When designing the strategic data center, one of the important design factors is the ability to add features into the data center without having to re-architect it. As the N1 Grid software continues to evolve the data center architecture plan must be flexible enough to enable the inclusion of these new products and technologies. The architecture plan is a road map, implemented in several phases.
During early N1 Grid enabled data center pilots, flexibility was achieved by phasing in technology and products as they matured. This can be done in accordance with other planned data center upgrades, perhaps in network or storage subsystems.
Phased implementation enables the IT staff to concentrate on systems that change rapidly or common platforms like web servers. This focus provides additional time to concentrate on more difficult automation tasks, such as the automatic installation of a database server or storage provisioning.
To demonstrate, TABLE 10-1 shows how a project is initiated and produces an overall road map for a phased release. Each cycle introduces additional functionality and expands the scope of the implementation.
Two environments are represented in TABLE 10-1: a pilot environment, which might be a quality assurance or test lab; and a production environment, which might be the production data center running the real services. In this example, four releases are split across production and the pilot environment. The dot releases are functional releases to the pilot lab, while the dot-zero releases are to the production data center. This enables adequate testing prior to production.
It is important to pick adequate example or target applications for pilot testing. In the example, the real-world applications for an e-commerce system were picked for the pilot. These are the same applications implemented in production. Because of the unique nature of application automation and provisioning, it is important to pay special attention to the target environment and to closely mimic production behavior. In the second set of releases, the environment is expanded to include Tier 3 applications for database servers, such as Oracle 9i.
In the first two releases (or iterations of the architecture), system builds are handled by basic automated operating system provisioning tools such as the Solaris OS JumpStart software. In the next two releases, system builds are provided by the N1 Grid PS.
In the first two releases, most of the focus is on automating application deployment across web servers and application servers (Tier 1 and Tier 2). Generally, the best targets for automation are commonly performed tasks, web servers, application servers (application containers or middleware), and the applications themselves. Applications change rapidly, as does web server content. Also, front-end (Tier 1) systems are usually deemed critical for security and other time-sensitive patches and software updates. After application deployments (and related changes) have been automated for these systems, the IT staff can then target other systems for true end-to-end provisioning. In the example in TABLE 10-1, the second set of releases targets deployment automation for database servers.
Observability systems, enterprise management systems, and system management integration are a complicated area. This chapter has concentrated on a hardware element manager, like core functionality within the Sun Management Center software, and an end-to-end service-level monitoring package, such as Micromuse's ISM (Internet Service Manager). The first iterations are focused only on Sun Management Center (Sun MC) software, while the second add end-to-end service monitoring, provided by ISM.
IT Management and the N1 Grid Software
The N1 Grid software changes how the data center is managed. It virtualizes many of the hardware systems and enables administrators to focus on higher-level problems and architecture, rather than time-consuming and repetitive duties.
Operators to System Administrators
Typically, operators focus on monitoring systems, performing backups, print jobs, and lower-complexity tasks, such as adding a user to an existing system. In some environments, basic operating system loads might be performed by operations staff if they have been automated, such as by the Solaris OS JumpStart software. Operators are responsible for acting as the interface between users and administrators, for basic troubleshooting, and for level one support. With the N1 Grid software, many of these monitoring and response tasks are automated. Using N1 Grid PS, operating system loads can be automated. Using the N1 Grid SPS, application provisioning tasks can be automated.
Operations will be responsible for creating an application instance or server load, including the specific service components necessary to deliver a service. Operations might be responsible for descriptively providing instance details, such as the total number of users that this node is intended to support. With this information, the initial N1 Grid system will build the correct resource. Typically, today's operations staff do not have this type of information, so processes and procedures will have to be designed to enable operations to perform this type of task, or it could be provided by a system administrator or architect. As the N1 Grid vision is achieved, operations staff starts taking on many more system administrator roles, such as creating the proper instance or deployment configuration of a service.
System Administrators to Architects
Today's system administrators are focused generally on server tasks, although they might be responsible for networks, storage, and backup. They have a wide variety of roles, however. They are usually very detail focused (for example, configuration of network interfaces and failover on a server).
Many of the common system administrator tasks will be automated with the N1 Grid software. System administrators will become more like architects. They will focus more on how the entire system operates the various system components necessary to provide the service and on higher-level issues such as interoperability. Network provisioning and many server configuration details will be automated, based on templates, so that systems are indeed wired once and provisioned many times.
The N1 Grid system administrator will use systemic patterns to help craft solutions for the data center. Patterns will be provided in such as way that constraints are detailed, and as long as the solution is deployed within those constraints, the system staff can make modifications as necessary. These patterns are then submitted to the operations staff to create and deploy the instances.
Examples of patterns include the J2EE design patterns that are used to help provide consistency and efficiency for common programming functions. Another example is Sun's Reference Architecture program that helps architects and administrators by providing baseline architecture, implementation, and performance information based on the design. Soon, these types of patterns will be available to the N1 Grid system, customized and deployed automatically. Of course, this type of automation requires policies to be well defined.
Architects to Pattern Developers
Architects who typically were concerned with interfaces (that is, how systems work together) often start from scratch to develop an architecture. Now they can use, and even create, systemic and software patterns to enable these architectures to be automatically deployed.
They will still be concerned with the systemic qualities (for instance, scalability and reliability), but these are channeled through the creation and modification of architectural patterns, not individual server instances. These patterns are contained within the N1 Grid software repository and can be used across multiple sites, even multiple customers.
Initially, pattern development will be based on reference architectures and solutions, such as those provided by Sun. At some point, vendors will be able to describe their system and software capabilities in such a way that the N1 Grid software can be responsible for the assembly of the entire architecture, microarchitecture, or runtime requirements.
Strategic Flexibility and the N1 Grid Software
A hierarchy of efficiency and value is achieved through the build, deploy, flex, and optimize cycles. Each component of the cycles provides particular additive value. Unlike the previous two individual solutions, data center optimization is highly dependent on the additive value of both infrastructure and application optimization.
The first two cycles are mostly focused on increasing utilization, decreasing server and other resource costs, and decreasing time to market for new services and changes to existing services. The capabilities of the N1 Grid software provide the following benefits for IT:
Comparing Operational Efficiency with Strategic Flexibility
The following product summaries illustrate how the N1 Grid software can be used to achieve both operational efficiencies (such as cost savings by improving utilization) and strategic business benefits (such as improving the time to offer new services by using flexible, reprovisionable servers).
N1 Grid Data Services Platform
The N1 Grid DSP provides basic value through the following efficiencies:
The N1 Grid DSP increases strategic flexibility by:
N1 Grid Provisioning Server
The N1 Grid PS provides basic value through the following efficiencies:
The N1 Grid PS software increases strategic flexibility by:
N1 Grid Containers
N1 Grid Containers provide basic value through the following efficiencies:
N1 Grid Containers increase strategic flexibility by:
N1 Grid Service Provisioning System
The N1 Grid SPS provides basic value through the following efficiencies:
The N1 Grid SPS increases strategic flexibility by:
The N1 Grid software provides a powerful shift in thinking for the data center. Implementing the N1 Grid technologies enables improvement, cost savings, and strategic capabilities. Application life cycle mobility emerges, automating flows from development through test and into production. Timeshare test environments become a possibility. A combined, optimized, and flexible use of shared infrastructure versus dedicated infrastructure becomes possible. Proactive capacity planning can be achieved through appropriate use of the observability infrastructure. The combination of all of these capabilities promises to deliver greater availability, shorter time to market, and better utilization, while delivering significant cost savings.