The devices of the two sides were connected through the POS interface and devices under the NE40 accessed the Internet through C7613 router.
The customer reflected low access speed and found incorrect CRC packets, which increased constantly, on the POS interface of Huawei’s device. The customer believed that Huawei’s device led to low access speed.
1. The parameters configured for the routers on the two ends are inconsistent. ->Check the parameters for the two ends, including the format of encapsulated frames, link layer protocol, C2 flag, whether to set scramble. All parameters that may affect services are consistent. ->The possibility that the parameters configured for the routers on the two ends were inconsistent was excluded.
2. The transmission in between is of low quality. ->The transmission in between underwent detection by the transmission provider and turned out to be of good quality.
3. The tail fiber is of low quality. ->Tail fibers were replaced on both ends but the problem persisted.
4. The optical module or the board of the device is faulty. ->The LPUF motherboard and the optical module were replaced (there were no spare parts of PIM sub-cards onsite for customer reasons), the problem persisted.
5. Substitution test ->The AR46 was substituted for C7613, the POS was used for interconnection, and Smartbits was used to simulate traffic. It turned out that CRC on the POS interface of the BNE40 did not increase. ->The possibility of hardware faults of Huawei’s device was excluded.
6. Checking C’s device again found increase in underrun count on its POS interface. C’s support engineer was contacted to confirm that the reason for the generation of underrun count was that the hardware buffer was insufficient when data was sent from the interface. In this case, data packets sent were of low quality, which might lead to CRC check errors on the peer device. C’s personnel confirmed that the generation of underrun count was caused by the inaccurate clock of the transmission in between.
7. Solution: The mode of clock source on the POS interface of Huawei’s device was change to slave to accept the clock from the transmission device. CRC count still increased, and however the error rate was reduced from 0.3% previously to 0.1%.
For low access speed, software from a third party was used on the device attached to the NE40 for tests, which found long delay primarily on the outbound interface for the Internet instead of on the Intranet. Confirmation showed that the problem was caused by the second optimal routes delivered by the ISP public network and it was not CRC that led to low access rate.
1. The parameters configured for the routers on the two ends are inconsistent.
2. The transmission in between is of low quality.
3. The tail fiber is of low quality.
4. The optical module or the board of the device is faulty.
1. Summary on the parameters for the interconnection between the POS interfaces of the NE40 and the device from Vendor C:
1) POS physical layer encapsulation
The device from Vendor C adopts SONET for physical layer encapsulation and the NE40 adopts SDH by default. The different configurations here can be compatible with each other, without affecting protocol layer being Up or service forwarding. However, in actual projects, physical layer encapsulation is configured to be consistent preferably.
2) POS link protocol
The link layer protocol of the device from Vendor C is HDLC and that of the NE40 is PPP by default. The two ends must be configured to be the same; otherwise the protocol layer of POS on the two ends cannot be Up.
CRC of Vendor C is 16 bits and that of the NE40 is 32 bits by default. The two ends must be configured to be the same; otherwise the protocol layer of POS on the two ends cannot be Up.
Note: Unable to be modified on the NE40, CRC bit can be modified only on C’s device.
4) POS scramble
The device from Vendor C is configured with the scramble function and the NE40 is not. The two ends must be consistently configured. Otherwise, the POS protocol layer of the two ends cannot be Up.
5) C2 flag bit
POS 155 of Vendor C works in HDLC mode, and C2 byte is 0xCF; the NE40 works in PPP mode and C2 byte is 0x16. If the default value is not modified, C2 byte does not match. However, as only an identifier, C2 does not affect services even if it does not match.
6) Clock source
If the POS interfaces of the two routers are directly connected, one end should be set to use the primary clock mode and the other end the secondary clock mode, or both ends the primary clock mode.
If the transmission in between is the SONET/SDH device, the POS interfaces of both end should be set to the secondary clock mode.
If the transmission in between is the SONET/SDH device, it is recommended to set the clock to the primary clock mode.
By default, the clock of the NE40 is the primary clock mode and that of the device from Vendor C is Line, that is, the secondary clock mode.
2. CRC increase
Small increase in CRC does not affect normal services. An error rate below 1% is generally acceptable, which should be explained to the customer.
The CRC error rate is the CRC statistics value after the clearing of the interface count or the number of packets received during that period of time.
CRC increases rapidly due to the algorithm just after the interface count is cleared. Make statistics after a period of time.
3. Transmission quality
In the case, the transmission provider was asked to check transmission quality twice, but limited to the transmission medium and without considering the impact of clock synchronization on data transfer in a digital network. Clock transmission can similarly lead to the abnormality of data transfer.