Riverbed Technology - WAN Optimization

WAN Optimization

Steelhead Appliances
Riverbed's main product is the Steelhead Appliance which is a box that is connected to a wide area network, often behind the WAN router and in front of the Local Area Network switch. Each Steelhead appliance interacts with one or more Steelheads in other locations across the WAN to increase the efficiency and performance of network traffic. Steelhead appliances work together to reduce network traffic in three ways:

1. Data Streamlining: Primarily, data streamlining involves the removal of redundant data from WAN traffic. In place of redundant data, the Steelhead sends a reference to the redundant data that has already been stored on the remote Steelhead appliance, which can then insert the proper data back into the data stream. The result is a significant reduction in data that gets sent across the network. Steelhead appliances recognize and eliminate duplicated data regardless of the direction in which it is sent, the user who sends it, or the application that sent the data.
2. Transport Streamlining: TCP works inefficiently on networks where latency is high. Transport streamlining combines the use of TCP options and Riverbed proprietary technology to reduce or eliminate the impact of high latency, limited TCP window size, jitter, packet loss, and out-of-order packet delivery in order to improve overall data transfer performance and throughput. Transport streamlining also includes the option to introduce alternative transport layer protocols, including High-Speed TCP, MX-TCP, and TCP-Westwood.
3. Application Streamlining: Applications such as Microsoft Exchange, NFS and CIFS require complex and chatty interactions between clients and servers, and these chatty interactions result in slow performance in high latency network environments. Some application protocols may require hundreds or thousands of round-trip interactions in order to transfer a relatively small amount of data. Application streamlining reduces the number of round-trips over the WAN to reduce or eliminate the impact of high latency. Application streamlining also addresses any application-specific data encoding or encryption mechanisms so that data deduplication mechanisms can be applied on the original data format.

The software that runs a Steelhead appliance is called the Riverbed Optimization System or RiOS. The current version of RiOS is 7.0. There are more than 20 different models of Steelhead appliances available, each with a different capacity for connections and data traffic, ranging from a small desktop unit up to a high-end 3U data center model with solid state disks.

Many Steelhead Appliance models support the Riverbed Services Platform (RSP), which enables them to run as many as five virtual machines in VMware Server directly on the Steelhead appliance.

Virtual Steelhead
While a standard Steelhead Appliance is a physical device, Virtual Steelhead provides the same RiOS-based WAN Optimization functionality in a VMware ESX or VMware ESXi environment, without requiring a dedicated physical device.

Steelhead Mobile
A third option for running RiOS is directly on a portable PC or Macintosh. Steelhead Mobile runs directly on the computer, and optimizes any WAN traffic that goes back to a data center where there is a Steelhead Appliance it can pair with.

Central Management Console (CMC)
The CMC simplifies the process of deploying, configuring, and managing Steelhead appliances. Administrators can manage, deploy, configure, update, and monitor as many as 2,000 Steelhead appliances or Virtual Steelheads from one web-based interface. Deploying, configuring, updating, and monitoring are all done through one web-based interface. The CMC is available either as a physical or a virtual device.

Interceptor
The Interceptor appliance extends the scaling and high-availability capabilities of Steelhead appliances to meet the requirements of the largest and most complex enterprise networks and data center environments. It clusters up to 25 Steelhead appliances so they can work together seamlessly to scale out to a million TCP connections and up to 12 Gbit/s of throughput.