Colocation facilities – the hidden enabler of the Next-generation Enterprise WAN

Last time we covered the network benefits to the enterprise WAN of server virtualization. While WAN Virtualization is the newest technology of the key components of the Next-generation Enterprise WAN (NEW) architecture, and WAN Optimization the most recognized, here let’s cover perhaps the least well understood component of the NEW architecture: colocation. Colocation facilities, whether used by enterprises directly, by newer WAN services providers who leverage them, or both, are set to play a critical role in next-generation enterprise WANs because they hold the potential to lower the price/bit of WAN bandwidth, to help address several of the key WAN factors affecting application performance, including the “chattiness” issues of certain applications, and to improve application performance and predictability for both intranet applications as well as access to SaaS and cloud-based services.

As with our previous column on server virtualization, rather than a detailed summary of colocation, I will mostly focus on how colocation relates to the Enterprise WAN, and to the other key components of the NEW architecture.

A colocation facility, or colocation center (hereafter also referred to simply as “colo”) is a type of data center “where equipment space and bandwidth are available for rental to retail customers. Colocation facilities provide space, power, cooling, and physical security for the server, storage, and networking equipment.” While we’re about to cover how colocation can be used to create a superior Enterprise WAN, many companies are familiar with colocation as the place to put public Internet-facing servers such as websites, either done by a hoster or managed by the enterprises themselves. Beyond public-facing web servers, colos are beginning to be used by some enterprises for disaster recover / business continuity.

While some telecom service providers offer colocation capabilities, for the most part here we are talking about so-called “carrier-neutral” facilities, such as those offered by Equinix and InterXion, where there are a wide variety of network carriers from which to choose. In fact, most frequently these colo facilities are colocated with the public peering points which help make the Internet the network-of-networks that it is. Being at a colocation facility puts your equipment/application/service close to the “core of the Internet” from a connectivity, performance and network reliability standpoint.

Almost by definition, there is “lots” of fiber connectivity to a given colo facility, meaning that at a carrier neutral facility, there is a lot of competition providing bandwidth. The combination of competition and capacity means that the price of bandwidth available at a colo facility is much lower than the price almost anywhere else.

Bandwidth rates at colo facilities as low as $1 – $2 per Gbps per month for U.S.-based colos when purchased at high volume are not unusual. And while rates in other countries, especially in Asia and South America, can be a lot higher, they are almost always 1/10^th to 1/100^th the price of IP bandwidth in a given geography at a general business/office location, serviced by at most one or occasionally two fiber providers.

This is consistent with the well known history of data service bandwidth pricing: core bits are cheap, last mile bits are expensive (and even more so when the last mile is a telco monopoly), fiber bits are relatively cheaper, copper bits are relatively expensive, and partly because backhoes don’t obey Moore’s Law, the places where fiber comes together are where the cheapest bandwidth can be found. Colocation facilities eliminate the last mile monopoly “rent”, and have many fiber providers motivated to connect to the facility. DWDM has helped to give these core fiber links unbelievably large capacity. With cheap fiber and other players there, multiple ISPs want to locate there, and Internet IP bandwidth here is close to being a “commodity”. [Now, ISP bandwidth isn’t completely a commodity yet, especially for enterprise use, but we’ll get back to that later…]

So colocation facilities offer inexpensive bandwidth. That’s good, but by itself, not enough. They are also close to the core of the Internet, which means that it’s possible to connect to other locations – say, other colo facilities around the world – very inexpensively, very reliably (the core of the Internet being highly reliable, even if sometimes the edge is not) and at fairly low latency. They also tend to be close, in terms of latency, to large numbers of users/business connecting to the Internet. CDN vendors like Akamai have taken advantage of this fact to speed up access to cacheable content for public websites for a long time now.

While rack space is relatively expensive at a colo facility (versus housing servers and other equipment at your own facility), as we saw last time, the small physical footprint made possible by server virtualization enables enterprises to cost effectively take advantage of colocation facilities.

So colocation offers cheap bandwidth, a way to build inexpensive, reliable, stable and relatively low latency connectivity worldwide, and frequently offers low latency connectivity to enterprise locations.

However, the data center consolidation trend of the last many years, enabled/accelerated by WAN Optimization and server virtualization, has for the most part not yet pushed those data centers running internal corporate applications to colo facilities (save for the previously noted use for disaster recover / business continuity), largely because of lack of affordable reliable bandwidth for corporate connectivity. While MPLS is available at many colo facilities, it remains extremely expensive and does not offer a meaningful price advantage over MPLS service at a corporate data center.

Lacking a means to connect to those colos that is both reliable and cost effective has until now inhibited their use for this purpose. Not coincidentally, while SaaS has clearly had some success in certain market segments, and public cloud computing has found certain niches as well, the vast majority of enterprise applications, especially those where reliability of connectivity and/or performance predictability is critical, continue to run on the internal corporate WAN rather than getting migrated to SaaS and/or public or hybrid cloud computing platforms.

Next time, we’ll look further into the benefits that colo facilities can bring to the Enterprise WAN, if that last mile connectivity issue can be sufficiently addressed.

A twenty-five year data networking veteran, Andy founded Talari Networks, a pioneer in WAN Virtualization technology, and served as its first CEO, and is now leading product management at Aryaka Networks. Andy is the author of an upcoming book on Next-generation Enterprise WANs.