For a long time, performance management has been a good idea, but now it's a must-have. Performance management defines, quantifies and reports on the value IT provides to a business. A CIO needs to understand the organization's IT value, which exists within the company's computing fabric. It takes more than IT to define this value; business owners and customers play a role in creating it, too.
In the past, management dollars have been at the bottom of the budget. Application and infrastructure led the priority list, and despite lip service for proactive management, it never became a necessity. The complexities of network, systems and application bottlenecks were not worth the effort to unravel. It was cheaper to buy more computing or bandwidth than to figure out what the problem was.
|
What Do Readers Think?
Check out our e-poll results on performance management.
|
Performance-management services are helping change this attitude. They have the presence and expertise that are costly to duplicate. In every case, these services have placed monitoring and load-generation engines around the world, collocated within ISP data centers and even on users' desktops. They have also made it their core business to be expert at creating and analyzing performance.
This core business focus requires that these services accurately convert the soft dollars of management into hard profits. This is serious leverage on the side of the service consumer; management products rarely come close because of implementation and maintenance expenses.
Service metrics are used to evaluate the performance of the service provider and carry clear monetary penalties. Performance management is critical as a monitor of the internal operational and capacity issues, but it is also required to give Web owners a view into their sites' performance, for promotional and development planning.
Additionally, just as Web users can abandon a slow-moving site, the Web has made it much easier for site owners to change hosting (an act known as deconversion). For this reason, the threat of deconversion is less of a reason to put up with poor Web-hosting performance, whether it's an internal IT service or not. Finally, a performance-management service provider's viability rests on correctly determining the costs of providing useful performance information to remain competitive.
The drawback to using a service is the loss of control over what data is collected when. This issue becomes less crucial if the only data that matters is availability and throughput, rather than the specifics of I/O, CPU, memory and bandwidth allocation. But those specifics help the provider deliver the higher-level service; as such, the need for granular performance data increases.
Data ownership was also a problem. Initially, enterprise performance-management products were adapted to run in multiuser service-provider environments. However, specific service-provider versions are now being deployed. If you ask for more control over your data, you're likely to get it.
What's IT All About?
Performance management helps IT departments set the agenda for all business units in terms of what value should be expected from the network, systems and applications. It's not just saying, "Look how well we're doing"; it is defining the metrics in a way that business owners can understand, buy into and find value in knowing.
A performance-management package should do more than report a cookie-cutter percentage or index. The goal is to determine the correlation of networking capacities with specific business goals. Over time, this relationship will create shared values for business owners and IT staff.
To deliver on these values, the underlying infrastructure's capacities must be accurately represented. Native network, systems and applications usage must be dead-on at all levels of summarization. Achieving this goal demands a common data description. At the lowest network and systems levels, these descriptions are well-understood, but unfortunately, applications and services do not yet derive or share a description with networks or systems performance.
Within the SNMP management name space, for example, network, systems and applications share a common data structure, the MIB. However, the MIB is unlikely to be implemented in systems and even less likely to be used in applications, leaving it focused solely on managing the network infrastructure. The Desktop Management Task Force (DMTF) has championed CIM (Common Information Model) as the way to describe systems, applications and network-management information in a common way. The implication is that management applications can interpret the effects of slow bandwidth and high Web-server CPU utilization as correlated performance events that will affect the user's experience and abandonment thresholds.
Once there is a common description, it must be instantiated in applications, system and network design. In other words, uniform management data that easily correlates across network, systems and applications isn't available and won't be for years. Nevertheless, demand for performance-management solutions is strong, and there has been strong product and service development here.
RSS
