Dr. Hossein Eslambolchi
April, 2012
In the mid-90s when the web was in its infancy, creating a web site was quite complex. It was not only necessary to design the content for the site (text, graphics, HTML…) but it was also necessary to build and operate the entire web server infrastructure: purchasing and installing servers, disk farms, routers, IP connections, web server software, monitoring system, raised floor computer rooms, backup power supplies, 24X7 operations staff needed to be hired, disaster recovery sites set up, and the latest security patches applied.
With the introduction of web hosting services, all the operational infrastructure and day-to-day operations became available as a service. Now, website creation simply requires a company (or an individual) to design the pages. The result was a dramatic reduction in the cost of creating and maintaining a web site as well as an equally dramatic expansion in the number of sites.
Enter cloud computing architecture. Comparable to how the introduction of hosting services dramatically changed the web, the cloud architecture will generate a similar shift in how companies network. Companies need to change the game for the customer by turning the network ‘inside out’, creating a user-centered, application-driven network.
Service providers have traditionally been viewed as networking companies providing voice, data, and IP. However, Tier 1 service providers are evolving their infrastructures, technology, and customer value in new directions. Applications will be able to execute in the network, providing unparalleled performance, scalability, response time, and application continuity, recoverability, and security.
Through this visionary network, customers will no longer need to purchase and operate a large and elaborate system infrastructure. They will instead pay only for the capacity needed to run their applications – even though this capacity may vary widely from moment to moment, day by day, or seasonally. Service providers are working to automate and develop intelligent operations systems which take on much of the responsibility for deploying and operating the application. This automation will increase or reduce the computing and network resources needed to keep the application running efficiently – while meeting customer-specified objectives for performance and response time. This secure, fault-tolerant, and global networking environment will provide true business continuity and ensure that both processor and network capacity is available to meet peak and required performance levels.
This approach should be part of any service provider’s overall utility-based computing strategy. It’s a network-centric approach that gives customers something other approaches don’t – true end-to-end optimization of application performance.
Architecture
Creating an environment that enables true end-to-end optimization of application performance requires a multifaceted approach. The grid computing architecture can be visualized as two separate but integrated environments: the network resources, including the network, systems and storage resources, and the software and service resources.
The combination of these resources will support multiple application types such as web services (which communicate with other applications), browser-based applications (where the user interface is via a web browser) and communications applications (where the application interacts with or controls a voice or multimedia calls). More importantly, these resources will enable business continuity, server performance optimization, and reductions in total cost of ownership.
Network Resources
Grid computing and networked utility computing strategy will need to support many different types of applications. Since one major focus of grid computing is web applications and web services, standard application servers, and related elements such as portal servers and database servers from industry software leaders will be deployed within the node. This deployment will provide a virtual environment that allows Sun Java 2 Enterprise Edition (J2EE) or Microsoft .NET applications to run on a variety of computing hardware. The standardized nature of these environments (J2EE and .NET) makes it possible to provide a common computing “fabric” that can be shared by different applications. Increasingly, new application development is migrating to these environments, as corporations drive to adopt more capable and standardized IP-related technologies.
Database servers will also be available to provide ‘local’ data management to support application operation. Doing so will optimize performance of the application by putting both the application and the database directly on the service providers’ backbones.
Grid computing operates within a shared infrastructure to minimize cost and achieve maximum flexibility, end-user responsiveness, and reliability. To ensure security and privacy, each application executes on one or more dedicated servers, which are dynamically assigned, as is a separate VLAN (Virtual Local Area Network). These private VLANs may also include secure VPNs to securely connect with databases and other resources operating in a corporate data center.
Grid computing incorporates the latest security technology, such as application-level firewalls, SSL accelerators and intrusion detection and prevention systems. In addition, operations personnel implement industry best practices, — the latest security patches are installed and all infrastructure is hardened against attack.
Software/Service Resources
With this in mind, the core of the grid computing architecture will be software and services resources — specifically, the dynamic resource allocation (DRA) layer, which is a software layer that automates the management of the computing, network, and application infrastructure.
Within the DRA resides a resource monitor that continually probes and monitors the computing, network, storage and application infrastructure, measuring performance, identifying overload conditions, detecting faults, and identifying the proximity of the demand to the various application service nodes. Changes in load and device status are passed to the DRA policy engine, where analysis determines if action is warranted and finds the optimal set of actions needed. The policy engine then manages the process of implementing these decisions. The actual detailed provisioning and control of individual resources is carried out by the resource manager, which communicates with each server, storage and network device using an adaptation layer, which allows new types of resources to be plugged in to the grid.
Grid computing will also provide a set of network-resident mediation services. Defined as common support functions, such as authentication, directory services, usage recording, device and user profiles, XML transformation, etc., mediation services simplify the creation of applications and web services that control real-time voice and multimedia communications. By utilizing these integrated services, a customer will be able to speed the development of new applications.
