Editor's note: This is part one of a two-part series examining the steps needed to unify wireless and wired networking....
See part two here.
In a world where wireless networks are critical infrastructure and wired desktops gather dust, many organizations need to refresh their legacy campus edge to better reflect the emerging status quo.
That said, wired and wireless networks still do not compete; each has its preferred use cases. A carefully considered deployment will mean that you can leverage the benefits of both. Even with the advent of technologies such as 802.11ac, wired still has superior per-client bandwidth. Conversely, the world hasn't seen much of an uptake on Ethernet adaptors for iPhones, so wireless utilization is clearly going nowhere but up. What's it all mean? Most organizations will need both for the foreseeable future.
Over the past several years, many organizations have been forced to hastily retrofit their existing networks with wireless. Access points (AP) were unceremoniously plunked on desktops or nailed to walls, without consideration for performance or capacity. To that end, the first step for a refresh is to audit the existing utilization of both wired and wireless devices. If you have upwards of 50 users attached to a single AP and a stack of Fast Ethernet switches with only a dozen ports "lit" in your wiring closet, then some fairly serious adjustment is required.
Survey the network to make sure you have accurate picture
Once you have a grip on actual utilization, it is worth investing some time and effort to perform a wireless survey to accurately map the optimum placement of APs. You can either hire a specialist contractor to conduct the search or use tools such as Fluke Networks' AirMagnet Survey or MetaGeek LLC's Chanalyzer and inSSIDer. Either way, a little planning will go a long way to create an accurate bill of materials.
When it comes to the congested 2.4 GHz band, the general rule of thumb is "fewer, more powerful" APs are better. It may be tempting to sprinkle cheap APs like fairy dust around the place, but the reality is that clients will have a tough job selecting the "best" AP and are thus more likely to have difficulty roaming. More APs also result in more beacon broadcasts, even for the same service set identifier (SSID). This eats into precious bandwidth, something that you are likely to run into even with a few hundred devices to support.
Once you have the optimum placement and quantity of APs, they'll need to be wired to the network. In most cases, this will be the local wiring closest already stuffed with alarmingly underused switches. Once again, an audit is necessary to understand what is actually being used. The devices still active will be a mix of the following:
- Docking stations for laptops;
- VoIP handsets (with or without wired desktops attached);
- Local printers and scanners; and
- Miscellaneous AV equipment such as Apple TVs or projectors.
It's likely the number of docking stations or wired workstations that have powered up in the past month is very low; these connections could be consolidated and not every wall socket needs to be connected. The reality is that 100% wireless coverage is probably more important than 100% wired coverage; that's because end users' perspectives on their "primary" device have changed. VoIP handsets are likely to remain relatively static; check the call logs to find out which wired handsets are actually used more than very occasionally.
The temptation is to move printers and scanners to wireless connections; however, the supplicants on many such devices are poor and difficult to troubleshoot. Furthermore, print traffic tends to be very bursty: long periods of idle followed by short exchanges of large amounts of data. Even with a Class of Service or Quality of Service policy in place, printer and scanner access is more reliably handled by a wired connection. Miscellaneous devices should also be audited and ideally standardized in terms of preferred connection method, with the primary concern support and maintenance.
Vendor selection, as always, is critical
If it’s not obvious, vendor selection is critical. When changing the campus edge, sticking with a single-solution vendor for both wired and wireless connectivity could save one's bacon. There are perhaps four or five major vendors that have both credible switching and wireless portfolios (i.e., one component is not provided by an OEM agreement). For the utopia of seamless transition between wired and wireless connectivity to be real, there are many dependencies that must work in concert: It's very easy for a single tuba to ruin the ensemble. When such support issues do occur, if you are able to quickly and easily reproduce it on the wired and wireless networks, it's much easier for the vendor to get a handle on it. Even if an issue only affects one type of connectivity, working with a single vendor gives technical support something concrete to work with and makes diagnosing the culprit(s) straightforward. This approach may mean you end up with a second-choice vendor for one aspect of the network, but ultimately, you should receive much better support in the event something breaks.
Selecting your core or distribution switching and VoIP provider, on the other hand, tends to be made independently of the campus edge. The core is likely to be a 10-year decision and voice is usually a different cost center altogether. To my mind, keeping these vendors distinct is the correct choice; looking for a vendor that plays in all four markets (wired, wireless, router and voice) leaves you with a single choice, which represents no choice at all.
Bringing it together to determine what's needed
So we know how many user devices we need to support and the quantity of Ethernet ports and APs. It's now possible to start sketching out some rough numbers in terms of what is needed on a per-floor basis.
The audit has probably shown that you need far fewer Ethernet ports per floor, and you need more APs than you have at the moment. Now would also be a good time to look at how everything is connected together. In many cases, remote wiring closets will be connected to the distribution/core networks via fiber, with one or two connections per switch. While this redundancy provides protection against a failed fiber or gigabit interface converter, it's not especially efficient. In Figure 1, we can see the "traditional" access and distribution core networks. I've picked arbitrary but common links to show how the edge is often connected to the core. Most network vendors are working toward collapsing two or three layers together, but the distinction is still useful for our purposes. The speeds and feeds of your own network will obviously differ, determined by the distance between your core and wiring closet(s).
Editor’s note: In the next part of this series, Kemp will disclose his advice for designs for an optimum network architecture to support a unified campus edge.
Glen Kemp is an enterprise solutions architect for a U.K.-based managed services provider. He designs and deploys network and application security tools, including access control, remote access, firewalls and other "keep the bad guys out" technologies. He is an experienced professional services consultant, delivering elephants and not hunting unicorns. His blogs can be found at sslboy.net and and at the Packet Pushers Podcast. Follow him on Twitter.