Network Computingwww.networkcomputing.com

Related Links

ArrowPoint CSS-100 Switch: Layer-by-Layer Load-Balancing, Sneak Previews, November 15, 1998

Other Workshops this issue

Java Connectivity With JDBC By Ahmad Abualsamid Understanding Internet Payment Protocols By Brian Walsh

Company Directory

Browse our directory to get data, starting with a particular company.

Reader Service

Allows you to request additional product information from our advertisers.

Print The Full Article

ClickHere

E-mail this URL

ClickHere

Buy the Book

One of the most frequently used--and most often misrepresented--terms is Layer 4 switching. No two vendors define it the same way, but the bigger the number the better, right? In this workshop we'll define Layer 4 switching and clarify vendors' numerous takes on the term.

To understand Layer 4 switching, you must delve into the world of the OSI model. Layer 4 switches operate at the transport layer of the TCP/IP stack (see ", The OSI Model. Layer 2 switches operate at the MAC (Media Access Control) layer--a sublayer of the Data Link layer--forwarding packets based on the MAC address of the destination host. Layer 3 switches operate at the network layer, switching packets based on IP address.

Logically, Layer 4 switches must operate at the UDP (User Datagram Protocol) and TCP level, making switching decisions based on information held in the transport layer. But, in fact, not all Layer 4 switches transfer packets based on this information. This is where the Layer 4 confusion begins.

Web Load-Balancing Alteon Networks coined the term Layer 4 switching. Its ACESwitch performs Layer 2 switching based on MAC address. But Alteon's solution also can dig deeper and forward packets destined for the same MAC address to different ports based on UDP or TCP port number. This lets the unit perform service-based load-balancing.

In this model, additional accounting must be performed at Layer 3 (the IP address) to ensure that each client session goes to the same Web or FTP server. With the switch as a front end to the Web server, the end user can be unaware of multiple Web or FTP servers on the back end. Because the switch is making load-balancing decisions based on the TCP port number of the incoming packet, Alteon dubbed this process Layer 4 switching.

With this method, all servers on the back end must have identically mirrored content. As long as this is the case, the scalability of the virtual Web server is limited only by the number of ports on the Layer 4 switch. Likewise, if one of the servers were to go down because of hardware failure, the Layer 4 switch can stop forwarding packets to it, adding a redundant element to Web and FTP hosting services.

Web load-balancing is just one possible application for Layer 4 switching. Alteon was among the first vendors to support this application, but more than 12 vendors now offer competing products. Most offer additional intelligence to help balance traffic better.

For example, instead of simply providing round-robin distribution of traffic, these products monitor server load and response time, and switch incoming requests to the least loaded server. Network Computing has tested several of these products. You can find our most recent tests, "Seven Web Load-Balancers Score With Round-the-Clock Access," at www.networkcomputing.com/913/913r2.html.

Access Control In marketing terms, Layer 4 switching has come to mean much more than Web load-balancing. Vendors with a significant presence in the server load-balancing arena--Foundry Networks, for example--also compete in the enterprise switching space. Naturally, these vendors use Layer 4 switching as a marketing point for their enterprise-class devices. As a result, other enterprise switch vendors have picked up the lingo. Here Layer 4 switching takes on a different meaning.