Latency, Bandwidth, and Scale – The Span

4.3.2 Latency, Bandwidth, and Scale – The Span

The span of the network requirements for latency, bandwidth and scale which are needed to support traditional enterprise business applications and those needed to support cloud based applications can be very wide. Accurate forecasting of the quality of user experience and potential business impact for any network failure is already a major challenge for IT managers and planners even for today’s tra- ditional enterprise business applications. This challenge will become even more difficult as businesses will depend increasingly on more high performance cloud based applications which often have more variability than traditional enterprise business applications.

Meeting this challenge is essential to the “quality of experience” required to get the user community to accept a shared set of standardized services which are cloud delivered. Without this acceptance, the transformation of today’s data center to a private or hybrid cloud environment with the dynamic and shared infrastructure needed for a reduced total cost of ownership will be much more difficult. Some users may choose to “get around” IT by trying to leverage services from the public cloud without the proper integration with existing IT and business processes which can have significant negative impacts on their businesses.

“Quality of experience” for access to some cloud based applications and services may require LAN-like performance to allow a portion of the user community to use real-time information to respond instantaneously to new business needs and to meet the demands of their customers. For these use cases, latency and bandwidth matter. Furthermore, there does not have to be problems at any one hop in order for end to end performance to be affected. Mild congestion at a number of hops can create problems in latency and packet loss. Therefore, content distribution, optimized rout- ing and application acceleration services are usually required especially for hybrid cloud deployments with regional to global network connectivity.

76 G. Lin and M. Devine Other users may only want the ability to simply request a new service without

needing to know how or where it is built and delivered. It is not that the performance is not important to these users. In fact, communication and application delivery optimizations may still be required to increase the performance of applications and data that must traverse the cloud. It is just that their main criterion for quality has more to do with the ability to easily provision as many services as needed than it is dependent on latency and bandwidth optimizations. For these use cases, the abil- ity to easily provision system resources including network resources (physical or virtual) is essential. It is also important to note that the two most prevailing tech- niques to help server-side scaling, i.e. physical density and virtualization, both drive an increased dependence on network integration.

For each user group and their corresponding use cases, the evolution to global- scale service delivery may best be accomplished via a hybrid cloud environment. The hybrid cloud can enable the visibility, control and automation needed to deliver quality services at almost any scale by leveraging not only the private network but also the public internet via managed network service providers.

Public clouds can be used for off-loading certain workloads. This off-load could

be so that the private network infrastructure can be available and optimized for other latency and bandwidth sensitive workloads and/or for the provisioning of additional services due to a shortage of available infrastructure on-premise. Application plat- forms and tooling available on the public cloud can also be used to provide even greater flexibility for development and test environments which are often the best workloads for this type of off-loading. SaaS applications can also be consumed by the user community within a hybrid cloud environment. Under the hybrid cloud model, the consumption of public cloud services can be fully integrated with the existing on-premise IT and business processes to maximize the return of investment as well as ensure regulatory compliance.