposed
> is 10 times slower than 802.3 that is directly mapped into SONET. (There
> are
> several transport vendors that are doing this.) There are other issues
with
> fiber maintenance that have not been issues before because the amount of
> aggregated data and the "round the clock" that 10GbE will likely see.
This
> makes the FT as much a function of the PHY as anywhere else.
>
> Thank you,
> Roy Bynum
> MCI WorldCom
>
> "Mick Sea,man" wrote:
>
> > What needs to be built in is the detection of failure. What we don't
need
> to
> > do is to build everything into the MAC. I suggest you look at the fault
> > tolerant capabilities provided by P802.3ad and the work on Rapid
> > Reconfiguration starting in 802.1.
> >
> > Both these (will) provide a degree of fault tolerance based on using
> > protocols that are independent of MAC details to allow network nodes to
> > precalculate their response to a low level indication of failure. There
is
> > really no need to build these protocols into the MAC.
> >
> > Mick
> >
> > -----Original Message-----
> > From: owner-stds-802-3-hssg@xxxxxxxxxxxxxxxxxx
> > [mailto:owner-stds-802-3-hssg@xxxxxxxxxxxxxxxxxx]On Behalf Of Joe Gwinn
> > Sent: Friday, July 16, 1999 3:15 PM
> > To: stds-802-3-hssg@xxxxxxxx
> > Subject: Does Ten-Gigabit Ethernet need fault tolerance?
> >
> > The purpose of this note is to present a case for inclusion of fault
> > tolerance in 10GbE, and to offer a suitable proven technology for
> > consideration. However, no salesman will call.
> >
> > Why Fault Tolerance? Ten-Gigabit Ethernet is going to be a relatively
> > expensive, high-performance technology intended for major backbones,
> > perhaps even nibbling at the bottom end of the wide-area network (WAN)
> > market. In such applications, high availability is very much desired;
> loss
> > of such a backbone or WAN is much too disruptive (and therefore
expensive)
> > to be much tolerated, and this kind of a market will gladly pay a
> > reasonable premium to achieve the needed fault tolerance.
> >
> > Why add Fault Tolerance now? Because it's easiest (and thus cheapest)
if
> > done from the start, and because having FT built in and therefore
becoming
> > ubiquitous will be a competitive discriminator, neutralizing one of the
> > remaining claimed advantages of ATM.
> >
> > Isn't Fault Tolerance difficult? In hub-and-spoke (logical star,
physical
> > loop) topologies such as GbE and10GbE, it's not hard to achieve both
fault
> > tolerance (FT) and military-level damage tolerance (DT). In networks of
> > unrestricted topology, it's a lot harder. The presence of bridges does
> not
> > affect this conclusion.
> >
> > How do I know that FT is so easily achieved? Because it's already been
> > done, may be bought commercially, and is in use on one military system
and
> > is proposed for others. The FT/DT technology mentioned here was
developed
> > on a US Navy project, and is publically available without intellectual
> > property restrictions. Why was the technology made public? To
encourage
> > its adopotion and use in COTS products, so that defense contractors can
> buy
> > FT/DT lans from catalogs, rather than having to develop them again and
> > again, at great risk and expense.
> >
> > What is the difference between Fault Tolerance and Damage Tolerance? In
> > fault tolerance, faults are rare and do not correlate in either time or
> > place. The classic example is the random failure of hardware components.
> > (Small acts of damage, such as somebody tripping over a wire or breaking
a
> > connector somewhere, are treated as faults here because they are also
rare
> > and uncorrelated.) In damage tolerance, the individual faults are
sharply
> > correlated in time and place, and are often massive in number. The
classic
> > military example is a weapon strike. In the commercial world, a major
> power
> > failure is a good example. Damage tolerance is considered much harder to
> > accomplish than fault tolerance. If you have damage tolerance, you also
> > have fault tolerance, but fault tolerance does not by itself confer
damage
> > tolerance.
