Today we dive deeper into the differences between centralized and distributed performance optimization solutions. The technical functions that Centralized and Distributed ADS solutions perform are often very similar—even identical. But there are major differences in the types of traffic that benefit and the span (or reach) of the two solution types. Here are two pictures to help you understand the difference.
The solid red lines in the pictures show the span of influence for Centralized and Distributed ADS techniques. In both approaches control and acceleration techniques can only be applied to traffic the devices directly process (in-line) or monitor (through a switch-mirrored port).
Centralized ADS solutions support the specific servers they front, and can only influence applications based on open standards such as Web (HTTP, HTML, Java), XML. Despite the growth of these standards-based applications, a recent NetForecast survey indicates that Web-based traffic represents only about one quarter and real-time traffic one fifth of a typical enterprise’s network load. This leaves just over half the traffic in the typical enterprise untouchable by Centralized ADS solutions.
Centralized ADS solutions exert influence over traffic traveling outbound across the WAN to Web browsers, and inbound across a LAN to the servers. Centralized ADS solutions cannot influence inbound WAN traffic because that is the realm of the browser.
Distributed ADS solutions, on the other hand, exert influence over outbound as well as inbound traffic traversing the WAN-LAN boundary at the data center and at remote sites. Therefore, these solutions can affect all WAN traffic (including Web, traditional client-server, UDP, and VoIP), but they cannot influence traffic that remains within the data center or within remote sites.
The dual-ended nature of most Distributed ADS implementations enables stronger versions of the same performance techniques used in Centralized ADS solutions. For example, distributed TCP optimization can leverage six separate TCP flows (two between the server and Distributed ADS, two between the Distributed ADS devices, and two between the remote device and the client), while centralized TCP optimization can only influence three TCP flows (two between the server and the Centralized ADS device, and one outbound from device to the browser).
Finally, as the dashed green lines in the picture illustrate, some Distributed ADS solutions can leverage peer-to-peer information exchange among remote devices and/or share a database of knowledge at the data center.
Distributed ADS solutions are typically dual-ended or synchronous, which means they operate as a pair. However, just to complicate matters, some Distributed ADS techniques—primarily control functions—can also operate alone. For example, a single distributed device is often used to manage college campus traffic to and from the Internet to prevent music and video downloads from swamping other applications. Although single-ended distributed implementations exist, we will focus on the more ubiquitous dual-ended solutions.
In contrast to Distributed ADS solutions, all Centralized ADS solutions are single-ended or asynchronous. This means they operate as standalone devices that do not rely on communications with peers.
