Grid computing is a very hot topic these days. Many major IT vendors are promoting and announcing "grid," "on-demand," "adaptive infrastructure" or some closely related initiative. It's likely the buzz will only increase as these firms reorient themselves to this emerging market.
Though it may seem to be yet another "next big thing," grid computing is in fact bringing real benefits to commercial enterprises. That's why enterprises and the software vendors that serve the analytics/business intelligence (BI) sectors are now partnering with the technology specialists in this space - or pushing initiatives of their own. It's particularly relevant in today's hyper-competitive yet cost constrained times when companies truly do need to do more with less.
Much of the talk around grid computing is often focused on hardware. The traditional view connotes aggregating numerous underutilized resources across an enterprise into a virtual supercomputer. However, while optimizing utilization is certainly beneficial, it's only part of the story. When grid computing technologies with an application focus are used, companies reap substantial benefits from both a business and IT perspective.
Overcoming application bottlenecks - which directly impact business capability and operational efficiency - should be the goal of any IT investment, be it software, infrastructure or hardware. With application-centric grid computing, enterprises can drastically reduce the time to results of software applications and enhance overall performance, as well as resiliency and reliability, in a cost-effective manner. Previously dedicated and/or underutilized resources enterprise-wide can be shared across business units, providing computing power as needed, when needed for business-critical analytics, BI and data warehousing applications. Not only does this approach leverage existing IT investments and new commodity computing models, it has been proven to generate substantial top-line growth and bottom-line savings.
What's Driving the Need?
An increasing array of competitive pressures, regulatory and compliance requirements, and more responsive decision support needs place ever-greater demands on IT organizations. Increasing demand for application performance and reliability continues to outstrip organizations' supplies of available computing resources, particularly in a time of constrained budgets, reduced staff and cost cutting. Many resources are dedicated to particular applications or groups, trapped in these suboptimized "silos." IT organizations must continue to meet stringent service levels while deploying and managing resilient and scalable applications across the enterprise.
There are a number of specific factors driving the need for more computing power. These increasing demands are fueled by more sophisticated algorithms and the need to analyze huge amounts of data. Many applications are compute- and/or data-intensive and require large numbers of simulations, while volume-intensive applications with large sets of service requests have rapid turnaround requirements. Intraday and real-time data analysis is often simply unavailable given the time constraints and processing power required. Business intelligence functions cannot be completed in a timely manner. Data analytics can take hours to days to process, making actionable "what-if" queries nearly impossible. With faster and faster changing market conditions, the need to react quickly is paramount. In addition, the risk of decision support delays or breakdowns due to dependencies on single points of failure (hardware, software or network failure) is high.
To remain competitive, it has become imperative to have easy access to enough computing capacity to power key applications to address business-critical issues such as:
- What types of marketing offers should be promoted to which customer segments?
- How is customer lifetime value modeled?
- How do you manage inventory to balance holding costs versus stockout costs?
- How do you leverage all the inherent customer data in your organization?
- How can you respond to and preempt competitive moves?
Constraints and Limits to Past Approaches
Until now, many compute-, volume- or data-intensive tasks have been tied to large, expensive, proprietary hardware. These "stovepiped" applications and their underlying models and analytics grew in sophistication and complexity as time pressures increased and more compute horsepower was needed. The only answer was to add larger, more expensive, proprietary boxes. With this type of resource-oriented approach, administration and maintenance of such dedicated environments was exceedingly difficult and not particularly cost-effective. Applications did not always scale when faced with demands for increased computing power.
Corporate IT organizations also needed to plan for the peaks in workload demand to prevent a computing shortfall when a larger than typical amount of computing power was needed. Different lines of businesses typically acted like spoiled children in that they wouldn't share - obviously a costly option, especially considering that hardware utilization typically ranges from anywhere from 10 to 30 percent in most companies.
