Gigabit Ethernet is a high-performance,
Ethernet networking solution for HP servers and workstations.
The Local Area
Network cards are data
link cards that support the IEEE 802.3z standard for 1000Base-SX
operation over multimode fiber, and the IEEE 802.3ab standard for
1000Base-T operation over 4-pair Cat-5 or Cat-5E UTP copper wiring.
1000Base-T brings high bandwidth with 10, 100 and 1000 Mbit/s
speeds, which makes more processing power available for applications.
Features |
 |
Gigabit Ethernet features include the following:
1000Base-T features:
Triple speed
10/100/1000 Mbit/s
Full-duplex operation at
10/100/1000 Mbit/s and half-duplex operation at 10/100 Mbit/s
(no half-duplex support for 1000 Mbit/s)
Autonegotiation and autosensing
to the highest available link speed
Minimized host CPU utilization:
Protocol off
loading through on board TCP, UDP, and IP checksum calculations
Adaptive interrupt coalesce
is based on traffic load
The IEther-00 driver
software bundle (iether driver) and the
GigEther-01 driver software bundle (igelan
driver) provide:
Starting
with the May 2005 release of HP-UX 11i v1 and v2, you have the
ability to turn on or off the receive and send sides of
the card’s Checksum Offload (CKO).
See details in “Checksum Offload Changes
for the iether and igelan Gigabit Ethernet Drivers”.
This capability is not supported on the gelan or btlan drivers.
Starting with the
May 2005 release of HP-UX 11i v1 and v2, the iether and igelan drivers
support 1000FD mode of switches (Table 1-1 “HP-UX 1000Base-T Supported Configurations”, row 2). Gigabit Ethernet cards in autonegotiation
mode will now support 1000FD manual/fixed mode on the link partner
(such as a switch). All supported configurations are shown in (Table 1-1 “HP-UX 1000Base-T Supported Configurations”) including those where the
gigabit Ethernet card and the link partner also have the same settings.
Set the ports on the card and on your switch according
to the following table. The newest supported configuration is shown
in the second row.
Table 1-1 HP-UX 1000Base-T Supported Configurations
| HP-UX 1000Base-T Port | Link Partner | Resulting Speed |
|---|
| AUTO | AUTO | Highest Common Speed (HP-UX supports 10/100/1000) |
| AUTO | 1000 FD[1] fixed/manual | 1000 Mbit/s FD |
| 100 HD | 100 HD | 100 Mbit/s HD |
| 100 FD | 100 FD | 100 Mbit/s FD |
| 10 HD | 10
HD (for example, a
10Base-T
Hub) | 10
Mbit/s HD |
| 10 FD | 10
FD | 10
Mbit/s FD |
TCP Segmentation
Offload (TSO). See details in “Support for TCP Segmentation
Offload (TSO)” of this document. TSO is not supported on
the gelan or btlan drivers.
64-bit management
information base (MIB) statistics. Directs the data
link service (DLS) provider to return 64 bit statistics. The GigEther-00
driver software bundle (gelan driver) also supports 64-bit
(MIB) statistics on HP-UX 11i v2 of September 2004.
IEEE 802.1Q Virtual
LAN (VLAN) tagging and IEEE 802.1p Priority Queuing on
HP-UX 11i v1 and the September 2004 Release of 11i v2.) The VLAN
feature was on HP-UX 11i v1 since June 2004 but was not on the HP-UX
11i v2 release of March 2004. This feature is described in Using
HP-UX VLANs on
http://docs.hp.com.
HP-UX LAN provider
support.
You can use WBEM-based clients to access the LAN Provider and collect
information about the Ethernet links on your system. For details
on the LAN provider, please refer to the documentation for the LAN Provider
on http://docs.hp.com.
The GigEther-00 driver software
bundle (gelan) driver and the always-installed Fast
Ethernet btlan driver provide:
64-bit management
information base (MIB) statistics. Directs the data
link service (DLS) provider to return 64 bit statistics.
Virtual LANs (VLANs).
This feature was on HP-UX 11i v1 since June 2004 but was not on
the HP-UX 11i v2 release of March 2004. This feature is described
in Using HP-UX VLANs on
http://docs.hp.com.
HP-UX LAN provider support.
You can use WBEM-based clients to access the LAN Provider and collect
information about the Ethernet links on your system. For details
on the LAN provider, please refer to the documentation for the LAN Provider
on
http://docs.hp.com.
Jumbo Frames support:
Large 9000-byte
maximum transmission unit (MTU) for improved efficiency and performance
with bulk data transfer. On HP-UX 11i v2, the jumbo frames range is
1501 - 9000. See Table 3-1 “Allowable MTU Sizes” for
allowable MTU settings per driver and HP-UX release
Supported only at 1000 Mbit/s
interface (link partner must also support Jumbo Frames)
MC/ServiceGuard and Auto-Port
Aggregation (APA) support for high availability
Configuration support through
SAM
Promiscuous mode (link and
SAP) inbound and outbound support
Supports TCP/IP, NFS and
DLPI applications (Ethernet and SNAP encapsulations are supported)
Supports OL* (Online Addition
and Replacement) (HP-UX 11i v1 PCI-X and 11i v2 PCI/PCI-X only)
Note that the SX and T cards are not considered
to be like cards for OL* that is, you cannot
replace an SX card with a T card, and vice versa.
