Technical data

ServerIron ADX Advanced Server Load Balancing Guide 47
53-1002435-03
Other transparent cache switching options
2
Resilient hashing for maximum cache persistence
In order to the maximize ‘cache-hit ratio’ which is a critical metric for architectures utilizing cache
servers, most vendors of application delivery controllers (ADC) utilize a mechanism that provides
persistence for traffic flows. When a new cache server is added to the mix for increased capacity
however or when a failed cache server is recovered from a failure event, then the traffic persistency
to cache is impacted. The network administrator then ends up either cleaning up the assembled
intelligence on traffic persistency or newly-available cache servers are not utilized until the next
available maintenance window.
Prior to software release 12.4.00a, the Brocade ServerIron ADX used a similar hashing mechanism
that required re-hashing of the hash table which resulted in the loss of persistency data with the
addition of new cache servers. Starting with software release 12.4.00a, the Brocade ServerIron
ADX is empowered with a new resilient hashing mechanism that allows the seamless addition of
newly-available cache servers while minimizing the impact on traffic persistency, which improves
the cache-hit ratio.
The new mechanism also reduces the burden of full hash-table sync-up between two
high-availability (HA) peer Brocade ServerIron ADX systems. Although this new mechanism doesn't
completely eliminate loss of data to ensure cache persistency for new flows, it drastically reduces
the impact by restricting the effect to a smaller sub-set of traffic flows.
The default settings for new resilient hashing mechanism are optimized to address common TCS
deployments. There can however be situations that require modification of the default settings to
improve cache re-use efficiency. Network configurations where you might want to alter the default
settings include: where a deployment has a very small number of cache servers or where the IP
addressing scheme of cache servers is causing non-optimal use of cache. In these situations, you
can alter the default settings to see if efficiency is improved.
The two settings that can be adjusted are described below:
Hash Size: This setting controls the overall size of internal hash. The default size is 4096. The
available settings are 2048, 4096, 8192 and 16384.
The following command is used to change the hash size.
ServerIronADX(config-tc-1)#resilient-hash hash-size 2048
Syntax: [no] resilient-hash hash-size <hash-size>
Hash Multiplier: This setting controls the granularity of hash usage against cache servers. The
default value is 60. The available settings are 20, 40, 60, 80 and 100.
The following command is used to change the hash multiplier.
ServerIronADX(config-tc-1# resilient-hash multiplier 100
Syntax: [no] resilient-hash multiplier <hash-multiplier>
NOTE
This mechanism is available only with Layer-7 TCS. The Layer-4 TCS utilizes the hash-mechanism
described in “Controlling traffic distribution among cache servers”.