ZXTM performance tuning

Trying to tweak ZXTM to get better performance? Here are a few suggested configuration changes to optimize ZXTM for performance.

Global settings page

• System settings

  Increase chunk_size to 32768 or 65536. This should improve raw throughput performance, especially for larger responses.

  Increase listen_queue_size - possibly might help for very high request rates.

• Connecting settings

  Increase max_keepalives and max_keepalives_pernode

• SSL configuration

  Increase ssl!cache!internal!size and external!size may help.
  It all depends on how much SSL session re-use there is.

Virtual server settings

HTTP: Set add_cluster_ip to No. This may stop nodes from logging source IP addresses correctly, however.

Increase client and server buffer size. This will allow ZXTM to buffer more data from the network. This is especially helpful when
using Apache web servers.

Increase connect_timeout to 60 seconds, if you want to improve performance at the cost of reliability

If you are running an exceptionally busy HTTP site, and ZXTM is not performing any sophisticated HTTP manipulation, you can change the protocol to 'client_first' for a speed improvement. Use with care!

Pool settings

Set max_connect_time to 60 seconds, if you want better performance but reduced node failure detection

Similarly, turn off all health monitoring for a slight performance advantage (but again, you are losing reliability)

See also

Tuning ZXTM for maximum performance

Ben [Zeus Dev Team] 01 July 2005