Re: Clock Tolerance and WAN PHY
Roy,
I'm not aware that IEEE 802.3 is developing 10GE for use on servers
within two years. This is not a stated objective of the P802.3ae
project. Gigabit Ethernet is still not prevalent on servers now, and
it's been well over two years since the Gigabit Ethernet standard was
complete. However, the Gigabit Ethernet market is huge and meets or
exceeds all the projections made at the time the 802.3z project was
under development. 10 Gigabit Ethernet will follow in its footsteps
while at the same time directly addressing new MAN and WAN applications
to increase its application space and lower cost for all applications.
Higher throughput server I/O is coming quickly as a result of new
several I/O technologies. Perhaps the most promising I/O technology
coming down the pike is InfiniBand, a switched-point-to-point technology
with wide server industry support which is directly related to the XAUI
Serial I/O technology being developed in P802.3ae. Other industry
efforts to increase I/O bandwidth in the interim, such as Rapid I/O,
LDT, PCI-X and follow on's, etc. are under way as well. I don't foresee
any I/O bottlenecks in the near future as there have been in the past.
In fact, I predict that I/O technology will not be the bottleneck of
future servers and that the bottleneck will move upstream to the CPUs
themselves.
Note that any bottleneck to storage required to ensure that storage
performance keeps up with CPU and I/O performance is also removed by
technologies such as Fibre Channel, including 10 Gigabit Fibre Channel,
and InfiniBand. 10GFC leverages the 10GE LAN Physical Layer, and
thereby, XAUI Serial I/O technology to do the job.
I additionally predict that 10GE-capable servers will meet the Economic
Feasibility criteria of the P802.3ae PAR, offering 10X performance over
Gigabit Ethernet at 3.5X cost at product maturity. So you can not only
expect to find 10GE capable servers at Frys, you'll be amazed at how
little they cost :-)
Best Regards,
Rich
--
Roy Bynum wrote:
>
> Rich,
>
> I am not sure if anyone has done in technical or economic study of how many
> servers within the next two years will be able push close to 10Gb of data
> into a LAN PHY. From what I have been told, computer systems today can not
> even push enough to fill a GbE link, much less a 10GbE link. Without a
> computer/server market, the 10GbE LAN PHY will be relegated to bandwidth
> aggregation within privately owned campus fiber plant or a large data
> room. The 10GbE WAN PHY, is easily used extended LANs over leased fiber;
> MANs over privately owned or leased fiber and wavelengths; WANs over DWDM
> wave lengths or leased fiber.
>
> Given the amount of GbE that is being used by "legacy-free carriers" in the
> U.S.A. and Europe, and the expansion of the Internet, the need for a
> "service provider" type of Ethernet is becoming a major market
> issue. Combined with the other uses of the 10GE WAN PHY, I believe that a
> very high percentage (75%?) of 10GbE will be implemented using the WAN
> PHY. The WAN PHY will be used because it has the operational management
> functionality required. If P802.3ae does not go ahead and put the
> operational management functionality that is being recommended, all of the
> vendors will be implementing "proprietary" versions that will have it. If
> P802.3ae does not do it, then the need to do it in proprietary
> implementations will give 802.3 a "major black eye".
>
> I tend to agree with Gary Nicholl in the need for a + 20 PPM
> clock. Perhaps someone could give a relative cost to implement a + 20 PPM
> clock instead of a + 100 PPM clock.
>
> Thank you,
> Roy Bynum
>
> At 08:47 PM 1/27/01 -0800, Rich Taborek wrote:
>
> >Roy,
> >
> >I'm sorry to have confused you with my comments. Please allow me to try
> >and clear up any confusion I may have caused.
> >
> >The primary purpose of the 10 Gigabit Ethernet project is to provide the
> >customer with the next logical speed upgrade to 1 Gigabit Ethernet,
> >which provide the same for 100 Megabit Ethernet, which in turn...
> >
> >Supporting Ethernet over SONET or SDH or OTN or Digital Wrapper, etc. is
> >NOT a purpose. This effort is merely a goal or objective which will be
> >met, one way or another, while addressing the primary purpose. The
> >P802.3ae Task Force has seen fit to set an objective to develop a WAN
> >PHY to support Ethernet over SONET. I merely pointed out in my previous
> >note on this thread that the WAN PHY is only one of at least seven
> >current industry and standards activities to map Ethernet to the
> >existing WAN infrastructure. I believe that at least five of the seven
> >methods have been mentioned in this thread alone.
> >
> >It should be clear that without an optional objective to support the WAN
> >PHY, that the 10 Gigabit Ethernet PHY would be developed to support LAN
> >and MAN applications anyway. This is because:
> >
> >a) IEEE 802 is the LAN/MAN Standards Committee (LMSC);
> >b) IEEE 802.3z (Gigabit Ethernet) has already opened up the barn door
> >with its Fiber Optic technology, significantly increasing the link
> >distance supported by prior Ethernet physical layers and capable of
> >addressing MAN and WAN applications in addition to LAN;
> >c) There are a plethora of startups and large corporations already using
> >or eying Gigabit Ethernet and 10 Gigabit Ethernet LAN PHY equipment for
> >MAN and WAN application without requiring ATM or SONET at all.
> >
> >Now back to the subject of this thread...
