DaveWentzel.com            All Things Data

My RML Utilities Notes


I haven't had a chance to play with this heavily, but I like what I see so far. 
PSS uses these tools.  RML stands for Replay Markup Language.  Will work with 2000/2005/2008. 
Helps with replay, comparing two workloads, lots of load-based reporting. 
ReadTrace takes .trc files (or even .zip or .cab that contain .trc files) and can output .RML, .TRC, or even the PerfAnalysis database.  It doesn't use fn_gettracedata or SMO. 
-a outputs the PerfAnalysis db, my favorite aspect. 
-o output directory
-I(-i) full path and file name of first trace file (or zip path)
-f disables .RML output if you don't need to replay. 
-M can separate spids into their own .TRC files
-r## only process a certain number of rollover files (vs all)
-d name of the PerfAnalysis database (PerfAnalysis is the default)
-T148 enables aggressive normalization.  Will identify RPC patterns.  Language events vs RPC events will be structured slightly differently but this flag enables them to normalize to the same call.  RPC calls use bound parameters and hence will likely perform better. 
More on Normalization
This is very similar to what my existing scripts do.  WHERE col = 'ABCD' becomes WHERE col = {str}.  It also handles sp_cursoropen calls in a similar manner as my scripts.  It will not do predicate reordering, but if the plans are the same you can use the plan tables to do this for you. 
PerfAnalysis Database
is required for Reporter functionality.  The ReadTrace schema is created as well as other necessary objects.  Different captures cannot be combined in a single db.  To compare data between runs use cross db queries. 
The samples.cab file contains .sql files that set up sql trace to capture the necessary information for reporting, replay, 'lightweight'. 
It uses rdlc files but it doesn't appear as though you can customize these. 
Performance Overview:  high level overview of the data capture, similar to the Performance Dashboard custom reports. 
Resource Consumption by:  what app/login takes the most resources?
Unique Batches
Unique Statements
Top Unique Batches

Add new comment