Re: Does Ten-Gigabit Ethernet need fault tolerance?
Rich,
At 3:09 PM 99/7/16, Rich Taborek wrote:
>
>I heartily agree with Mick's point. Fault Tolerance in Ethernet, if implemented
>at all, is generally implemented in at L3 or above (the MAC). I believe this to
>be appropriate for Ethernet networks. It seems to me that FT requirements are
>orthogonal to data transport requirements which are the purview of Ethernet.
>Here are some examples of networks which support FT:
I don't know that we disagree. Fault tolerance needs to be wholly in the
hardware, to keep software from screwing things up (never underestimate
software), but aside from that, I don't really have an opinion.
Actually, I would defer to your expertise on exactly where in the ethernet
hardware stack one should put the various pieces of the RTFC algorithm. I
wasn't proposing that the algorithm be in the MAC or anywhere else; this is
a detail to be decided later, if and only if people are interested in
including FT in 10GbE.
The MAC would however have to detect and report changes in network topology
(the gain or loss of links), as only the MAC can detect such things
directly. As discussed below, the necessary signals are probably already
available.
>1) WANs implementing SONET FT at level 1: Very, very expensive but capable of
>reliably transporting any type of data reliably upon protocol conversion. FT at
>level 1. FT at other levels is essentially ancillary.
Yes, but what does this have to do with the rostering algorithm? All of
SONET is expensive, as it's intended for large-scale Telco use. This is an
apples-to-oranges comparison.
>2) Mainframe channels: Fairly fault tolerant individually, and virtually
>fault-free considering multi-path configurations and multiple hosts controlled
>by a single operations point. I would put these in the very expensive category.
>FT at levels 1, 2 and 3 and above.
Yes, but what does this have to do with the rostering algorithm? This is
an apples-to-oranges comparison.
>3) Commodity servers, Ethernet and FT extensions such as Link Aggregation and
>Rapid Reconfiguration: Fault tolerant as a system, expensive only in comparison
>to non-fault tolerant systems (i.e. dirt cheap compared to other FT
>alternatives). FT at levels 3 and above.
My instinct is that if one were to implement rostering in 10GbE, most of
the logic would be in the hubs, and a small bit would be in the NICs, and
nothing more.
>I also agree that sufficient failure detection is already built into Ethernet.
You and Mick are very probably right. The existing RTFC implementations
use the existing synch-achieved signal from the vanilla gigabit TriQuint
Fibre-Channel chipset as the topology-change detector (qualified by a
optical-level signal from the optical transceiver). These signals are
probably already available within GbE and thus one assumes in10GbE to come.
>I'd put my money at the threshold of door #3.
I think we almost agree. I am simply proposing a FT technology to be
implemented in the hubs. I claim the algorithm is simple and cheap to
implement, perhaps cheap enough to compete with non-FT solutions on price
per NIC, hub, and fiber, which would be a change. (The overall FT network
will always be more expensive, because of the need to duplicate hardware.)
When we (the RTFC Team) looked around for suitable FT/DT products some
years ago, we found nothing suitable, and so were forced to invent RTFC and
the rostering algorithm. Granted that our military requirements were
lunatic fringe (survive three weapon strikes), but this had the odd effect
of forcing us to find a dead simple algorithm, because nothing more complex
could be guaranteed to survive even one cruise-missile strike, never mind a
salvo of three.
Joe
>--
>
>"Mick Seaman" 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.
>>
[snip]
>> 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.