IPG rate estimation for XGENIE
All,
I have examined the IPG bandwidth for the XGENIE approach.
This is to answer some of the XGENIE issues (2a,2b) raised by Mr. Paul
Bottorff farsightedly.
http://grouper.ieee.org/groups/802/3/10G_study/email/msg02127.html
Note that my current position is on (3); using LAN-PHY + XGENIE as the
WAN-PHY alternative to the ELTE, and the SONET side of the ELTE is the
Uni-PHY (64b/66b on SONET). Here we will use the SONET-compatible
XGENIE ordered-set for end-end OAM&P, instead of its loose relaying
to the SONET path overhead bytes shown in my previous PDF file.
http://grouper.ieee.org/groups/802/3/10G_study/email/pdf00009.pdf
As has been pointed out by Mr. Roy Bynum repeatedly, our approach
shoud pay 3% tax for 64b/66b on SONET. I believe this is worth
wasting since this approach has the following advantages;
a) any work done on 64b/66b for the LAN is leveraged.
(single packet delineation mechanism in 802.3ae)
b) future-proof for providing non-SONET dark-fiber solution that
will be connected seamlessly with LAN-only PHY. It will provide
full 10Gb/s MAC rate instead of 9.58464 Gb/s.
c) XGENIE approach allows both serial PHY and CWDM PHY to be WAN-PHY.
It would be appreciated if someone could confirm the following estimation
for the IPG bandwidth to be valid or have some mis-understanding.
Best Regards,
Osamu
(1) LAN-PHY with XGENIE, without Open Loop Control*.
*Mr. Shimon Muller's Albuquerque proposal
http://grouper.ieee.org/groups/802/3/ae/public/mar00/muller_1_0300.pdf
|<-------------------- 125us -------------------------->|
10.00000 Gb/s |<-------------------- 156,250 bytes ------------------>|
[001_1530][12][002_1530][12]......[12][101_1530][12][508bytes
IPG#1 IPG#2 IPG#101
Assuming that each IPG is /A/K/R/ and we can replace a
half of /R/ by XGENIE /O/, we can transmit 50 ordered
sets per 125 us. This is enough to carry 30 overhead
bytes defined for full-SONET compatibility.
If we consider 'JUMBO frame', full-SONET-compatible
XGENIE would be difficult to be guaranteed. However
we still have enough space to carry B1,J0, B3, G1.
e.g. 8kB Jumbo frame reduces the IPG rate to one fifth,
and hence guarantees only 10 ordered set per 125 us.
(2) LAN-PHY with XGENIE, with Open Loop Control to 9.58 Gb/s MAC rate.
Later the Ethernet frame may be accomodated in SONET payload by
zero-overhead delineation mechanism such as EOS or SLP.
9.95328 Gb/s |<-------------------- 155,520 bytes ------------------>|
9.58464 Gb/s | <--------> <-- ...149,760 bytes <------->|
[1_640][1_16,640][2_640][2_ .......... ][9_640][9_16,640]
SONET-frame [S_OH1][Payload1][S_OH2][Pay . .... ][L_OH5][Payload9]
Open Loop Control will reduce the MAC rate by 4.1536 %.
This corresponds to increasing the averaged IPG to 67 bytes
assuming that the maximum-size packets are continuously transmitted.
10.00000 Gb/s |<-------------------- 156,250 bytes ------------------>|
[001_1530][67][002_1530][67].....[67][097_1530][67][1341bytes
IPG#1 IPG#2 IPG#97
9.58464 Gb/s |<-------------------- 149,760 bytes ------------------>|
SONET-payload [001_1530][12][002_1530][12]......[12][097_1530][12][186bytes
IPG#1 IPG#2 IPG#97
Assuming that each IPG is /A/K/R/K/R/..../A/K/R/....and
we can replace a half of /R/ by XGENIE /O/, we can
transmit 15 ordered sets per IPG, and hence 1,455 ordered
sets per 125 us at the 10.3 Gb/s LAN-PHY section.
Therefore, we will have no difficulty even with the
Jumbo frame.
In the SONET payload, we can not expect the IPG
transparency. Hence we need some OAM&P info relay
at ELTE. (Please see my previous PDF picture.)
(3) LAN-PHY with XGENIE with Open Loop Control to 9.29 Gb/s MAC rate.
Later the Ethernet frame may be accomodated in SONET payload by
64b/66b according to Mr. Howard Frazier's Uni-PHY proposal.
9.95328 Gb/s |<-------------------- 155,520 bytes ------------------>|
9.58464 Gb/s | <--------> <-- ...149,760 bytes <------->|
9.29420 Gb/s | <--------> <-- ...145,222 bytes <-------> |
[1_640][1_16,640][2_640][2_ .......... ][9_640][9_16,640]
SONET-frame [S_OH1][Payload1][S_OH2][Pay . .... ][L_OH5][Payload9]
Open Loop Control will reduce the MAC rate by 7.058 %.
This corresponds to increase the averaged IPG to 117 bytes
assuming that the maximum-size packets are continuously transmitted.
10.00000 Gb/s |<-------------------- 156,250 bytes ------------------>|
[01_1530][117][02_1530][117].....[117][94_1530][117][1432bytes
IPG#1 IPG#2 IPG#94
9.29420 Gb/s |<-------------------- 145,222 bytes ------------------>|
SONET-payload [001_1530][12][002_1530][12]......[12][094_1530][12][274bytes
IPG#1 IPG#2 IPG#94
Assuming that each IPG is /A/K/R/K/R/..../A/K/R/....and
we can replace a half of /R/ by XGENIE /O/, we can
transmit 27 ordered sets per IPG, and hence 2,538 ordered
sets per 125 us at the 10.3 Gb/s LAN-PHY section.
In the SONET payload, we will be able to transmit 47
ordered sets per 125 us. This is enough to carry 30
overhead bytes defined for full-SONET compatibility.
Moreover, in this case, we may have no difficulties
even with the Jumbo frames since the open loop control
will assure the wider IPG when we encocunter them.