If you're an IT professional you've probably been hearing a lot about cloud computing lately. I know I've sat through a number of seminars and sales pitches where people have been touting public cloud services on the merits of lower cost, reducing infrastructure and quicker implementation of services. However, I've noticed that almost none of these presentations discuss the increased reliance on Internet connectivity. With all the focus on the benefits of cloud computing, it's easy to forget that there has to be a trade-off. In order to offer reliable, quality access to public cloud services, your Internet connectivity likely needs some tuning.

Below are some areas that should be investigated.


Unlike the data center where traditional applications are hosted and where everything right from the server NICs all the way up into the core of the network is redundant, a corporation's Internet connectivity often times is not engineered for such high availability. Connectivity is often delivered from a single service provider and is terminated on a single router or firewall. This creates a number of single points of failure. An Internet outage used to mean that people couldn't read the news or maybe some email messages got delayed. With services in the cloud, it now means lost productivity at best and lost revenue at worst. Treating your Internet connectivity as a non-critical service might not be an option any more; it needs to be highly available.

Maintenance and Support

A key side component to high availability is maintenance and support. If you have a device or a service go down, how quickly can you get a hold of someone to start fixing it? What sort of service level agreements do you have with your hardware vendors and service providers? If you're a smaller shop and don't have full-time IT support, how quickly can you get your IT consultants engaged?

This all ties back to what I mentioned above about treating your Internet connectivity as a critical service. Maintenance contracts need to be in place on all equipment and SLAs with service providers and consultants need to be well documented and understood.


The next big consideration is capacity. Can your Internet connection handle the expected number of users who will be accessing the cloud services? What sorts of data will be moving back and forth? Any large files? Do you have the bandwidth to spare? Do you have enough bandwidth in the upstream direction? Is the application interactive? What kind of latency is typical on your Internet connection?

An analysis needs to be done to determine the bandwidth usage and latency characteristics on your Internet connection. That needs to be compared with the estimates for what's required by the cloud services. If there's a shortfall, plans need to be drawn up to increase capacity where needed. Keep in mind that unlike traditional software, cloud services don't have much in the way of "system requirements". Don't expect to see cut and dry recommendations about bandwidth or latency. It's likely you will need to do your own calculations.


Another area that is often overlooked when sending and receiving data with cloud services is the confidentiality and integrity of that data. Is the data being sent back and forth being encrypted or digitally signed? Although this isn't really a network layer problem, it might be up to you as the network person to raise these questions because, as the network person, you are more likely than the application people to understand how easy it is to sniff clear text data and the risk that poses.

Bandwidth Cost

There are basically two billing models when it comes to Internet bandwidth:

  1. A monthly allowance for how many bytes you can transfer where you pay extra for overages. This is typical for cable and DSL connections.
  2. A monthly committed rate, billed on the 95th percentile, where you pay extra if your rate was higher than your commitment for the month. This is typical in business class services such as those delivered by your hosting provider.

In the first model you have to be concerned with how many overall bytes are transferred during the month. This model is particularly susceptible to large file transfers as they can quickly eat into your base allowance and cause you to start paying extra.

In the second model it doesn't matter how much data is transferred, it matters at what rate the data is transferred. You can move big files back and forth all day and will only be charged extra if those transfers force your billable utilization to rise above your committed rate.

It may be tough to estimate ahead of time how your bandwidth costs will change but it's important to realize that they can change and to make some allowance for it.


IT architects should carefully consider their Internet connectivity if cloud services are being considered by their organization and/or customers. Without appropriate Internet connectivity, access to cloud services, and therefore the benefits that they can provide, are put at risk.