Poor performance, such as low throughput, long response times, and so on, might be suddenly and consistently experienced by a certain set of users or applications. Also, traps might be generated by the Network SLAPM2 subagent. When this happens, the problem might be the way the QoS policy is defined for the corresponding set of users or applications.
For example, the IPv4 ToS/DS or IPv6 traffic class value might be set incorrectly to a lower QoS level than is intended, for example, medium or low priority instead of high priority. It is important to remember that given a fixed amount of network resources, changing some traffic demand from a lower to higher QoS level will mean that other traffic demands will be affected. Therefore, use care to ensure that in attempting to meet one set of QoS requirements, different or worse problems do not result.
Another cause for poor performance might be in the way the bandwidth allocation defined via the DiffServ token bucket parameters, or TCP maxrate or minrate, is not adequate to accommodate the traffic demand. Yet another possibility might be that either network or the server capacity is not adequate to handle the traffic demand. This is evident when a majority of users or applications do not have their QoS requirements met. When this happens, the network planning process must be revisited.
For more information, see Using the Network SLAPM2 MIB to monitor policies.