Published: 18 Feb 2014
Moving applications to the cloud has a dramatic effect on networking requirements, which can affect cloud computing costs. "Pay for what you use" is a popular and seductive selling tool for cloud providers, as CIOs often believe that they are paying for unused capacity, infrastructure and software that the organization does not need. Cloud's consumption pricing means that bandwidth is charged per megabyte, network appliances are charged for CPU and memory consumption, and logging requires ever-increasing amounts of storage. The result is that the total monthly charges of a cloud service can vary significantly from month to month.
Networking infrastructure and cloud computing
Part 1: Is networking infrastructure the Achilles' heel of cloud computing?
Part 2: Reducing network latency means focusing on location, location, location
Part 3: Networking costs swell pay-per-use public cloud bills
When building out a cloud environment, don't be lulled into thinking that the level of cloud resources you use today will be the same as those you use tomorrow. That changing metric will be reflected in the month-to-month price you pay for cloud services.
Compare the pay-per-use model with existing networking budgets, which have known costs. An IT manager can easily predict future IT needs, and thus budget. This is a good thing, since many internal accounting systems will not permit or cope with variable charges.
Organizations that move to the cloud will need to do a lot of work to develop systems to predict and manage ongoing costs. Consumption-based pricing requires constant monitoring to ensure that plans and actual costs are correct. The ensuing weekly cost-control meetings can cut into useful IT activities that would have improved service and functionality in other areas.
'Data gravity' increases cloud costs over time
Data gravity refers to the fact that data attracts more data. Consider that a firewall collects a lot of log data. This log data will grow over time. A server application will need backups, which will increase the amount of stored data over time.
Inevitably, this data must be moved between locations, perhaps to a long-term archive. Or the business might decide to store some of the existing backups in the cloud to comply with an off-site backup policy.
As data volumes grow, this creates even more load and network traffic. This cost is hidden in a private data center since the bandwidth, either Internet or WAN, is prepaid. But in a cloud environment, bandwidth consumption and its related costs will consistently increase over time because of data gravity. The amount of log data stored will increase and incur even more charges. If you plan on storing the log data in some other location, then network charges will be incurred to move the data. Data gravity is an ever-expanding problem and cost.
Hidden network costs turn up on your cloud bill
Cloud service providers market availability via the Internet as a feature enabling employees to work from home or on the road. This creates a major data security challenge. In particular, the network team is often responsible for policing bring-your-own-device (BYOD) programs. It must also manage VPNs and policy gateways, so controlling connectivity may incur unexpected costs for network appliances.
Another less obvious consideration is the level of Internet security inspection. Performing these tests extensively will increase network and application latency. Since such checks are for internal consumption, the correct security policy is to bypass these cloud services or classify them as "trusted." When configuring security for cloud applications, don't give in to overzealous security posturing.
Finally, don't expect to save much money on network firewalls and routers, which are key components of most IaaS deployments. So far, vendors are not discounting software appliance costs, and yearly software maintenance is still required.