Monitoring Cube Performance and Usage - Expert Cube Development with SSAS Multidimensional Models - page 342

XMLA
We can also initiate a trace by executing an XMLA command. The columns and events
used in the trace are defined in the XMLA command itself. An example of an XMLA
script is available at http://tinyurl.com/XmlaTrace .
Flight Recorder
Flight Recorder is a feature of Analysis Services that maintains a log of all events
that happened in the recent past, which might be useful when investigating crashes
or performance problems; it works by running a trace. By default, it doesn't capture
all events and only keeps data for a limited time (so as not to fill the disk with trace
data), but we can customize it by changing both the length of time it keeps data and
the events that it records. We have to remember though that Flight Recorder can
affect performance: the more events it records, the more I/O operations are required
to update the trace files it generates.
Flight Recorder trace files can be opened with SQL Server Profiler and are stored
in the OLAP\Log folder (usually found at C:\Program Files\Microsoft SQL
Server\MSSQL.2\OLAP\Log ). To customize the trace definition it uses, we can
use a SQL Profiler template in the same way we did with ASTrace.
Flight Recorder and Trace Architecture
A more complete description of the Trace architecture of Analysis
Services and the behavior of Flight Recorder is available in Chapter 38
of Microsoft SQL Server 2008 Analysis Services Unleashed , Gorbach , Berger ,
and Melomed , Sams Publishing .
Monitoring processing with Performance
Monitor counters
In the first part of this chapter, we saw how to use Performance Monitor to monitor
the operating system. Analysis Services has its own set of Performance Monitor
counters too. The most important ones for analyzing Analysis Services processing
operations can be found under the MSOLAP Processing category. Total Rows
Converted , Total Rows Read , and Total Rows Written are respectively the count
of the number of rows converted, the count of the number of rows read from
all relational data sources, and the count of the number of rows written during
processing. These values allow us to monitor the amount of data processed, and can
be useful for evaluating if the amount of time taken for processing is related to the
amount of data processed.