Colleague Phoneboy made some really insightful observations about my recent brief mention of A partial QoS approach for SMBs. I'm back home from several days in Chicago and want to thank Dameon for helping trigger me to cycle back to this important topic. It really raises several key issues.

SMBs have never been an important market to the high-end QoS community at large. Neither have end users. As Dameon pointed out - "QoS makes the assumption that there is more bandwidth than there is
bits going across it and that those bits need to be prioritized to get
an optimal experience."

The foremost point, that's too often overlooked is that Quality of Service (QoS) = OVERHEAD. Period. QoS means you must overlay some prioritization scheme, simple or complex, to your traffic. It's overhead somewhere. In packet payload as tags, in CPU processing cycles, in bandwidth utilization. Somewhere in your topology QoS adds overhead. The underlying assumptions is that your network is over-engineered to a degree that will support QoS in addtion to your regular traffic.

One approach to QoS has been gigabandwidth. Deploying significantly more bandwidth than necessary is certailny one for of QoS. The overhead in that case being the network pipe and hardware needed to deploy more bandwidth than necessary.

In our home networks today, we are often subscribing to more bandwidth that "required" for our critical applications. Dameon describes his home office network. It's not unlike mine. And he's absolutely right that it's not unlike many SMB networks. How many of us are truly approaching that maximum capacity threshold of our network today. The wide availability of broadband solutions, notably cable modem and DSL, provide more resources that most SMB, home office and sole proprietorships really require to conduct core business. A WiFi network within the LAn environment, or even a 100Mb wired LAN environment is, as they say, cheap, fast and easy to deploy.

Dameon's important note is that QoS is an end-to-end solution. To implement it effectively, you must control both ends and the middle. Once your service leaves your network, QoS is unknown or non-existent. There are a few exceptions that may apply. For example, some large enterprises use contracted WAN services from a carrier. These carriers may in fact negotiate contracts ensuring they will accept and honor Class of Service tags associated with MPLS for enterprise traffic. Those contracts are costly and not within the scope of SMB solutions.

QoS is about getting guarantees that your service is good enough, not about making service instantaneous and "perfect." That Valhalla is reserved for those organizations that have more money to spend on extraordinary measures in network design, not most of us.

Note: Dameon's post really just triggered more thinking on my part about the whole issue of QoS for SMBs. My thoughts here are not comprehensive and don't fully elaborate on his own thoughts. I encourage you to read his post. And think about what quality of service is good enough for your needs before proceeding down a complex path to over-engineering a solution that you may not require.


Technorati Tags: VoIP, Quality of Service, QoS, SMB