Unlimited system performance
In this work, the correctness is a hard requirement that must be added to 100% in all situations. CEP systems are generally real-time systems, which means that certain deadlines must be respected. So must be within a predetermined maximum time a situation have been recognized since the information in the best case is otherwise worthless. For critical systems, such as the monitoring of a chemical plant, the belated recognition of a leak or building up of dangerous gas mixture can have fatal consequences, so that regular queue methods a timely response to security not be queue management Kane. Night rarely the correctness requirement is in this particular case difficult to reconcile. Nevertheless, both can be met together, then this is usually done under the constraint of system performance. This work has also aimed to timeliness.
All operator of a CEP system wants to have both and possible underutilized operators as well an as little as possible loaded network. However, due to heavily loaded node in case of failure threatens quickly overload, because by eliminating a component the remaining units have to process more orders. It must therefore be quickly taken queue management care of replacement, but this can not always happen immediately, because as until a new node provisioned and then has to be applied to this one operator. If a replacement component permanently on standby, whilst it can quickly respond to outages, but this leads inevitably to reduced resource efficiency. The alternative, to be able to intercept incoming error using redundant event processing, reduces the performance of the system because both additional operators are needed and the network through multiple distribution is more stressed.
A paralleled and distributed CEP system can be represented by a directed graph G Operator operator. In turn consists of several operator instances j i, which make the actual correlation work. The on the various input streams incoming events are queue management serialized by the operators. Brought into a unique order. Start and generate serialized to the current selections, which are finally sent from the splinter as a complete unit to the operator instances for correlation and filtering over more. There is a correlation function can be performed f that maps the input current j in the output current. Individual input currents are referred to as an operator Iik.
Current queue management methods
Omega and run after merging on the splinter in queue management the total input current Ii. A subset of this current Ii is then forwarded to the operator instances and is there as I j i available. The output current O j i the operator instance j i via the sequence number generator operator. In its total output current Oi. Figure 3.1 shows the operator ω1 with all relevant streams. An event. A concrete data object of a specific data type. As in many cases, such a is also in this thesis exchanged against real-time and accuracy. In this chapter, I introduce the system model of paralleled and distributed complex event-processing system, which is also in the practical part for application.
In addition to the actual user data, it also contains various metadata. This includes a steady time stamp and a uniquely assigned by the producer sequence number that can not be changed as well. Events are the information or data with which a CEP system is operating. They encapsulate all the necessary information needed for processing in the system. Both innerhalb- between the operators queue management and the Instance- and outside the- from the source to the first operator and the last operator to consumers- events are sent and received for data exchange.
