The DBA Issue Description:
Initial Issue:ONT2-32 are passing 25Mbps untagged upstream, and ONT2-32 are passing 20Mbps p-bit 0 and 15Mbps p-bit 4 downstream.
All traffic passing as expected. ONT1 is passing 90Mbps upstream and ONT1 is passing 60Mbps p-bit 0 and 30Mbps p-bit 4 downstream
All downstream traffic is passing as expected.
The Upstream traffic is being dropped. Getting 37M instead of the expected 90M
According to the upgrade solution, the issue could be summarized as followings:
To Update the MA5603U from V800R202C01 to V800R203C00. The Configuration of the DBA Calculation Period Supported by the version R203C00:
The default mode is min-loop-delay, and the DBA calculation period is 2 GPON frame periods.
A)Optimal upstream data transmission delay is achieved in the GPON system.
B)The low buffer requirement for the ONT optimizes the maximum upstream bandwidth of a single or a few ONTs (without a packet length restriction).
C)Bandwidth is fairly allocated to short and medium-length packets (for example, 256 bytes).
Bandwidth is unfairly allocated and bandwidth usage is low in some scenarios.
A)In the max-bandwidth-usage mode, the DBA calculation period is 8 GPON frame periods.
Higher bandwidth usage is obtained at the cost of a longer delay.
B)In the manual mode, users can set the DBA calculation period to 16 GPON frame periods in the diagnose mode.
Higher bandwidth usage is obtained at the cost of an even longer delay.
Huawei recommends using the default DBA calculation period mode, min-loop-delay.
The bandwidth is wasted in certain conditions when bandwidth is allocated using NAB of the DBA algorithm. The analyse is as the following:
1. The ETH packets of ONTs 2�32 enter a queue in an average of every four GPON frame periods, which occurs randomly. When the scenario in the preceding figure occurs, the total bandwidth usage will drop.
2. If the packet is short (for example, 256 bytes), packets of ONTs 2�32 enter queues in each GPON frame. Therefore, bandwidth is allocated more fairly and the bandwidth usage increases.
*In each frame from ONT 1, there is a 1450-byte packet requesting 720 Mbit/s bandwidth.
*In every four frames from ONTs 2�32, there is a 1450-byte packet, and there are two packets in each ONT queue after 8 GPON frame periods. Each ONT requests 180 Mbit/s bandwidth.
3. Assume that the DBA calculation period of the OLT is eight GPON periods. Now each ONT plays a part in bandwidth allocation. The OLT equally allocates the total bandwidth requested by the ONTs in the eight frame periods to the eight frames, that is, ONT 1 requests 90 Mbit/s and ONTs 2�32 each request 25 Mbit/s. Bandwidth allocated according to the DBA algorithm is used more efficiently.
4. After the DBA calculation period is lengthened, bandwidth is allocated more fairly and used more efficiently. However, the lengthened DBA calculation period greatly increases the data transmission delay.
In addition, the lengthened DBA calculation period requires more buffer capability of the ONT. Especially, short packet transmission significantly increases ONT buffer occupancy.
It also greatly influences the upstream throughput in the following scenario:
A few (for example, two or three) ONTs under a PON port are VIP users, whose maximum upstream bandwidth reaches 500 Mbit/s. The upstream throughput performance of a single ONT is greatly influenced by the buffer.