Per Flow Hashing - Alcatel-Lucent 7750 Configuration Manual

Per Flow Hashing

Per flow hashing uses information in a packet as an input to the hash function ensuring that any
given flow maps to the same egress LAG port/ECMP path. Note that because the hash uses
information in the packet, traffic for the same SAP/interface may be sprayed across different ports
of a LAG or different ECMP paths. If this is not desired, other hashing methods outlined in this
section can be used to change that behavior. Depending on the type of traffic that needs to be
distributed into an ECMP and/or LAG, different variables are used as input to the hashing
algorithm that determines the next hop selection. The following outlines default per flow hashing
behavior for those different types of traffic:
•
•
•
•
7750 SR Interface Configuration Guide
VPLS known unicast traffic is hashed based on the IP source and destination addresses for
IP traffic, or the MAC source and destination addresses for non-IP traffic. The MAC SA/
DA are hashed and then, if the Ethertype is IPv4 or IPv6, the hash is replaced with one
based on the IP source address/destination address.
VPLS multicast, broadcast and unknown unicast traffic.
→ Traffic transmitted on SAPs is not sprayed on a per-frame basis, but instead the
service ID is used to pick ECMP and LAG paths statically.
→ Traffic transmitted on SDPs is hashed on a per packet basis in the same way as VPLS
unicast traffic. However, per packet hashing is applicable only to the distribution of
traffic over LAG ports, as the ECMP path is still chosen statically based on the service
ID.
Data is hashed twice to get the ECMP path. If LAG and ECMP are performed on the
same frame, the data will be hashed again to get the LAG port (three hashes for LAG).
However, if only LAG is performed, then hashing will only be performed twice to get
the LAG port.
→ Multicast traffic transmitted on SAPs with IGMP snooping enabled is load-balanced
based on the internal multicast ID, which is unique for every (s,g) record. This way,
multicast traffic pertaining to different streams is distributed across different LAG
member ports.
→ The hashing procedure that used to be applied for all VPLS BUM traffic would result
in PBB BUM traffic being sent out on BVPLS SAP to follow only a single link when
MMRP was not used. Therefore, in chassis mode D, traffic flooded out on egress
BVPLS SAPs is now load spread using the algorithm described above for VPLS
known unicast.
Unicast IP traffic routed by a router is hashed using the IP SA/DA in the packet.
MPLS packet hashing at an LSR is based on the whole label stack, along with the
incoming port and system IP address. Note that the EXP/TTL information in each label is
not included in the hash algorithm. This method is referred to as Label-Only Hash option
and is enabled by default, or can be re-instated in CLI by entering the lbl-only keyword. A
couple of options to further hash on the header of an IP packet in the payload of the MPLS
packet are also provided.
Interface Configuration
Page 121