By Tom Schmidt
In most modern data centers today, application management is a major undertaking. Large enterprises may have hundreds to thousands of servers, hosting thousands of applications. For a subset of applications needing specific uptimes, enterprises may deploy traditional high availability clustering. But these applications generally represent less than 10% of the total applications within a data center. While the rest may not have application clustering deployed, end users are still affected if outages occur.
For this large application set, most data centers rely on a combination of scripting and monitoring to keep applications available to users. Large companies generally deploy an enterprise monitoring framework to monitor the availability of specific network, server, and application components. This framework provides status and notification to operations staff.
In large enterprises, a front-end operations staff monitors the enterprise notification system. This staff is typically capable of receiving monitoring alerts and performing basic troubleshooting and limited corrective actions, which may be as simple as restarting the application or the server. For these actions to occur, however, the operator must be able to log in to the system, perform a basic analysis, and restart the application. This usually involves running scripts designed to stop and then restart the application. Although this sequence sounds simple, it involves significant effort.
Compounding the situation is the fact that many enterprises have adopted large-scale shared architectures running complex multi-tier applications across a broad, distributed collection of physical and virtual servers, accessing terabytes of shared storage. If enterprises are to keep up with the relentless growth in demand for data center services while keeping costs under control, they need an automated way to control applications, virtual machines, servers, and storage.
As Donna Scott, vice president and distinguished analyst of Gartner Inc., has observed:
"As data center management becomes more complex, IT managers struggle to find the right tools to automate business system administration. Enterprises need a consolidated approach that provides both visibility into their health and control for start, stop, and failover."
This article looks at one such approach.
The scale of the challenge
According to IDC, U.S. server installed base is expected to climb to 16 million by 2009, growing at a 14% compound annual growth rate. Worldwide, the installed server base is expected to top 39 million by 2009. Drivers for this growth include the wide deployment of low-cost Windows servers and Linux servers, along with the growth of blade servers for a range of workloads in the high-performance computing and commercial enterprise market spaces.
At the same time, IT budgets have remained flat and are just now growing again after significant declines earlier in the decade. According to Gartner, "IT spending as a percentage of revenue will actually decline in some organizations [and] flat IT staff budget plans suggest that organizations are still unwilling to significantly bolster the number of IT professionals in the ranks."
Gartner predicts that, by using an approach called Real Time Infrastructure (or RTI, a technology architecture that enables companies to fulfill IT business process, application, and infrastructure requirements from "resource pools" rather than dedicated resources), large organizations will reduce IT hardware costs 10% to 30% and labor costs 30% to 60% while improving quality of service.
Real data center automation
A data center automation solution should enable the management of large pools of enterprise applications based on policy and priority, using intelligence to optimize hardware resource usage, and helping maximize enterprise-wide application availability. It should also provide the benefits of reduced labor and hardware costs, increased IT responsiveness, and improved IT quality of service. There are four primary uses for this type of solution today:
- Stand-alone application management The solution is used to manage the start/stop/monitor of applications on single hosts. These will typically be applications where a specific high availability solution is not needed, but the organization would like a simple management process for starting potentially complex applications on a large number of nodes without the need to maintain start and stop scripts on each node.
- Multi-tier application management This refers to the ability to control a modern-day distributed application that typically consists of a back-end database, one or more application servers, and some number of Web servers. These components will typically run on separate servers and will also likely run on different operating systems.
- Server consolidation management In most enterprise environments, server utilization is a major concern. At the back end, many organizations see less than 20% utilization. Server consolidation management in this context refers to the ability to control "application stacking" on a single physical server and single instance of an operating system. At the most common level, this would require determining what applications are compatible to run together, then ensuring that the built-in resource manager of the operating system is properly configured to ensure that applications do not "run away" and take excess capacity.
- Reduce planned maintenance windows According to industry analysts, 80% of downtime is planned downtime, or time taken to perform scheduled maintenance. Although advances have been made in configuration and patch management, they still do not adequately address all the challenges of system maintenance, including easily migrating applications as well as being able to quickly roll back and restore operability. The growing scale and complexity of the modern data center require new tools for centralized control to help keep applications available.
Conclusion
As IT complexity continues to increase at a dizzying rate, a flexible data center infrastructure that allows enterprises to prioritize customer needs is necessary for delivering on application service level agreements. The centralized, policy-based management capabilities of a superior data center automation solution can help enterprises maximize the availability of critical applications and more effectively respond to the needs of their customers.
Tom Schmidt writes frequently about information security topics. He has more than 15 years' experience as a writer and editor in high-tech publishing.