Simplicity and Flexibility
Grid computing is being developed with simplicity and flexibility in mind. The simplicity to monitor and manage all the network and application resources via a single, integrated portal and the flexibility to design, develop, and deploy applications when, where, and by whom you want. All by tapping into a secure, 24 x 7 managed environment that will allow enterprises to maximize under-utilized resources and take advantage of a robust computing and network infrastructure.
Network-Centric Approach
Unlike other approaches that focus primarily on data center optimization, grid computing takes a network-centric approach. This leverages service providers’ deep knowledge of network performance, as well as server and storage performance, to achieve optimal end-to-end application performance. The network is a major enabler of utility computing — it sets service providers apart in this area.
Grid computing will allow you to take a holistic management and monitoring approach to both the applications and the network, all over a single integrated system. It provides a simple, standard way to dynamically create and control flows of data between different systems. Applications and management systems alike will be able to create and/or control connectivity functions such as specifying end-points, bandwidth allocation, QoS, security settings, monitoring, etc.
When you combine grid computing’s support for Intranet, Extranet, and Public Internet applications with advanced network features, it’s clear that a network-centric approach is essential to achieving end-to-end optimization.
Business Continuity and Global Optimization
Historically, business continuity capabilities (fault tolerance and disaster recovery) have been pursued using a combination of fault-tolerant techniques. Spare processors are maintained at each data center and the entire data center is replicated at a geographically distant site. This approach can be extremely costly — extra, unproductive capacity is required and different fault tolerant technologies must be used and made to work together. It may be necessary to coordinate fault tolerant capabilities in CPU’s operating system, the data center’s management system and the systems and architecture that links data centers together.
Business continuity features are an integral part of grid computing capabilities. Because grid computing can dynamically add processors and move loads between sites, it can easily route around failing components. In the case of failure, new processors are rapidly and automatically allocated to ensure that customer applications have access to the resources they need. No expensive special purpose hardware or specialized disaster recovery techniques are required. In a fully developed system, more than 90 percent of installed capacity is available for production purposes.
And since the load balancing will be delivered across an IP global application service node infrastructure, global load balancing and disaster recovery can be achieved.
Grid computing will automatically deploy servers where they are needed to improve performance. A global corporation may be receiving requests from around the world. As the sun moves, the source of load will shift, from Europe, to the Americas, to the Far East. To meet this shift, grid computing will automatically deploy servers in the US while reducing capacity in Europe, and then later in the day, begin to deploy servers in Asia while reducing capacity in the US.
This visionary model may also place the burden of integrating and managing complex computing and network systems on service providers. It will be up to us to ensure that the infrastructure is secure, the necessary patches are applied, capacity is available, and the network is resilient – all things that service providers have been doing successfully for years.
Total Cost of Ownership:
The complete functionality of grid computing will have a significant impact on CXOs total cost of ownership.
Applications today are deployed and operated using application-specific, highly customized infrastructures, leading to a complex and difficult-to-change environment. The result is an inefficient utilization of computing resources with the average customer utilizing only 10-15 percent server capacity, which is far superior than today’s network. Furthermore, such applications are very expensive to deploy globally with mission-critical reliability. All this effort only serves to distract an enterprise from its core business and its ability to provide innovative differentiation to its customers.
Grid computing addresses these issues by providing a standard, shared, global, secure, flexible execution environment for applications. Thus, applications that run within grid computing achieve a much higher degree of resource utilization and, therefore, lower unit cost. The combined effect reduces the total cost of ownership.
In addition, the pay-for-use pricing model will allow customers to pay only for the capacity they need and actually use. Fixed capital expenditures and fixed monthly recurring operating expenses are now converted to variable costs as servers, storage and operational systems move from the customer’s books to service providers. Because infrastructure does not need to be purchased, the economic penalties associated with accelerating, delaying or even canceling the deployment of a new project are significantly reduced. The risks associated with deploying a new application that might experience unpredictable demand are also reduced.
Summary:
Welcome to the new world of optimized, intelligent and adaptive, globally distributed applications. The deployment of grid computing, along with cloud computing, will help companies simplify, integrate, and optimize system infrastructure. At the same time, the total cost of ownership to manage, monitor and deploy those systems drops. This isn’t too good to be true . . . service providers are changing the game for customers, enabling companies to utilize integrated, secure, end-to-end application networking solutions on each enterprise’s terms.
Customer Benefits
Grid computing:
- Dynamically deploys applications to maximize server utilization and performance, improving the customer experience and reducing capital investments.
- Anticipates peak usage with the intelligence to handle spikes in demand by automatically allocating anticipated capacity.
- Reduces total cost of ownership by leveraging operational support infrastructure (i.e. systems, people, etc.).
- provides hands-free, end-to-end flow through the process, enabling service providers to deliver services to customers in real-time, eventually achieving zero cycle time and zero defects
- Better aligns customer IT infrastructure operations and network costs with customer revenues, this reducing the variability of customer profitability.
- Results in fewer or no end-users turned away due to lack of capacity, since capacity scales automatically with demand.
- Improves response time and reduces latency for interactive applications by automatically invoking additional processors close to where demand is originating.