> >
> > How is this Damage Tolerance achieved? All changes in LAN segment
> topology
> > (the loss or gain of nodes (NICs), hubs, or fibers) are detected in MAC
> > hardware by the many link receivers, which report both loss and
> acquisition
> > of modulated light. This surveillance occurs all the time on all links,
> and
> > is independent of data traffic. Any change in topology provokes the
> > hardware into "rostering mode", the automatic exploration of the segment
> > using a flood of special "roster" packets to find the best path, where
> > "best" is defined as that path which includes the maximum number of
nodes
> > (NICs).
> >
> > Just how fault tolerant and damage tolerant is this scheme? A segment
> will
> > work properly with any number of nodes and hubs, if sufficient fibers
> > survive to connect them together, and will automatically configure
itself
> > into a working segment within a millisecond of the last fault. If the
> > number of broken fibers is less than the number of hubs, all surviving
> > nodes will remain accessible, regardless of the fault pattern. If the
> > number of fiber breaks is equal to or greater than the number of hubs,
> > there is a simple equation to predict the probability of loss of access
to
> > a typical node due to loss of hubs and/or fibers, given only the number
of
> > hubs and the probability of any fiber breaking: Pnd[p,r]= ((2p)(1-p))^r,
> > where p is the probability of fiber breakage and r is the number of
> > surviving hubs (which ranges from zero to four in a quad system). This
> > equation is exact (to within 1%) for fiber breakage probabilities of 33%
> or
> > less, and applies for any number of hubs.
> >
> > The simplicity of this equation is a consequence of the simplicity of
this
> > protocol, which is currently implemented in standard-issue FPGAs (not
> > ASICs), and works without software intervention. It can also be
> > implemented in firmware.
> >
> > To give a numerical example, in a 33-node 4-hub segment, loss of 42
fibers
> > (16% of the segment's 264 fibers) would lead to only 0.5% of the nodes
> > becoming inaccessible, on average. Said another way, after 42 fiber
> breaks,
> > there are only five chances out of a thousand that a node will become
> > inaccessible. This is very heavy damage, with one fiber in six broken.
To
> > take a more likely example, with three broken fibers, all nodes will be
> > accessible, and with four broken fibers, there is less than one chance
in
> a
> > million that a node will become inaccessible. Recovery takes two ring
tour
> > times plus settling time (electrical plus mechanical), typically less
than
> > one millisecond in ship-size networks, measured from the last fault.
> > Chattering and/or intermittent faults can be handled by a number of
> > mechanisms, including delaying node entry by up to one second. Few
current
> > LAN technologies approach this degree of resilience, or speed of
recovery.
> >
> > In commercial systems and some military systems, a dual-ring solution is
> > sufficient. Up to quad-ring solutions are comercially available, needed
> > for some military systems. However, the ability to support up to quad
> > redundant systems should be provided in 10GbE, for two reasons. First,
> > quad is needed for the military market, which may be economically
> > significant in the early years of 10GbE. Second, quad provides a clear
> > growth path and a way to reassure non-military customers that their most
> > stringent problems can be solved: One can ask them if their needs really
> > exceed those of warships duelling with supersonic missiles.
> >
> > 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" and "http://grouper.ieee.org/
> > groups/802/3/10G_study/ public/ email_attach/ gwinn_2_0699.pdf". I was
a
> > member of the team that developed the technology, and am the author of
> > these documents.
> >
> > Although these documents assume RTFC, a form of distributed shared
memory,
> > the basic rostering technology can easily be adapted for Gigabit and
> > Ten-Gigabit Ethernet as well. For nontechnical reasons, RTFC originally
> > favored smart nodes connected via dumb hubs. However, the overall
design
> > can be somewhat simplified if one goes the other way, to dumb nodes and
> > smart hubs. This also allows the same dumb nodes to be used in both
> non-FT
> > and FT networks, increasing node production volume, and does not force
> > users to throw nodes away to upgrade to FT.
> >
> > I therefore would submit that 10GbE would greatly benefit from fault
> > tolerance, and also that it's very easily achieved if included in the
> > original design of 10GbE.
> >
> > Joe Gwinn