However, existing approaches such as cluster management and batch scheduling have never delivered enterprise-ready solutions. It has been impossible to provide the necessary power due to cost, scalability, management and maintenance challenges. The current environment creates a new mandate: applications must be cost-effective, scalable and operate in a heterogeneous environment, and lines of business must share resources across the enterprise. Importantly, processing the volumes of data in near real-time modes has necessitated access to power as needed, when needed, automatically. To meet their business requirements, companies need to optimize the use of their existing, underutilized resources and deploy resilient, scalable architectures to support their applications.
Figure 1: On-Demand Computing Environment
Why Application-Centric Grid Technology?
With application-centric grid computing, rather than the traditional resource-centric approach that only manages compute resources, traditional obstacles are overcome and significant gains are realized. Such solutions decouple stovepiped applications from their dedicated resources, enable cross-unit resource sharing, focus on real-time business applications and are based on open standards. They are changing the economics of corporate IT and even how business groups can run their operations.
Application-oriented grid computing solutions result in faster time to results, increased accuracy and more sophisticated analysis and modeling. Application uptime is guaranteed. And not only are all facets of application performance improved, but application cost of ownership and overall IT costs are reduced through lower maintenance, optimized resource utilization and use of superior price/performance technologies.
New grid solutions are also leveraging important technology trends. For example, moving from proprietary architectures to commodity processors provides a way to meet the never-ending demand for more power - overcoming the "More Law" of resources: the more users have, the more they need and want. Commodity processors such as those from Intel can offer an almost limitless, cost-effective supply of computing power - if they can be harnessed effectively.
Trends toward open source software such as Linux also hold promise if enterprise-level functionality such as fault-tolerance and scalability can be integrated as well. Additionally, new techniques such as "reusable" components for business applications, as emerging from Web services, also show promise, again, if they can be made "enterprise-ready."
However, simply deploying a large compute farm of commodity processors, implementing a Linux environment and/or moving to Web services are approaches that do not address the problem of getting computing power to users when needed, where needed, as needed. Delivering such power on-demand, efficiently and manageably is crucial.
With grid technologies, firms can perform real-time analyses they couldn't previously, run more sophisticated simulations driving improved product development and react quicker to market opportunities through more responsive decision support.
Grid-Enabling Your Applications
The key to deploying such infrastructure and leveraging new cost-effective technologies is an operating environment that easily "grid-enables" a wide range of applications and services. Application-centric grid computing products unite all these technologies and their benefits, and offer new paradigms for corporate technology. By providing an application with substantial levels of application scalability, performance and reliability are realized, expanding capabilities to drive revenue opportunities. Resources are easily shared across the organization; thus capacity is available whenever needed - leveraging underutilized hardware and optimizing hardware investments.
By focusing on the application - the workload (i.e., the business logic and process execution) - itself and creating a "virtual" pool of all of an enterprise's computers (from mainframes to servers to desktops), grid technology can easily extend applications so their components can be distributed and processed across this pool, invisibly to the user. The user simply obtains results faster with guaranteed uptime and as much power as needed, whenever needed without having to "stock up" (i.e., spend) for a worst-case scenario.
The latest grid computing advances have enabled Web services and service-based applications to be scaled over the same computing infrastructure along with traditional applications, providing a common application operating environment across organizations, greatly simplifying IT maintenance and speeding new application development and deployment.
Current products provide a distributed computing environment that makes it easy to write, deploy, manage and administer application components on a grid. Applications are typically grid-enabled within days to a few weeks. In addition, more third-party applications are incorporating grid technology to overcome performance and scalability constraints.
How is this done? Compute- and data-intensive applications typically have large workloads that can be divided into discrete, constituent components and then processed in parallel, and are obvious candidates to move to a grid environment. Service-oriented application approaches are also well suited for grid environments. By supporting multiple programming paradigms, these solutions capture basic parallel programming and service-oriented computing models that exist across many application domains, enabling rapid grid deployment.
Grid computing is applicable cross-industry over many sectors. Figure 2 depicts the wide range of applications across verticals that may be "grid-enabled."
