IBM Brocade, Инструкция по установке и эксплуатации

Страница: 22 / 74

10 Brocade 2-Port 10GbE Converged Network Adapter for IBM BladeCenter Installation and User’s Guide
Features, specifications, and standards
1
•
Interrupt Coalescing
This feature provides a method to delay generation of host interrupts and thereby combine
(coalesce) processing of multiple events. This reduces the interrupt processing rate and
reduces the time that the CPU spends on context switching. You can configure the following
parameters per port to adjust interrupt coalescing:
-
Interrupt time delay. There is a time delay during which the host generates interrupts. You
can increase this delay time and thereby coalesce multiple interrupts events into one. This
results in fewer interrupts for interrupt events.
-
Interrupt latency timer. An interrupt is generated when no new reply message requests
occur after a specific time period. You can adjust this time period and thereby minimize
I/O latency.
CEE and networking features
The Brocade 2-Port 10GbE Converged Network Adapter supports the following features for
Converged Enhanced Ethernet (CEE):
•
10 Gbps throughput per port full duplex
•
1500 or 9600 byte (Jumbo) frames
These frames allow data to be transferred with less effort, reduces CPU utilization, and
increases throughput. Mini-jumbo frames are required to encapsulate FCoE frames on CEE.
Network administrators can change the jumbo packet size from the default setting using host
operating system commands as described in the “Adapter Configuration” appendix in the
Brocade Adapters Installation and Reference Manual. Note that the MTU size refers to the
MTU for network configuration only. Internally, hardware will always be configured to support
FCoE frames that require mini-Jumbo size frames.
NOTE
The jumbo frame size set for the network driver cannot be greater than the setting on the
attached FCoE switch or the switch cannot accept jumbo frames.
•
Checksum/CRC offloads for FCoE packets, IPv4/IPv6 TCP and UDP packets, and IPv4 header.
The checksum offload supports Checksum offloads for TCP & UDP packets and IPv4 header.
This enables the adapter to compute the checksum, which saves host CPU cycles. The CPU
utilization savings for TCP checksum offload can range from few percent with MTU of 1500,
and up to 10-15% for MTU of 9000. The greatest savings are provided for larger packets.
•
Data Center Bridging Capability Exchange Protocol (DCBCXP) (802.1)
Used between adapter and FCoE switch to exchange configuration with directly connected
peers. DCBCXP uses LLDP to exchange parameters between two link peers.
•
Enhanced transmission selection (802.1Qaz)
Provides guidelines for creating priority groups to enable guaranteed bandwidth per group.
More important storage data traffic can be assigned higher priority and guaranteed bandwidth
so it is not stalled by less-important traffic.