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

RE: [EFM] EPON TDMA - Small timeslots/DBA

Title: RE: [EFM] EPON TDMA - Small timeslots/DBA
you are absolutely right on 1 and 2.  The problem is that we have to be sensitive (or compliant) to 802.3 format. How much change do we want to impose on 802.3 frame structure? Or a harder question: how much change are we allowed to impose on 802.3 and at the same time to benefit the simplicity of 802.3 frames?
On the 3rd point, transport efficiency can only be measured in a statistical sense. This is not the first time people transport complete Ethernet frames using fixed time slots. The game is to build a balance between TDM and the variable packet lengths, more importantly, with considerations given to the characteristics of intended network application, such as RTD, shared fiber in the case of EPON, or memory based switching fabric and addressing scheme in the case of a switch.
-----Original Message-----
From: Harry Hvostov [mailto:HHvostov@xxxxxxxxxxxx]
Sent: Friday, July 13, 2001 11:42 AM
To: Jian Song; mike.obrien@xxxxxxxxxxxx;
Subject: RE: [EFM] EPON TDMA - Small timeslots/DBA

1. For delay sensitive traffic requests could be implicit - periodic fixed size grants sent by OLT will work.
2. Requests could be "piggy-backed" on data transmissions, further reducing request/grant delay and overhead.
3. Fixed size cells don't work for variable size Ethernet frames. If you don't use fragmentation, worst case is transport of 64 byte frames in 1518 byte fixed size slots - efficiency of 4%.
-----Original Message-----
From: Jian Song [mailto:jsong@xxxxxxxxxx]
Sent: Friday, July 13, 2001 10:44 AM
To: mike.obrien@xxxxxxxxxxxx;
Subject: RE: [EFM] EPON TDMA - Small timeslots/DBA

The real issue is that we don't want to fragment 802.3 MAC frames as what ATM-PON did by using SAR devices on both ends. QoS may be affected by the following delays:

1) grant/request due to the long round trip time of 200uS for 20Km or 100uS for 10km;
2) the time slot duration from 50uS to 200uS;
3) guard time between two bursts from a few hundred nS to 10uS (see Mile's e-mail);
4) long Ethernet packet of 1500kB = 12uS.

So you see that long Ethernet packets is not worst offender here.  Mike's comments on efficiency is correct given the fact that guard time of sub 10uS is desirable at this time to address costs and feasibility issues.

Within a switch, the problem is different.  Use of a fixed size of switching cells is to reduce the system latency (therefore FIFO sizes) and easy operation (memory addressing). Also you don't really have any standard on fabric design.

-----Original Message-----
From: mike.obrien@xxxxxxxxxxxx [mailto:mike.obrien@xxxxxxxxxxxx]
Sent: Friday, July 13, 2001 9:55 AM
Subject: RE: [EFM] EPON TDMA - Small timeslots/DBA

Hi Xu,
        I'll take the liberty of answering this for Glenn. EPON uses a much
larger 'guard-band' than APON. This greatly simplifies the system timing,
however the trade off is a loss of efficiency. A typical EPON guard band -
this includes laser turn-off, turn-on, reciever AGC settling time, and PLL
locking - can be 5 to 10 micro-sec. At 8 ns per byte a 10 micro-sec gaurd
band is 1250 bytes. So using 64-byte timeslots would lead to an efficiency
of 64/(1250 + 64) ~ 4.8% .

        I hope this answers your question on why we dont use a small
timeslot. We will have to take another approach to DBA.

-----Original Message-----
From: xu zhang [mailto:zhangxu72@xxxxxxxxx]
Sent: Thursday, July 12, 2001 7:01 PM
To: glen.kramer@xxxxxxxxxxxx
Subject: [EFM] EPON TDMA

hi, glen:
 I had seen your presentation file about EPON TDMA in
PHY, it help me a lot to understand your EPON system.
We had developed the first APON system in china, when
I think of the TDMA of EPON, I think though the uplink
data rate is 1Gbits/s when shared by 16 or 32 users is
still not enough, so the dynamic bandwidth
allocate(DBA) protocal must be a requiremant
especially when take care of the QoS performance. In
DBA protocal, in order to achieve high performance the
time slot need be to small, I think why not we divide
the ethernet packet to 64 byte per solt, it is often
used in ethernet switch when store packet in SRAM.

best regards
xu zhang

Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail