Protocol Profiles

When you implement a Fast L4 profile, you can instruct the system to dynamically offload flows in a connection to ePVA hardware, if your BIG-IP system supports such hardware. When you enable the PVA Offload Dynamic setting in a Fast L4 profile, you can then configure these values:

• The number of client packets before dynamic ePVA hardware re-offloading occurs. The valid range is from 0 (zero) through 10. The default is 1.
• The number of server packets before dynamic ePVA hardware re-offloading occurs. The valid range is from 0 (zero) through 10. The default is 0.

The tcp-lan-optimized profile is a pre-configured profile type that can be associated with a virtual server. In cases where the BIG-IP virtual server is load balancing LAN-based or interactive traffic, you can enhance the performance of your local-area TCP traffic by using the tcp-lan-optimized profile.

If the traffic profile is strictly LAN-based, or highly interactive, and a standard virtual server with a TCP profile is required, you can configure your virtual server to use the tcp-lan-optimized profile to enhance LAN-based or interactive traffic. For example, applications producing an interactive TCP data flow, such as SSH and TELNET, normally generate a TCP packet for each keystroke. A TCP profile setting such as Slow Start can introduce latency when this type of traffic is being processed. By configuring your virtual server to use the tcp-lan-optimized profile, you can ensure that the BIG-IP system delivers LAN-based or interactive traffic without delay.

A tcp-lan-optimized profile is similar to a TCP profile, except that the default values of certain settings vary, in order to optimize the system for LAN-based traffic.

You can use the tcp-lan-optimized profile as is, or you can create another custom profile, specifying the tcp-lan-optimized profile as the parent profile.

The tcp-wan-optimized profile is a pre-configured profile type. In cases where the BIG-IP system is load balancing traffic over a WAN link, you can enhance the performance of your wide-area TCP traffic by using the tcp-wan-optimized profile.

If the traffic profile is strictly WAN-based, and a standard virtual server with a TCP profile is required, you can configure your virtual server to use a tcp-wan-optimized profile to enhance WAN-based traffic. For example, in many cases, the client connects to the BIG-IP virtual server over a WAN link, which is generally slower than the connection between the BIG-IP system and the pool member servers. By configuring your virtual server to use the tcp-wan-optimized profile, the BIG-IP system can accept the data more quickly, allowing resources on the pool member servers to remain available. Also, use of this profile can increase the amount of data that the BIG-IP system buffers while waiting for a remote client to accept that data. Finally, you can increase network throughput by reducing the number of short TCP segments that the BIG-IP^® system sends on the network.

A tcp-wan-optimized profile is similar to a TCP profile, except that the default values of certain settings vary, in order to optimize the system for WAN-based traffic.

You can use the tcp-wan-optimized profile as is, or you can create another custom profile, specifying the tcp-wan-optimized profile as the parent profile.

The tcp-mobile-optimized profile is a pre-configured profile type, for which the default values are set to give better performance to service providers' 3G and 4G customers. Specific options in the pre-configured profile are set to optimize traffic for most mobile users, and you can tune these settings to fit your network. For files that are smaller than 1 MB, this profile is generally better than the mptcp-mobile-optimized profile. For a more conservative profile, you can start with the tcp-mobile-optimized profile, and adjust from there.

This list provides guidance for relevant settings

• Set the Proxy Buffer Low to the Proxy Buffer High value minus 64 KB. If the Proxy Buffer High is set to less than 64K, set this value at 32K.
• The size of the Send Buffer ranges from 64K to 350K, depending on network characteristics. If you enable the Rate Pace setting, the send buffer can handle over 128K, because rate pacing eliminates some of the burstiness that would otherwise exist. On a network with higher packet loss, smaller buffer sizes perform better than larger. The number of loss recoveries indicates whether this setting should be tuned higher or lower. Higher loss recoveries reduce the goodput.
• Setting the Keep Alive Interval depends on your fast dormancy goals. The default setting of 1800 seconds allows the phone to enter low power mode while keeping the flow alive on intermediary devices. To prevent the device from entering an idle state, lower this value to under 30 seconds.
• The Congestion Control setting includes delay-based and hybrid algorithms, which might better address TCP performance issues better than fully loss-based congestion control algorithms in mobile environments. The Illinois algorithm is more aggressive, and can perform better in some situations, particularly when object sizes are small. When objects are greater than 1 MB, goodput might decrease with Illinois. In a high loss network, Illinois produces lower goodput and higher retransmissions.
• For 4G LTE networks, specify the Packet Loss Ignore Rate as 0. For 3G networks, specify 2500. When the Packet Loss Ignore Rate is specified as more than 0, the number of retransmitted bytes and receives SACKs might increase dramatically.
• For the Packet Loss Ignore Burst setting, specify within the range of 6-12, if the Packet Loss Ignore Rate is set to a value greater than 0. A higher Packet Loss Ignore Burst value increases the chance of unnecessary retransmissions.
• For the Initial Congestion Window Size setting, round trips can be reduced when you increase the initial congestion window from 0 to 10 or 16.
• Enabling the Rate Pace setting can result in improved goodput. It reduces loss recovery across all congestion algorithms, except Illinois. The aggressive nature of Illinois results in multiple loss recoveries, even with rate pacing enabled.

A tcp-mobile-optimized profile is similar to a TCP profile, except that the default values of certain settings vary, in order to optimize the system for mobile traffic.

You can use the tcp-mobile-optimized profile as is, or you can create another custom profile, specifying the tcp-mobile-optimized profile as the parent profile.

The mptcp-mobile-optimized profile is a pre-configured profile type for use in reverse proxy and enterprise environments for mobile applications that are front-ended by a BIG-IP^® system. This profile provides a more aggressive starting point than the tcp-mobile-optimized profile. It uses newer congestion control algorithms and a newer TCP stack, and is generally better for files that are larger than 1 MB. Specific options in the pre-configured profile are set to optimize traffic for most mobile users in this environment, and you can tune these settings to accommodate your network.

The enabled Multipath TCP (MPTCP) option enables multiple client-side flows to connect to a single server-side flow in a forward proxy scenario. MPTCP automatically and quickly adjusts to congestion in the network, moving traffic away from congested paths and toward uncongested paths.

The Congestion Control setting includes delay-based and hybrid algorithms, which can address TCP performance issues better than fully loss-based congestion control algorithms in mobile environments. Refer to the online help descriptions for assistance in selecting the setting that corresponds to your network conditions.

The enabled Rate Pace option mitigates bursty behavior in mobile networks and other configurations. It can be useful on high latency or high BDP (bandwidth-delay product) links, where packet drop is likely to be a result of buffer overflow rather than congestion.

An mptcp-mobile-optimized profile is similar to a TCP profile, except that the default values of certain settings vary, in order to optimize the system for mobile traffic.

You can use the mptcp-mobile-optimized profile as is, or you can create another custom profile, specifying the mptcp-mobile-optimized profile as the parent profile.

The TCP Profile provides you with multipath TCP (MPTCP) functionality, which eliminates the need to reestablish connections when moving between 3G/4G and WiFi networks. For example, when using MPTCP functionality, if a WiFi connection is dropped, a 4G network can immediately provide the data while the device attempts to resume a WiFi connection, thus preventing a loss of streaming. The TCP profile provides three MPTCP settings: Enabled, Passthrough, and Disabled.

You can use the MPTCP Enabled setting when you know all of the available MPTCP flows related to a specific session. The BIG-IP^® system manages each flow as an individual TCP flow, while splitting and rejoining flows for the MPTCP session. Note that overall flow optimization, however, cannot be guaranteed; only the optimization for an individual flow is guaranteed.

The MPTCP Passthrough setting enables MPTCP header options to pass through, while recognizing that not all corresponding flows to the sessions will be going through the BIG-IP system. This passthrough functionality is especially beneficial when you want to respect the MPTCP header options, but recognize that not all corresponding flows to the session will be flowing through the BIG-IP system. In Passthrough mode, the BIG-IP system allows MPTCP options to pass through, while managing the flow as a FastL4 flow. The MPTCP Passthrough setting redirects flows that come into a Layer 7 virtual server to a Fast L4 proxy server. This configuration enables flows to be added or dropped, as necessary, as the user's coverage changes, without interrupting the TCP connection. If a Fast L4 proxy server fails to match, then the flow is blocked.

An MPTCP passthrough configuration

When you do not need to support MPTCP header options, you can select the MPTCP Disabled setting, so that the BIG-IP system ignores all MPTCP options and simply manages all flows as TCP flows.