Figure 2: Applications That May Be Grid-Enabled
Driving Business Value
This next generation of computing infrastructure truly is enabling companies to change business practices by performing activities that were previously impossible, redirecting resources to more value-added tasks, entering new markets, offering new services (both internally and externally) and directly impacting revenue and profit contribution. Grid computing is aggregating computing power and converting it into business value.
Moving to an application operating environment also enables new revenue-producing applications to be deployed much quicker, as the IT "plumbing" infrastructure to scale applications - so it is powered appropriately - is no longer an issue.
Reduced Operating Costs
Another key benefit sometimes overlooked when considering a grid operating environment is the significant cost savings that are realized. In fact, grid computing provides a powerful cost justification that is likely to have major impact on near-term IT projects - especially given today's limited budgets and focus on cost control. Existing grid deployments have demonstrated that dollars spent today on grid initiatives result in multiple times such investments in cost savings.
Optimized Capital Investment
Reduced hardware capital investments are another significant and attractive cost saving yielded by grid computing. As applications increase in breadth, volume and complexity, they increasingly require more compute horsepower. This necessitates that more and more funds be spent on hardware. However, as specified previously, with resource utilization across most enterprises estimated at anywhere from 10 to 30 percent, funding the purchase of new hardware is ineffective.
More importantly, as the overall machine volume increases, support costs can skyrocket; operating costs can range from five to eight times the initial hardware expenditures, dominating application TCO (total cost of ownership). Grid computing solutions have a significant impact in this area.
Deploying New Technologies
IT costs are also being reduced through the types of resources being deployed. Traditionally, many compute-, volume- and data-intensive applications used large, expensive SMPs (symmetric multiprocessors) to provide the requisite power. Grid computing is enabling a move to commodity computing (away from the "big box" world to more of a Wintel/Lintel environment) with obvious lower initial expenditures and follow-on support costs. By harnessing all the resources of these machines, a grid infrastructure can create a "compute pool" that can power applications in an infinitely scalable manner, and much more cost-effectively than adding expensive, proprietary ones as was done previously. Grids are even off-loading processing from expensive mainframes to cheaper desktops while maintaining the mainframe's expensive cycles for the workload with which it's best suited.
Companies with grid deployments have increased utilization of their hardware assets from 20 to more than 95 percent, reduced hardware costs by 30 percent and slashed transaction costs by 90 percent.
Grid computing is enabling more wide-scale deployment of new technologies with powerful price/performance benefits. Whether it's migrating from Unix to Linux servers, SMPs to blade servers or client/server strategies to service-oriented architectures, grid computing can enhance the inherent advantages in each of these initiatives, ultimately helping IT organizations cut infrastructure costs while providing higher levels of service. By reducing people and infrastructure costs, and avoiding unnecessary growth costs, grids are a proven approach to slashing IT expenditures while driving superior business performance.
These technologies are also driving structural and organizational changes, such as centralized "ownership" of IT resources. This enables business units to simply pay for the computing power they need (via internal, or even external, utility computing services) - knowing that whatever computing power is needed will be available when it's needed, where it's needed - without user intervention. With this, organizations can support new business given this newfound computing power.
Transformation of Enterprise IT
With an on-demand computing environment, organizations capture existing power from underutilized heterogeneous resources and optimize added hardware investments to power a wide range of analytics, BI and data warehousing applications and services in real time. This can lead to measurable cost reductions as well as revenue increases by enabling new businesses and operations that previously could not be performed simply because "we don't have enough computing power to do it as quickly and as accurately as we need to."
Grid solutions are creating quantifiable bottom-line impact across organizations today. Firms not aggressively pursuing this new approach of computing run the real risk of putting themselves at a competitive disadvantage. Shortly, deploying intra-enterprise grids will not be an option for corporations; such technology will become a standard part of corporations' IT infrastructure due to the compelling benefits and overriding cost of falling behind.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access