The Internet has grown rapidly in last decade. Its growth has both posed opportunities and threats to many business activities. It has become a global channel of communication to include all businesses. The internet is expected to become a medium of communication that would provide a differentiated level of service on the same internet infrastructure. Thus, this differentiated level of service would include data and voice communications that should also be able to provide real and reliable network services as well. In order to provide many services to the same infrastructure network, IETF has proposed many QoS technologies like DiffServ, IntServ and MPLS.
Presently, there is huge gap between network lower-level specifications and the application level service level require-ments. Application level QoS which are specificed as SLAs (Service Level Agreements) are specified and usually dealt with only at the application level. Network lower level details which are specified in terms of jitter, delay, throughput are dealt separately with at the network layer and it is usually, difficult for application developers to specify their network level QoS requierments at design or run-time. In this work, we shall restrict our attention to enterprise networks and enterprise applications. Internet is too big a domain to start tackling, even though the protocols may be the same.
SLA and QoS parameter at the application layer reflects the quality of service to the customer. It should be easy for customers to understand the degree of assuring the service. They can be different according to the types of service. On the other hand, network QoS parameters means the basic metric of performance measurement of network level QoS at the network management layer.
Objective in the whole process is the mapping of NPMs to QoS parameters (SLAs) and then giving the view of the net-work in terms of SLAs to applications, pricing, conformance etc and not in terms of NPMs to customer, network provider and application provider.
Here, we collect NPMs from the network through SLA agent and then map them to QoS.
Say, some application wants a delay of less than 10 sec. We should be able to collect the Network state (which is nothing, but NPMs which are: traffic congestion, throughput, availability and latency).
As an example for the IP enterprise network, the SLA of some service provider with the customer can be as follows: The availability over 99.99 % should be assured. The average RTT (round trip time) should be less than 50 msec. The delivery ratio should be more than 98.0%.
The QoS network parameters would be in terms of availability, delay, jitter, latency and loss. Since, the network performance metrics are gathered through network monitoring methods, we need to convert these network statistics to QoS network parameters at the network management layer which needs to be mapped to the SLA at the application layer above.
The network monitoring methods use to measure network performance and QoS parameters can be : active monitoring, passive monitoring and through SNMP agents.
The active monitoring method obtains the current status of the network by setting up the test machine from the point which one wishes to measure, and then sending extra traffic from one machine to another during a specific time. Passive monitoring can monitor the network status without additional traffic. But limited NPMs can be measured comparing to the active monitoring method. By using SNMP agents, we can measure the status of the network device. For example, RMON monitors traffic information with SNMP agent. Therefore, Loss, Delay and Connectivity are the NPMs (Network performance metrics) that can be mainly measured by active monitoring. Utilization and Throughput are the NPMs which can be measured by passive monitoring. Functionality and Throughput can be measured using SNMP agent.
Although we may obtain various NPMs using these network monitoring methods,it is difficult to apply these values to SLA parameters directly. The contract between service providers and customers is performed using SLA Parameters and the network quality is measured using NPMs. To measure the SLA Parameters, first the NPMs for each SLA parameter should be decided. A QoS parametercan be mapped to one NPM or many. This mapping depends on the type of services and can be very complicated. And the quality information of service should be presented in customer friendly form, SLAParameters, not NPMs. Therefore it is necessary to translate the measured NPMs to SLA parameter.In this report, SLA is used interchangeably with QoS parameters at the application layer. And the QoS parameters at the network level is used interchangeably same as NPMs (network performance metrics).