Private-cloud projects will put intolerable strains on existing networks and punish the ill-prepared, says Lori MacVittie.
Any IT organisations thinking about a private cloud inevitably have to examine the health of their network and its readiness to support the stresses that accompany such implementations.
As with Web 2.0, cloud computing inherently increases the use of the network — from the increase in application packaging due to virtualisation to cloud computing's reliance on collaboration and integration across the entire infrastructure.
While these changes may lead to re-engineering of processes and a new architecture for the core network to improve its general performance and reliability, they can also lead to increased stress on every layer of the network, and impair the utilisation of existing and new application deployments.
Operational efficiency
There is no way of avoiding the increase in network utilisation that comes from a private cloud-computing implementation. Not only are there generally more devices communicating with one another, but they also do so more often — at least in a well-designed cloud implementation that exploits automation and orchestration for operational efficiency.
This increased utilisation leads to greater stress on the core network as well as on individual infrastructure components, which must manage the extra traffic and more frequent changes in configuration and policy management caused by the altering application infrastructure.
If the network itself is not capable of providing the same level of service to customers while handling this additional communication and traffic, it can effectively die. It can leave operations and customers frustrated by poor response times for applications and the infrastructure.
A cloud-computing implementation can quickly become a Catch-22, in which customer requests and infrastructure communication create conflicting demands on performance.
Implementation strategy
The key to improving your network with a cloud-computing project is having a clear implementation strategy that includes evaluation of existing operational processes and the ability of the core network to support increases in traffic flow.
Core network readiness is important. It may be the case that you will need to upgrade bandwidth — from 100Mbps to 1Gbps or 10Gbps, depending on the size of your datacentre. And while you are examining each infrastructure component for readiness, it is an appropriate time to look at consolidation possibilities in the infrastructure.
Can you combine that caching solution with your load balancer? What about the intrusion prevention system (IPS) and the host-based IPS? Do you need both or can you manage with one? Streamlining the flow of data through the core network can drastically improve performance of applications and reduce the impact of integration by eliminating unnecessary touch points in the network.
Evaluating the operational processes tied to infrastructure components and delivery of applications is a critical factor in whether cloud computing will make or break a network.
By closely evaluating those processes before codifying them with automation and orchestration, you can identify and remove redundancies and discover more efficient methods of accomplishing necessary tasks.
This process reduces the strain placed on infrastructure in general while handling a higher volume of change, and improves the efficiency of the overall infrastructure.
Changing environment
Without a plan, it is likely that the implementation of a private cloud-computing environment will kill your network. Perhaps it will not die immediately, but eventually the strain of a constantly growing and changing environment will drive components into the ground and failures will occur.
A strategy to address the readiness of your network is necessary to ensure the benefits of cloud computing will be enjoyed by your organisation.
Implementing a strategy that considers physical and operational readiness will go a long way towards ensuring a cloud-computing implementation will cure your network, not kill it — certainly something to consider when developing cloud-computing strategies in 2010.
Lori MacVittie is responsible for application services education and evangelism at application delivery firm F5 Networks. Her role includes producing technical materials and participating in community-based forums and industry standards organisations. MacVittie has extensive programming experience as an application architect, as well as in network and systems development and administration.
