KEMP Technologies LoadMaster 1500 Installation And Configuration Manual Download

Page: 21 / 71

G. Layer 4 Persistency

1. Source IP Address Based Persistency

The LoadMaster can balance TCP or UDP based traffic based on source and destination IP addresses. All
packets are passed through to one of the real servers. It can also ensure that requests from one particular

host always go to the same server. This is known as connection persistency.

Persistency in the context of load balancing means that multiple requests from a client to a virtual service are

redirected to the same real server as that selected for the first request.

The duration period for which the LoadMaster maintains the persistency for a given source IP address can be
controlled via the persistency timeout. The persistency timeout is specified in minutes; otherwise the default

value of 5 minutes (300 seconds) will be used.

It is also possible to specify an optional persistency mask that determines the granularity of the source IP

match. For a persistency mask of 255.255.255.255 (the default) every single source IP address is considered

separately, whereas with a persistency mask of 255.255.255.0 only the leading 24 bits of the source IP
address are considered for the persistency i.e. any source IP address from this network is directed to the

same real server.

This feature can be used when clients are connected via Proxies, where the client source IP seen by the
balancer may vary between requests. The thought here is that the addresses used by the proxy fall in

contiguous ranges or blocks (e.g. class c net) that can be masked with the persistency mask.

In some situations (most notably from sites such as AOL) all requests from AOL users come from only one IP

address. This totally defeats IP source address based persistency, all requests from these so-called “Mega

Clients” would go to only one real server, while the rest of the server farm idles. This may be a reason to
employ the Layer 7 Persistency options.

H. Layer 7 Persistency

The LoadMaster supports several content-based persistency methods. These are normally termed layer 7
persistency methods because they use the content of the message to determine which real server should be

used. When using these methods, the LoadMaster intercepts a request and looks at the start of the message

from the client, using this information, the LoadMaster can determine to which real server the connection
should be routed.

These methods make most sense for protocols such as HTTP and SSL and are therefore explicitly supported by

the LoadMaster.

1. SSL session ID Based Persistency

The SSL (Secure Socket Layer) protocol is used on the World Wide Web to protect confidential information, by

performing authentication, data encryption and ensuring message integrity. The LoadMaster uses the SSL

session ID to ensure that all traffic for an SSL transaction reaches the same real server. This is a “common”
feature for commercial, financial and shopping-cart based web sites.

Note: This mode is not available if SSL acceleration is enabled. SSL in this case would be achieved via the

Real Server, which can lead to performance hits.

2. URL Based Persistency

Using this method, the LoadMaster will direct requests for the same URL to the same real server as long as

the persistency duration is valid.