Microsoft SQL Server 2008 R2 Specifications download pdf (Page 167)

StreamInsight Architecture CHAPTER 8 147

Data Structures

The high-throughput data that StreamInsight requires is known as a stream. More speci-

cally, a stream is a collection of data that changes over time. For example, a Web log contains

data about each server hit, including the date, time, page request, and Internet protocol (IP)

address of the visitor. If a visitor clicks on several pages in the Web site, the Web log contains

multiple lines, or hits, for the same visitor, and each line records a different time. The informa-

tion in the Web log shows how each user’s activity in a Web site changes over time, which

is why this type of information is considered a stream. You can query this stream to nd the

average number of hits or the top ve referring sites over time.

StreamInsight splits a stream into individual units called events. An event contains a header

and a payload. The event header includes the event kind and one or more timestamps for the

event. The event kind is an indicator of a new event or the completeness of events already in

the stream. The payload contains the event’s data as a .NET data structure.

There are three types of event models that StreamInsight uses. The interval event model

represents events with a xed duration, such as a stock bid price that is valid only for a certain

period of time. The edge event model is another type of duration model, but it represents an

event with a duration that is unknown at the time the event starts, such as a Web user session.

The point model represents events that occur at a specic point in time, such as a Web user’s

click entry in a Web log.

The CEP Server

The CEP server is a run-time engine and a set of adapter instances that receive and send

events, as shown in Figure 8-1. You develop these adapters in a .NET language and register

the assemblies on the CEP server, which then instantiates the adapters at run time. Input

adapters receive data as a continuous stream from event stores, such as sensors on a factory

oor, Web servers, data feeds, or databases. The data passes from the input adapter to the

CEP engine, which processes and transforms the data by using standing queries, which are

query instances that the CEP engine manages. The engine then forwards the query results

to output adapters, which connect to event consumers, such as pagers, monitoring devices,

dashboards, and databases. The output adapters can also include logic to trigger a response

based on the query results.

1 2 ... 162 163 164 165 166 167 168 169 170 171 172 ... 235 236

No comments

Microsoft SQL Server 2008 R2 Specifications Page 167