HP ProLiant ML310 G3 Data Protection Storage Server administration guide (June 2006)

Switch-assiste
dLoadBalancingwithFaultTolerance(SLB)
Switch-assiste
d Load Balancing with Fault Tolerance (SLB) is a team t ype that allows full transmit and
receive load ba
lancing. SLB requires the use of a switch that supports some form of Po rt Trunking (for
example, EtherChannel, MultiLink Trunking, and so on) . SLB doesn’t support switch redundancy since all
ports in a team must be connected to the same switch. S LB is similar to the 802.3ad Dynamic team type
discussed late
r.
Transmit Load Balancing with Fault Tolerance (TLB)
Transmit Load Balancing with Fault Tolerance (TLB) is a team type tha t allows the server to load ba lance
its transmit traffic. TLB is switch independent and suppor ts switch fault tolerance by allowing the teamed
ports to be connected to more than one switch in the same L AN. With TLB, trafficreceivedbytheserveris
not load b alanced. The primary teamed port is responsible for receiving all traffic destined for the server.
In case of a failure of the primary teamed port, the NFT mechanism ensures connectivity to the server
is preserved by selecting another teamed port to assume the role.
Transmit Load Balancing with Fault Tolerance and Preference Order
Transmit Load Balancing with Fault To lerance and Preference Order is identical in almost every way to
TLB. The only difference is that this team t ype allows the server administrator to prioritize the order in
which team
ed ports should b e the Primar y team ed por t. This abilit y is important in environments where
one or more teamed p orts are m ore preferred than other ports in the same team. The need for ranking
certain tea m ed por ts better than others can be a result of unequal speed s, better adapter capabilities (for
example, higher receive/transmit descriptors or buffers, interrupt coalescence, and so on), or preference
for the team’s Primary port to be located on a specificswitch.
Network Fault Tolerance Only (NFT)
Network Fault Tolerance (NFT) is the foundation of HP ProLiant Network Adapter Teaming. In NFT mod e,
from two to eight teamed ports are teamed together to operate as a single virtual net work adapter.
However, only one teamed port—the Primary teamed port— is used for both transmit and receive
communication with the server. The remaining adapters are considered to be stand-by (or secondary
adapters) and are referred to as N on-Primary teamed ports. Non-Primary teamed ports remain idle
unless the Primary teamed port fails. All team e d ports m ay transmit and receive hear tbe ats, including
Non-Primar y adapters.
The fault-tolerance feature tha t NFT represents for HP ProLiant Net work Adapter Teaming is the only
feature found in every other team type. It can be said that the foundation of every team type supports NFT.
Network Fault Tolerance Only with Preference Order
Network Fault Tolerance Only with Preference Order is identical in almost every way to NFT. The only
diffe
rence is that this team type allows the server administrator to prioritize the order in which teame d
ports should be the Primary team ed port. This abilit y is important in environments where one or m ore
teamed ports are more preferred than other ports in the same team. The need for ranking certain teame d
ports better than others can be a result of unequal speeds, bet ter adapter capabilities (for example,
high
er receive/transmit descriptors or buffers, interrupt coalescence, and so on) , or preference for the
tea
m’s Primary port to be located on a specificswitch.
Transmit load balancing m ethods (alg orithms)
All load-balancing team types (TLB, SLB, 802.3ad Dynamic, and Dual Channel) load balance transmitted
frames. There is a fundamental decision that must be made when determining load balancing
mechanisms: whether or not to preserve frame order.
Frame order preser vation is impor tant for several reasons—to prevent frame retransmission because
frames arrive out of order and to prevent performance-decreasing frame reordering within OS protocol
stacks. In order to avoid frames from being transmitted out of order when c ommunicating with a target
network device, the team’s load-balancing algorithm assigns outbound conversations to a par ticular
HP ProLiant ML310 G3 Data Protection Storage Server 45