A slow application is a broken application. In some ways a slow application is worse. It affects the productivity of corporate users, causes churn for service providers as customers leave in frustration, and costs e-commerce vendors business as potential customers leave for other sites with better response time.
White Paper
The Importance of
Response Time
Management
Eric Stinson,January 2006Introduction Monitoring
A slow application is a broken application. In some ways a Monitoring is the process of comparing the measurementslow application is worse. It affects the productivity of to a threshold and generating an alarm for thresholdcorporate users, causes churn for service providers as violations. This threshold is based on the question, "Whatcustomers leave in frustration, and costs e-commerce is bad response time?" For example, e-commercevendors business as potential customers leave for other customers may be willing to wait 60 seconds for a page tosites with better response time. load, but they would not likely wait five minutes. In thisarbitrary example, the threshold for web site responseThis paper will define Response Time Management and time, therefore, is somewhere between 60 seconds anddiscuss its importance to enterprise IT departments, five minutes. It obviously varies from user to user andservice providers, and e-commerce businesses. It will also from application to application, but there is a valueprovide an overview of response time testing methods, somewhere in that range above which we have a responseand introduce CA's Response Time Management solution time problem worthy of our attention.based on the SPECTRUMŽ platform. CA's solutionleverages existing response time technology to better "Good" Response Timemanage performance in the IT infrastructure. "Good" response time is subjective. Suppose the accountspayable application typically responds in five seconds. IfWhat is Response Time response time were to suddenly increase to 10 seconds,users would be unhappy. If it increased to 20 seconds,Management? they would stop using it. In contrast, for a time-criticalResponse Time Management is the process of measuring application such as tracking production on an assemblyand monitoring the time it takes for services to respond to line, response time must be sub-second. When ituser requests and aiding in the isolation of the root cause increases to as little as a second or two, products back upof poor response time. and production suffers, resulting in a loss of productivityfor the company.Services Root Cause IsolationA "service" is the unit of work a user has requested,whether directly or indirectly, via the infrastructure. This Root Cause Isolation is the process of examining what iscould be a complete transaction, such as submitting an known about the components involved in a unit of work toorder via a web site, downloading a file from an ftp server, narrow the focus of the investigation. Accuratelyor logging into the mail server. "Service" could also refer pinpointing the location of the problem increases theto underlying infrastructure services, such as assignment speed at which a problem is resolved and greatly reducesof an IP address from the DHCP server or a request for the cost of the repair.name services resolution from the DNS server. In many "Response Time" typically refers to the amount of timecases a service makes use of other services. For example, required to complete a unit of work called a "transaction."to access a web site the user enters a URL into their Another frequently-used performance term, "Latency",browser, which issues a request for name resolution (using refers to the delay introduced by subcomponents of athe DNS server) and then attempts to connect to the web transaction. For example, the time from when a serverserver before requesting the web page specified in the receives a request from the network until the response isURL. From all end users' perspectives, the service is the transmitted to the network is server latency. The timeweb page, but in actuality they have performed three required for data to travel from the server to the client isseparate transactions using three difference services network latency.(DNS, TCP, and HTTP). Several factors contribute to the latency in a userMeasurement transaction. Propagation delay is the time required for anMeasurement of the service is typically made from the electron to travel the length of a network segment.time the request is issued (user presses "ENTER", for Transmission delay is the time required to send a packetexample) until the entire response is received. In some across a network segment. Routers and other processingcases a single user transaction can be measured as an introduce more delay as pac... [download for more]
