Cisco Network Mgmt Protocol FAQ: Service Level Management
Q1. Both the provider and the customer of a service should be concerned with service level monitoring. Why is this so?
Answer: The customer might not know for sure whether he is indeed getting the service level that was agreed to.
Q2. Assume that you are a service customer and are about to enter an SLA with a service provider to provide you with video phone service across your enterprise network. Can you think of some service level parameters that you might want to include in service level objectives of the SLA? (Pick three.)
Answer: Examples might include picture resolution, picture jerkiness, time to set up a video call when dialed, overall video call quality (a video call mean opinion score that captures also factors such as whether the voice and audio are synchronized).
Figure: Differences in Service Level Parameters, Depending on the Layer of the Service
Q3. In addition to a set of service level objectives, what else should an SLA spell out?
Answer: An SLA should also spell out how service level objectives are measured and what happens when service levels are violated. The latter includes the course of action that is to be taken when service level violations are detected, financial penalties that are incurred, and possible legal ramifications, for example, regarding early termination of service agreement.
Q4. Give an example of a voice service level parameter that cannot be derived from an underlying network parameter.
Answer: For example, the time to set up a call. This involves a sequence of signaling activities and processing that the underlying transport has no notion of. Another example is the mean opinion score, which judges the voice quality of a call that is impacted by aspects such as static, echo, and clipping.
Q5. Give an example of a service level parameter of a voice service that, when decomposed, contains a network service level parameter as a component.
Answer: Voice delay. Although it is influenced also by dejitter buffers, network delay is the main variable component of the voice delay that is experienced during a call.
Q6. Assume that your SLA specifies a maintenance interval of 1 hour per month. When measuring the availability of your network, you forget to take this maintenance interval into account and accidentally include 30 minutes that your service was not available during this interval. How much is your availability number skewed?
Answer: One month has 30 days with 24 hours, with 60 minutes, for a total of 43,200 minutes—excluding the maintenance interval of 43,140 minutes. Of that, 30 minutes constitutes roughly 0.07 percent by which the availability number is now skewed. Even if availability had otherwise been perfect, that would result in 99.93 percent availability—not bad for many scenarios, but two orders of magnitudes off from the coveted five nines carrierclass availability.
Q7. Name a technique that can be used to set up an early warning system for impending service level violations.
Answer: Threshold-crossing alerts, to indicate when service levels get dangerously close to missing service level objectives and when components of service level parameters are exceeding their allotted budget.
Q8. Assume for a moment that you are a service provider. A customer complains about the level of service he received. Your first impulse is to give little credibility to that complaint—you are quite certain that the level of service that you provided was well within the targets set by the service objectives. However, the customer produces as evidence a log file with service level measurements that were taken over the past month. Which three questions might you want to go over in your mind to assess the validity of the customer’s claims?
Answer: Are the measurements statistically relevant, or is it possible that they paint a picture that is skewed? How were the measurements taken—are they sure to accurately reflect the service level parameters? Is the data authentic, or could important samples have been inadvertently left out or manipulated?
Q9. What trade-off does a service provider need to assess when deciding whether and by how much to oversubscribe resources?
Answer: The trade-off is between cost and revenue. Cost is associated with resource utilization, which needs to be high because having resources idling is expensive. Revenue is associated with service levels that are achieved, which need to be high to charge top dollar for a service and keep customers satisfied. A service provider needs to find the spot at which maximum resource utilization is achieved without violating service level objectives.
Q10. What strategy might a service provider take when realizing that resource contention occurs, to prevent everybody’s service level from dropping?
Answer: Until the resource-contention situation passes, impose some form of admission control to prevent additional users from generating additional load on the network and underlying resources.