The chaotic data center LAN

Opinion
Apr 27, 20112 mins

As mentioned in the last newsletter, LAN design used to be quite staid. Now it is anything but staid. A good example of the dynamic nature of LAN design is the data center LAN. As we will explain in the next two newsletters, it is only a bit of hyperbole to say that everything that you once knew about designing a data center LAN needs to be re-examined in light of shifts in demand, technology and the vendor landscape.

In terms of design guidelines for data center LANs, in the late 1990s the industry settled on a three-tier architecture for large data center LANs that was comprised of access, distribution and core switches. This architecture was not created because somehow having three layers of switching was intrinsically correct or because that architecture provided real value. It was created in large part because the LAN switches of that era had relatively low port density and it was not possible to design a large data center LAN with fewer than three tiers. While the design certainly worked, it was inefficient and complex in part because in addition to having switches switch traffic between servers, there were many switches in the typical design whose sole role was to switch traffic between switches.

Juniper vs. Cisco

Juniper is a good example of how the data center LAN environment has changed over the last decade. Unlike vendors such as Cisco and HP, Juniper is a relative new comer to LAN switching. In fact, Juniper introduced its first LAN switches to the market just over three years ago. More recently, Juniper announced what it refers to as a new “3-2-1” data center network architecture. The goal of that architecture is to flatten and simplify legacy data center networks, reducing them from three to two layers today, and to a single layer in the future.

Juniper is certainly not the only vendor to be talking about flattening the data center LAN. In the next newsletter we will continue this discussion and will include how the traditional assumptions and LAN design guidelines are giving way to a new set of assumptions and guidelines.

Jim has a broad background in the IT industry. This includes serving as a software engineer, an engineering manager for high-speed data services for a major network service provider, a product manager for network hardware, a network manager at two Fortune 500 companies, and the principal of a consulting organization. In addition, Jim has created software tools for designing customer networks for a major network service provider and directed and performed market research at a major industry analyst firm. Jim’s current interests include both cloud networking and application and service delivery. Jim has a Ph.D. in Mathematics from Boston University.

More from this author