HP does not support OL* on A6794A.
Checksum Offload Changes
for the iether and igelan Gigabit Ethernet Drivers |
|
Beginning with the HP-UX 11i v1 and v2 OE release of May 2005,
the iether and igelan gigabit
Ethernet networking drivers allow you to turn the CKO feature of
the gigabit Ethernet interface ON or OFF for both the receive and
send sides of the card. With this and subsequent versions of the iether and igelan drivers:
The default for receive side
CKO is OFF (disabled). Prior to the May 2005 release, the receive
side CKO was ON (enabled). This setting can be restored as explained
in the examples that follow.
The default for the send side
is CKO ON (enabled) -- the same setting as on prior releases.
Example settings for iether driver:
The following description is for the iether driver;
if you’re using the igelan driver,
just substitute igelan for iether in
the following strings.
Recommended: To set the CKO options
permanently (so they’re saved across reboots), edit the
following values in the /etc/rc.config.d/hpietherconf file:
HP_IETHER_SEND_CKO and
HP_IETHER_RECV_CKO and set them to ON or OFF
To determine the current CKO settings, enter the following
command:
lanadmin -x cko ppa
Optional: If desired, you can also use
the lanadmin -X command to temporarily set these values; but if you
use the lanadmin command, your settings will not be saved across a reboot.
The exact lanadmin commands are as follows (note that the -X is capitalized
in these commands):
lanadmin -X send_cko_on ppa
lanadmin -X recv_cko_on ppa
Please refer to the lanadmin man page for details of these CKO options.
Support for TCP Segmentation
Offload (TSO) |
|
TCP Segmentation Offload (TSO) is a mechanism by which the
host stack offloads certain portions of outbound TCP packet processing
to the Network Interface Card (NIC) thereby reducing host CPU utilization.
This functionality can significantly reduce the load on the server
for certain applications which primarily transmit large amounts
of data from the system. Examples include web servicing, NFS, and
file transfer applications.
How It Works
The reduction in CPU
utilization is achieved primarily by allowing the host to transmit large
frames (frames larger than the links Maximum Transmission Unit or
MTU) to the NIC which are subsequently carved up into smaller, MTU-sized
frames by the NIC, before transmission on the wire. Thus instead
of processing many small MTU-sized frames during transmit, the host
sends fewer larger VMTU (Virtual MTU) sized frames thereby increasing
the efficiency of the data transfer in the host. The VMTU is typically much
larger than the links MTU; for example, on a typical Ethernet card,
the link MTU is 1500 bytes while a VMTU could be as large as 64Kbytes.
Greater than 50% reduction in CPU utilization has been observed
on some FTP workloads.
|
| |
|
 | NOTE: Not all applications benefit from the TSO mechanism.
Only data intensive applications which transmit large data buffers
using TCP over IPv4 are improved. Other types of applications will
not significantly benefit from the TSO mechanism. Performance improvements
vary depending upon the platform used. Systems which support hardware
partitioning will notice a decrease in per-card throughput in addition
to the significant reduction in CPU utilization. |
|
| |
|
Managing TSO
TSO is disabled by default. To enable TSO on each specific
interface, use the lanadmin command to set the “vmtu” tunable to a non-zero value. See
the next section for details.
New lanadmin Command Option: vmtu
The lanadmin command is updated to get or set
a new tunable called “vmtu” -- the virtual MTU of a link that supports
TSO. The following command will list the TSO capability of the link:# lanadmin -x vmtu <ppa>
Driver/Hardware supports TCP Segmentation Offload. Current VMTU = 32160.
The syntax of the command to set the vmtu is as follows (note:
the X is capitalized):# lanadmin -X vmtu new_vmtu_value <ppa>
|
| |
|
| NOTE: Currently only two values are allowed for the tunable “vmtu”:
0 and 32160. Setting “vmtu” to 0 (default) disables
the TSO feature, and setting it to 32160 enables it. |
|
| |
|
Changes to the Output of Current Commands
If you are using NetTL or tcpdump to trace outbound packets
may see differences from current behavior. Tracing outbound packets
in the host will display large packets (> MTU size, as
large as 64K) being transmitted on the link, while tracing the packets
on the wire or at the receiving end will display only packets that
are lesser than or equal to the link MTU size.
TSO Interaction with Other Software
TSO is supported on virtual LANs (VLANs). It is supported
at all MTU values (that is, TSO is supported over Jumbo frames)
and speed settings on the link. Both Ethernet and SNAP encapsulations
are supported.
TSO is supported on link aggregations starting with the September
2004 version of the HP Auto Port Aggregation [APA] product. For
further details, please refer to the latest release notes for the
HP APA product. When used with HP APA product versions prior to September
2004, the TSO functionality is automatically disabled on those link aggregations.
Printable version
