HP APA Configuration Examples

This section shows some sample HP APA configurations. Select a configuration that most closely matches the environment into which you want to configure HP APA on your system.

Enterprise Intranet Client/Server Environment

Figure 3-1 shows a sample enterprise client/server environment. This type of environment is a good candidate for HP APA link aggregations, and has the following characteristics:

Requires a switch capable of trunking or load balancing.
Many clients produce many connections. This makes effective use of the HP APA outbound network traffic distribution algorithms. The HP APA MAC address load-balancing algorithm is a good choice. The IP address and TCP/UDP port address load-balancing algorithm also works effectively in this configuration.
The switch typically provides good inbound traffic distribution. Most switches use the data packet's source MAC address, or a combination of the packet's source and destination MAC addresses, to provide inbound load balancing.
Depending on the network traffic bandwidth requirements, you can use two to four (two to eight for the December 2005 release (B.11.23.10)) 100BT interfaces or two to four (two to eight for the December 2005 release (B.11.23.10)) Gigabit interfaces in an PAgP or MANUAL link aggregation. With LACP, you can use up to 32 interfaces in the link aggregation. This enables bandwidth scalability as network loads increase as the organization grows.

Figure 3-1 Sample Enterprise Intranet Client/Server Configuration

Sample enterprise intranet client/server environment

Internet or Large Enterprise Environments Using Routers

You can use HP APA link aggregation successfully in certain environments employing routers. You must be careful because a particular router might not have a load balancing capability. Additionally, switches employed between the server employing HP APA and the router inject another level of complexity that you should analyze before determining that the environment is a candidate for HP APA link aggregations.

Figure 3-2 shows a sample router and server configuration with no switch. This configuration makes the following assumptions:

The router or switching router connected to the server provides trunking or load balancing using an IP address-based load-balancing algorithm.
There will be many TCP/UDP client connections. The HP APA IP address load-balancing algorithm provides effective outbound network traffic load balancing, as does the TCP/UDP port address algorithm. Do not use the MAC address algorithm because all packets transmitted from the server would contain the same source and destination MAC addresses.

Figure 3-2 Sample Router and Server Configuration (No Switch)

Figure 3-3 shows a sample router and server configuration with a switch. In this configuration, the switch might present problems because switches typically use a MAC address load-balancing algorithm. This might make the switch a bottleneck point because the packets from the router and from the server will contain the same source and destination MAC addresses, thus defeating the load-balancing algorithm for both inbound and outbound data at the server.

This condition might be acceptable if the load balancing of inbound traffic to the server is not a concern and the link between the switch and the router has greater bandwidth capacity than the server's link aggregation. For example: The server's link aggregation is composed of 100BT links and the link between the switch and the router is a Gigabit link.

Figure 3-3 Sample Router and Server Configuration (Switch)

Server-to-Server (Back-to-Back)

Figure 3-4 shows a sample server-to-server configuration. You create server-to-server aggregations by directly connecting the physical ports in one server's link aggregation to the physical ports in the other server's link aggregation. This configuration has the following characteristics:

It needs many TCP/UDP client connections between the servers in order for load balancing to be effective. Therefore, use the HP APA TCP/UDP port load-balancing algorithm.
Depending on the network traffic bandwidth requirements, use two to four (two to eight for the December 2005 release (B.11.23.10)) 100BT interfaces or two to four (two to eight for the December 2005 release (B.11.23.10)) in an PAgP or MANUAL link aggregation. With LACP, you can use up to 32 interfaces in the link aggregation. This enables bandwidth scalability as network loads increase as the organization grows.

Figure 3-4 Sample Server-to-Server Configuration (Back-to-Back)

Hot Standby for High Availability

Figure 3-5 shows a sample MANUAL (Hot Standby) mode configuration. These link aggregations provide high availability network access with an active link and a standby link.




	NOTE: HP strongly recommends using failover groups (LAN_MONITOR mode) rather than Hot Standby mode.

This configuration has the following characteristics:

The Hot Standby active link carries network traffic until it or its link partner fails. In that event, the standby link takes over the responsibility for delivering network traffic. If the previous active link is configured with a higher port priority than the current active link, when it recovers it resumes being the active link delivering the network traffic. If the port priorities are the same, the current active link continues as the active link.
The active and standby links must both be the same type of device: 100Base-T or Gigabit.
Hot Standby link aggregations can be connected to any switch or hub. The ports must be cabled to a switch and the switch ports must not be configured for an aggregation.
Dual switches or hubs (as used in Figure 3-5) are not required. But dual switches and hubs provide a more reliable network environment by removing single points of failure. Both switches or hubs must be on the same subnet.

Figure 3-5 Sample Hot Standby Configuration for High Availability

Server-to-Server with Switch (Not Recommended)

Figure 3-6 shows a sample server–to–server HP APA link aggregation configuration with a switch between the servers. This configuration will not work as intended for the following reasons:

The switch nullifies any load balancing of network traffic provided by HP APA.
The switch uses a MAC address load-balancing algorithm. Because the servers' link aggregations have fixed MAC addresses, the switch will not load balance; it will only transmit data on one physical link.

Figure 3-6 Sample Server-to-Server Configuration with Switch (Not Recommended)

Failover Group

Figure 3-7 shows a sample failover group (LAN_MONITOR mode) configuration. This configuration provides high availability network access with an active link and a standby link, and has the following characteristics:

Dual switches or hubs are not required. However, dual switches and hubs provide a more reliable network environment by removing the switch or hub as a single point of failure. If two switches or hubs are used, there must be a data path between them to allow them to be on the same subnet.
You can connect failover groups to any switch or hub.
The link partner does not require trunking to be enabled.

Figure 3-7 Sample Failover Group (LAN_MONITOR) Configuration

Failover Group Using Link Aggregates

Figure 3-8 shows a failover group that uses link aggregates as the active and standby devices to increase the network bandwidth through load balancing across the physical links. This configuration has the following characteristics:

You can use any HP APA link aggregate, except Hot Standby, as a device in the failover group.
The standby link does not have to be a link aggregation. It can be a single physical link of the same type as used in the link aggregation.
Dual switches are not required. However, dual switches provide a more reliable network environment by removing the switch as a single point of failure. If two switches are used, there must be a data path between them.
LAN Monitor failover groups using link aggregates are restricted to switches supported by HP APA link aggregates.

Figure 3-8 Sample Failover Group Using Link Aggregates Configuration

HP APA Configuration Examples

Technical documentation

» Table of Contents

» Glossary

» Index