> >
> >Clock Tolerance: +/-100 PPM is more than adequate for the 10 Gigabit
> >Ethernet LAN PHY. In fact, this tolerance is fairly difficult to achieve
> >as it corresponds to the 10.3125/3.125 Gbps line rate for the
> >Serial/WWDM, respectively, rather than the clock oscillator. The clock
> >oscillator frequency may actually be ~1/100 the line rate and is likely
> >multiplied up to the line rate.
> >
> >In the interest of standing a chance of meeting the IEEE P802.3ae
> >economic feasibility criteria, I strongly suggest keeping the clock
> >tolerance of the Ethernet PHY at +/- 100 PPM. This goes for both the LAN
> >and WAN PHYs since the WAN PHY would benefit greatly from the sharing
> >common components with the LAN PHY.
> >
> >
> >
> >Roy Bynum wrote:
> > >
> > > Rich,
> > >
> > > You have a very good a presenting that would seem reasonable to those who
> > > don't have any experience in attempting to implement what you are
> > > proposing. The objectives of P802ae include a WAN PHY. What constitutes a
> > > WAN PHY has been explained to the group by those of us that have worked in
> > > a WAN optical environment. You keep miss representing the requirements of
> > > a WAN PHY by presenting a LAN implementation as a WAN. It works very well
> > > at confusing those that are attempting to gain an understanding of what the
> > > issues are.
> > >
> > > Those of us that have worked in the WAN optical environment are not
> > > confused by your comments. Those of us that have worked in the WAN optical
> > > environment would like to have the opportunity to educate those that would
> > > actually like to gain a understanding of what the real world
> > requirements are.
> > >
> > > Thank you,
> > > Roy Bynum
> > >
> > > At 06:39 PM 1/24/01 -0800, Rich Taborek wrote:
> > >
> > > >Boaz,
> > > >
> > > >I strongly agree with your suggestion below, and Brad Booth's note along
> > > >the same vein, to leave any conversion between Ethernet and SONET at the
> > > >ELTE level. This would allow the LAN PHY, Serial or WWDM, it doesn't
> > > >really matter since the ELTE does the required conversion, to natively
> > > >serve LAN, MAN and WAN applications at the lowest possible cost. At the
> > > >same time, the same LAN PHY can attach to an ELTE and operate in "WAN
> > > >PHY mode" to support SONET/SDH. The latter would be applicable to those
> > > >legacy SONET/SDH core applications, again, AT THE LOWEST POSSIBLE COST.
> > > >
> > > >Of paramount importance to the customer is our ability to meet the
> > > >Economic Feasibility PAR criteria. In a nutshell this criteria states
> > > >that the cost of 10GE shall be ~3.5X the cost of Gigabit Ethernet at
> > > >product maturity. I don't visualize copying SONET/SDH, requiring +/-20
> > > >PPM clock tolerance, supporting overhead bytes, etc. as helping us meet
> > > >our economic feasibility objectives. However, I do see straightforward
> > > >mappings, the same clock tolerances as Gigabit Ethernet (+/-100 PPM),
> > > >protocol simplicity, relaxed jitter specifications, etc. as meeting the
> > > >objectives. The latter are characteristics of the LAN PHY, the former
> > > >are characteristics of the WAN PHY.
> > > >
> > > >Ethernet has met its economic feasibility objectives for three
> > > >generations through selection and leverage of low cost physical layers.
> > > >This will happen again through simple extension of Gigabit Ethernet
> > > >physical layer technology as is the case for the LAN PHY. It should be
> > > >obvious to the most casual observer that the more that SONET/SDH is
> > > >leveraged for 10GE, the higher the total 10GE solution cost will be.
> > > >
> > > >Just reading though the notes of this thread, it looks like there's
> > > >about as many ways to map Ethernet to SONET as there are to skin a cat.
> > > >Here are some of the ones from this thread:
> > > >
> > > >1) Ethernet over LAPS, ITU-T SG7
> > > >2) Packet over SONET, ???
> > > >3) Ethernet over SONET/SDH, ITU-T SG7 X.86
> > > >4) Ethernet over SONET/SDH, T1X1
> > > >5) Digital Wrapper, ITU-T ???
> > > >6) G.709, ITU-T (Same as Digital Wrapper???)
> > > >7) IEEE P802.3ae WAN PHY
> > > >
> > > >I'm kind of partial to Ethernet over Ethernet.
> > > >
> > > >KISS
> > > >
> > > >Best Regards,
> > > >Rich
> > > >
> > > >--
> > > >
> > > >Boaz Shahar wrote:
> > > > >
> > > > > MAC - (Serial lan phy) - ELTE - (sonet ring) - ELTE - (serial lan phy)
> > > > - MAC
> > > > >
> > > > > Anyway, the WIS takes 64/66 frames and encapsulates them into the SONET
> > > > > frame. So just take the 66/64 bit stream that comes to the ELTE
> > through the
> > > > > serial LAN and do the same there with a full SONET compliancy. Is that
> > > > > correct? What is the advantage comes from doing the WIS in the PHY?
> > (The
> > > > > rate is not a problem Just operate the MAC in SONET rate as you do
> > anyway
> > > > > and put the FIFO in the ELETE as you anyway have in the WIS)
> > > > >
> > > > > Thanks (Sorry for the long mail)
> > > > > Boaz
-------------------------------------------------------
Richard Taborek Sr. Phone: 408-845-6102
Chief Technology Officer Cell: 408-832-3957
nSerial Corporation Fax: 408-845-6114
2500-5 Augustine Dr. mailto:rtaborek@xxxxxxxxxxx
Santa Clara, CA 95054 http://www.nSerial.com