About autodiscovery settings
Ensure that the autodiscovery settings on peered server-side and client-side appliances match. If they do not, optimization might not properly function.
Enable Enhanced IPv4 Auto-Discovery enables enhanced autodiscovery for IPv4 and mixed (dual-stack) IPv4 and IPv6 networks. With enhanced autodiscovery, the SteelHead automatically finds the furthest SteelHead along the connection path of the TCP connection, and acceleration occurs there. For example, in a deployment with four appliances (A, B, C, D), where D represents the appliance that is furthest from A, the SteelHead automatically finds D. This feature simplifies configuration and makes your deployment more scalable. Enabled by default.
Enhanced autodiscovery peering is enabled by default. Without enhanced autodiscovery, the SteelHead uses regular autodiscovery. With regular auto-discovery, the SteelHead finds the first remote SteelHead along the TCP connection path, and acceleration occurs there.
This option uses an IPv4 channel to the peer SteelHead over a TCP connection, and your network connection must support IPv4 for the inner channels. If you have an all-IPv6 (single-stack IPv6) network, select the Enable Enhanced IPv6 Auto-Discovery option.
Enable Enhanced IPv6 Auto-Discovery enables enhanced autodiscovery for single-stack IPv6 networks. Disabled by default.
Enable Extended Peer Table enables support for up to 20,000 peers on high-end server-side SteelHeads The appliance’s data store maintains the peers in groups of 1,024 in the global peer table. We recommend enabling the extended peer table if you have more than 4,000 peers. By default, this option is disabled (the option is unavailable on models that don’t support it).
Before enabling this feature, you must have a thorough understanding of performance and scaling issues. When deciding whether to use extended peer table support, you should compare it with a serial cluster deployment. For details on serial clusters, see the SteelHead Deployment Guide. After enabling this option, you must clear the RiOS data store and stop and restart the service.
Enable Latency Detection enables peer appliances to pass through traffic without optimizing it when the latency between the peers is below the configured threshold. The latency threshold is in milliseconds and the default is 10 ms. The client-side SteelHead calculates the latency. When latency between peers is low enough, simply passing through unoptimized traffic can be faster than transmitting optimized traffic. When enabled, you can specify the Ignore Latency Detection flag in peer in-path rules, as needed.