Enterprise clouds: The secret is the workload

The cloud is the long-term strategic delivery model that will enable IT as a service while reducing virtual sprawl in the datacenter -- if firms design, build, and operate it correctly

IT has struggled to solve the business needs that have surfaced due to a mismatch in service expectations between business and the delivery that IT has accomplished. IT is being forced to restructure itself by offering IT as a service, which changes the mind-set of the datacenter as an infinite resource to the datacenter as a utility, with service levels tied to costs that the business can relate back to growth. The business cannot do that now, and the reasons are worth explaining.

There are a variety of applications that support numerous business lines across an enterprise. These applications exhibit a wide range of operational characteristics as they service the diverse business demands. That diversity is the key to business success, but it has consequences.

Ask any senior enterprise application architect about how IT should run a datacenter and you will find that the range of behaviors exhibited across the portfolio of applications cannot be run on one or even a small set of standard platform configurations. They will emphasize that the most critical applications often have extreme operational requirements that require specialized adjustments to operating environments. Ask any datacenter manager about accommodating those different application behaviors and they will tell you that the proliferation of platforms makes it impossible to contain costs, manage complexity, or maintain reliability, especially for those critical applications. Ask any business executive about how business is running, and they will tell you that applications cost too much to develop, run, and maintain. They cannot correlate growth to investment in IT, and that is a very dangerous state of affairs for any business.

Offering IT as a service requires a paradigm shift in technology, operational processes, and thinking about how business demand, manifested by applications, affects the use of all datacenter resources. Enterprise cloud computing enables IT as a service and is the next phase in datacenter evolution to solve this spiraling dilemma. The question is, how does business successfully operate this very expensive investment?

The enterprise cloud combines a series of evolving and maturing technologies with associated techniques that make it easier to leverage the power of distributed processing while minimizing the complexity and difficulty associated with such distribution. The location of compute, network, and storage resources required to operate applications becomes transparent through virtualization and dynamic allocation. Applications are serviced by pools of infrastructure services that do not have to be dedicated to an application. This fluidity is supported by advances in resource monitoring, advanced security, and faster networking. The promise of cloud computing is providing a significantly improved user experience, while balancing cost and efficiency -- three critical pillars that must be satisfied for a business to succeed. But IT promises rarely live up to the hype, and the reason is simple: IT typically builds such projects with little input from the business on growth and general usage.

A successful cloud meets the needs of the business

The cloud, even with all its new technologies, is not magic. It is an operating environment that needs to be designed to perform optimally. The design must start top-down with business requirements, taking into account:

  • How does the business operate?
  • What are the peaks, cycles, and key operating drivers?
  • How vital is availability or processing massive throughput?

In a sufficiently large enterprise, the answer will be yes for many requirements questions, but not for all the applications, nor for all lines of business. The problem has been that in the absence of that knowledge, the answer has been to design from the bottom up, resulting in overengineering the operating platform, which increases cost and complexity while causing datacenter sprawl and reducing reliability. The consequences of overengineering are far ranging, as it is not just the addition of servers, but also the increased costs and complexity that result from the need for more power, cooling, cabling, floor space, and support personnel. Despite all these extra costs, it is not unusual for performance and reliability to suffer, a businessperson's true nightmare.

Understanding demand

There is nothing more important than properly servicing business demand, so the questions are how to do so and what happens once you understand that demand?

Despite the large variety of applications that exist, they all can be categorized into a small set of workload characteristics. Initially this will seem counterintuitive to many, but that is why characterizing actual demand is necessary. The brief workload-type descriptions below cover a wide range of applications running in the enterprise. Some applications will exhibit multiple workload-type behaviors.

Numerical processing intensive — Applications that process large data sets from many sources (i.e. dozens of sources, millions of records, gigabytes of data), involving iterative calculations, data transformations and data driven matching. These applications typically run for many hours and must complete in tight operational time frames (i.e. statement generation and rendering, risk calculators, general ledger, etc.).

Request/response — Applications that provide a large population of users' business intelligence by making access to disparate data sources transparent.

Event driven — Users are presented with data changes every second that affect models they are running. Users must make decisions frequently based on those models (e.g. trading environments).

High concurrency/high throughput — Typified by Internet-facing applications that must be responsive to massive fluctuations in demand while providing rich media content to end-users.

Ubiquitous user — Typified by the need to allow a user multiple presences through different media (Webcasts, telepresence, etc.).

Aligning the cloud

Aligning the cloud is imperative if the investment in this evolutionary technology is to be realized. Racing to implement technology without a proper understanding of the business demand it will service is a mistake. This is a cultural change across IT and business, so below are a few guiding principles to consider:

Do:

  • Be specific about workload-type requirements by establishing the means to measure them through the use of defined qualities (i.e. response time, availability and opportunity cost).
  • Challenge conventional wisdom about quality values; don't accept broad statements about 24/7 availability without some business qualification.

Don't

  • Don't allow application owners to pigeonhole their applications into specialized classes. Often, the owners of very high-performing applications consider all aspects of an application to be unique. Make sure the work is decomposed well enough by workload type to challenge those assumptions.

Conclusion

The cloud is the long-term strategic delivery model that will enable IT as a service while reducing virtual sprawl in the datacenter -- if firms design, build and operate it correctly.

To do this, firms must start with a top-down design, not bottom-up design, because top-down means starting with the business demand that is indicative of the business growth direction. IT as a service, delivered through a cloud, will be possible because the pressures on IT to deliver will force alignment with the business holistically.

From CIO: 8 Free Online Courses to Grow Your Tech Skills
Join the discussion
Be the first to comment on this article. Our Commenting Policies