Avaya P333R-LB Manuel D’Utilisation
Chapter 14
Load Balancing in the P333R-LB
42
Avaya
P333R-LB User’s Guide
Client persistency is achieved either by using naturally persistent load balancing
schemes (such as Hash or MinMiss Hash), or by forcing persistent load balancing
decisions on non-persistent load balancing schemes (such as Round Robin).
Decision forcing is performed by storing the history of the latest decisions in a cache
for a limited time, and sending the packets to the appropriate server based on
previous load balancing decisions.
Regardless of the client persistency nature of the selected load balancing metric, the
P333R-LB offers a unique client persistency feature that is available in all load
balancing metrics. Client persistency is based on a "persistency cache". Load
balancing decisions are recorded in a persistency cache for a specified time
configured by the user. When a new session that matches an entry in the persistency
cache is processed by the P333R-LB, it is directed to the same server pointed by the
cache (provided, of course, that the server is considered healthy).
The key to the persistency cache is based on the client IP, in combination with a
wildcard. This allows persistency to be configured per an exact IP address, or per a
group of addresses. For instance, in cases where clients hide behind a NAT device
which selects NAT addresses from an address block of 255 addresses, enabling the
persistency cache with a wildcard of 0.0.0.255 will map all clients to a single entry
and a single Real Server.
schemes (such as Hash or MinMiss Hash), or by forcing persistent load balancing
decisions on non-persistent load balancing schemes (such as Round Robin).
Decision forcing is performed by storing the history of the latest decisions in a cache
for a limited time, and sending the packets to the appropriate server based on
previous load balancing decisions.
Regardless of the client persistency nature of the selected load balancing metric, the
P333R-LB offers a unique client persistency feature that is available in all load
balancing metrics. Client persistency is based on a "persistency cache". Load
balancing decisions are recorded in a persistency cache for a specified time
configured by the user. When a new session that matches an entry in the persistency
cache is processed by the P333R-LB, it is directed to the same server pointed by the
cache (provided, of course, that the server is considered healthy).
The key to the persistency cache is based on the client IP, in combination with a
wildcard. This allows persistency to be configured per an exact IP address, or per a
group of addresses. For instance, in cases where clients hide behind a NAT device
which selects NAT addresses from an address block of 255 addresses, enabling the
persistency cache with a wildcard of 0.0.0.255 will map all clients to a single entry
and a single Real Server.
Selecting a Load Balancing Metric and Persistency Option
The different load balancing metrics and persistency options provide you with
flexibility.
Round Robin generally gives you the best load balancing solution.
Hash, with the key set to Src IP, gives a predictable metric, required when client
persistency is needed and a redundant load balancer is used.
MinMiss Hash, with the key set to Src IP, gives the best persistency, without using
the persistency cache.
The persistency cache mechanism enables you to specify the required client
persistency time.
The wildcard feature allows the client to maintain persistency when the same client
is mapped to a different IP address by a NAT device. However, persistency cache
size is limited, and it somewhat increases the session processing overhead.
flexibility.
Round Robin generally gives you the best load balancing solution.
Hash, with the key set to Src IP, gives a predictable metric, required when client
persistency is needed and a redundant load balancer is used.
MinMiss Hash, with the key set to Src IP, gives the best persistency, without using
the persistency cache.
The persistency cache mechanism enables you to specify the required client
persistency time.
The wildcard feature allows the client to maintain persistency when the same client
is mapped to a different IP address by a NAT device. However, persistency cache
size is limited, and it somewhat increases the session processing overhead.
Port Re-mapping
The P333R-LB may be configured to re-map the destination port number when
performing load balancing. For example, you might want to run the HTTP process
on real servers using a different port number in order to enable more than one
concurrent HTTP service.
performing load balancing. For example, you might want to run the HTTP process
on real servers using a different port number in order to enable more than one
concurrent HTTP service.