has been on my mind lately.
I suppose there are a number of reasons.
We’re in the process of re-architecting our data network at York
. We’re starting off by adding redundancy in various ways, and anticipate the need to address QoS in preparing for our future deployment of a VoIP service.
Of course, that doesn’t mean we don’t already have VoIP or VoIP-like protocols already present on our existing undifferentiated network. In addition to Skype, there are groups that have already embraced videoconferencing solutions that make use of protocols like RTP
. And given that there’s already a Top 50 list of Open Source VoIP applications
to choose from, I’m sure these aren’t the only examples of VoIP-like applications on our network.
At the moment, I have more questions about QoS than answers.
- If we introduce protocol-based QoS, won’t this provide any application using the protocol access to a differentiated QoS? I sense that QoS can be applied in a very granular fashion, but do I really want to turn my entire team of network specialists into QoS specialists? (From an operational perspective, I know I can’t afford to!)
- When is the right time to introduce QoS? Users are clamoring for QoS ASAP, as it’s often perceived as a panacea – a panacea that often masks the root cause of what really ails them … From a routing and switching perspective, do we wait for tangible signs of congestion, before implementing QoS? I certainly have the impression that others managing Campus as well as regional networks plan to do this.
- And what about standards? QoS isn’t baked into IPv4, but there are some implementations that promote interoperability between vendors. Should MPLS, used frequently in service providers’ networks, be employed as a vehicle for QoS in the Campus network context?
- QoS presupposes that use is to be made of an existing network. Completely segmenting networks, i.e., dedicating a network to a VoIP deployment, is also an option. An option that has the potential to bypass the need for QoS.
I know that as I dig deeper into the collective brain trust answers, and more questions, will emerge.
And even though there are a number of successful deployments of VoIP that can be pointed to, there still seems to be a need to have a deeper discussion on QoS – starting from a strategic level.
As I reflect more and more on QoS I’m thinking that a suitably targeted BoF, at CANHEIT 2008
for example, might provide a fertile setting for an honest discussion.