Lenovo 3200 Series, Installation Manual

Page: 45 / 132

8200 Series Converged Network Adapter and 3200 Series Intelligent Ethernet Adapter Page
27
Switch-Independent Load Balancing Mode Switch-independent load balancing mode provides a
failsafe feature and supports transmit load balancing. For receive load balancing, use the 802.3ad modes.
In this mode, the outbound traffic is efficiently distributed across the member adapters to increase the
transmit bandwidth. Traffic load balancing is connection-based to avoid out-of-order packet delivery. The
administrator can select one of the following load distribution types:
• Auto Select indicates that the load is distributed based on the target IP address (IPv4 or IPv6) and port
number. This option ensures a one-to-one correspondence between a traffic flow and a team adapter.
• MAC address based indicates that the load is distributed based on the target MAC address.
In switch-independent load balancing, a team receives the traffic on the preferred primary adapter. If the
preferred primary adapter fails, the receive load switches to a secondary adapter (failover operation). If the
preferred primary adapter becomes operational again, the receive load fails back to the preferred primary
adapter (failback operation). Thus, a switch-independent load balancing team also behaves like a failsafe
team. Each time the preferred primary changes due to failover or failback, other network elements are
notified of the change in the primary adapter through team gratuitous address resolution protocols (ARPs).
Link Aggregation Mode link aggregation provides increased bandwidth and high reliability by combining
several NICs into a single, logical, network interface called a link aggregation group (LAG). The link
aggregation is scalable, meaning an adapter can be added or deleted either statically or dynamically from
a team.
Traffic from all the team ports that form a LAG have the same MAC address, which is the MAC address of
the team. If a new adapter joins the LAG, or an adapter forming the LAG fails, the LAG becomes
operational again after a brief exchange of protocols between the switch and the server. QLogic Adapters
are rapidly aggregated, with a latency of 1 to 2 seconds.
Two options are available in the link aggregation mode:
• Static link aggregation
• Dynamic link aggregation
Static Link Aggregation (SLA) Static link aggregation (SLA, 802.3ad static protocols with generic
trunking) is a switch-assisted teaming mode, where the switch must be 802.3ad compliant. The switch
ports must be configured so that the switch perceives adapters from a LAG as a single, virtual adapter.
In SLA, the ports on the switch are active by default. There is no negotiation between the switch and the
teaming driver to decide on adapters participating in a LAG.
In SLA mode, the protocol stack responds to ARP requests with a single, advertised MAC address, and an
IP address corresponding to the LAG. Each physical adapter in the team uses the same team MAC
address during transmission. As the switch (at the other end of link) is aware of the trunking teaming mode,
it appropriately modifies the forwarding table to indicate the trunk as a single virtual port. This modification
ensures correct traffic routing on the receive side as well. In this mode, the switch also distributes receive
traffic across the member adapters.
Dynamic Link Aggregation (DLA) Dynamic link aggregation (DLA) with LACP is similar to SLA except
that LACP allows self configuration of LAG through handshaking between the switch and the intermediate
driver. For the team to function, LACP must be enabled at both ends of the link: the server and the switch.
LACP (802.3ad dynamic) allows switch ports to dynamically communicate with the teaming driver, allowing
controlled addition and removal of ports from the team.
Note: The switch must support the IEEE 802.3ad standard for the preceding two link aggregation
modes to work.