The
following table view screen shot will be used
when describing the table visualizations in
JXInsight. Multiple screen grabs where made
and then layered on top of each other to show
the tool tips displayed over different parts
of the table screen space. The table view is
taken from the transactions perspective within
the profile mode of the performance management
console and displays statistical information
for transaction paths occurring in the BEA WebLogic
Medical Records J2EE™ sample application.
For database bound applications this is one
of the most important views to start from.
Transaction
Paths
A transaction path is an ordered set
of tuples each one uniquely identified
by the SQL, Java call stack, Trace stack
and position in set. The JXInsight server
will aggregate execution resource counters
- clock time, cpu, thread monitor waiting
and blocking, GC time, object allocations
and clock time adjusted. Count, maximum,
minimum, average, and total statistics
are recording for each counter.
|
Lets
break down each graphical element in the table
view. Following this breakdown
you should have a better appreciation of the
amount of information that is presented in such
a small screen real estate.
Call Stack Prefix
The
'Call Stack Prefix' column provides a graphical
display of the common call stack prefix between
the first and last visit executions in the transaction
path.
The icons to the left of the text, the method
name and line number of the transaction entry
point, represent a partial sequence of the Java
call stack when the entry point occurred. The
icons represent
classified Java classes with methods on the Java call stack.
Classes with consecutive method invocations
are compressed to a single graphical icon entry.
The sequence when processed from left to right
represents the incoming caller list thus it
is the reverse of the typical stack display
where the top element is displayed first.
This is an extremely useful visualization of
the call stack because one can quickly see that
that first path, an XA transaction, was started
after the following call sequence: message driven
bean -> ejb object -> session bean ->
entity home -> entity bean -> xa resource.
The time to scan and understand is very fast.
The graphic also helps in distinguishing the
type of incoming requests to the J2EE container.
A moment processing of the second row and one
understands this is an incoming web request
that was processed by a filter, then a servlet,
which was then forwarded to a Struts specific
servlet and action, eventually arriving at an
ejb object.
Access
The next column, 'Access', states the type of
transaction based on the DML type of SQL statements
executed, Read-Only = {
SELECT},
Read-Write = {
SELECT,
INSERT |
UPDATE |
DELETE},
and Write-Only = {
INSERT |
UPDATE |
DELETE}.
The
graphic to the left displays the types of operations
executed within the transaction.
Table
Access
The next column, 'Table Access',
is extremely powerful in helping to quickly
identify transactions paths (history patterns)
that have similar table access but slight differences
in the Java call. Each
table identified via parsing of all SQL statements
in the model is represented as a column of 3
pixels. Each column has 4 rows for the type
of access (
INSERT,
SELECT,
UPDATE and
DELETE).
A tool tip displays the database table name
associated with each graphic column. In the
following screen shot
UPDATE and
SELECT statements where executed against the record_seq
table.
Trace Stack
The
next column, 'Trace Stack, displays
the transactions entry point trace stack. The
top most (in terms of the stack) is displayed
in the cell. A tool tip is available when moving
over the cell displaying the complete entry
point trace stack.
The first graphical icon indicates whether the
trace is explicit and has a trace point (arrow
drawn in box) or that the trace identifier has
been added for delineation for trace stacks
trees as well as aggregation
of statistics.
The second graphical icon indicates the type
of trace identifier.
Depth
The depth column has a common cell rendering
technique used in all console table columns
with numeric values - the slight shading of
cells with values above the average for the
column. The shading is subtle because in a table
with a large number of columns any other rendering
effects would loudly distract the user. Please
note that maximum values in a column are prefixed
with '*' .
Count
All table column headers have 2 rows. The top
row for the label. The bottom for a statistic
value of the values in the column. This can
be a total, max, min, or average. When a column
is selected in a table, as is the case for 'Count'
column in the main screen shot above, then a
color scheme is applied. This same color scheme
is used across different visualizations in the
console. The color selected for rendering of
a cell's background is based on the ratio of
the cell's value to the maximum column value.
Clock
Avg
The 'Clock Avg' column has an embedded graphic
showing a miniature bar chart for the distribution
of all recorded values. A tool tip is displayed
with a larger scale of the bar chart when a
mouse is moved over.
Bar
Chart Row Header
A bar chart is displayed in the table row header
when a numeric column is selected in the table.
This graphical element compliments the color
shading of the selected column.
When the selected column contains non-numeric
values then the row number is displayed.
A tooltip is displayed when the mouse is moved
over the row header. The tool tip provides a
quick comparison of a transaction path's metrics
and statistics. This view is very helpful in
assessing what events (GC and thread monitor
contention) contribute to the total and maximum
wall clock times. The max graphical bars have
lines drawn indicating the minimum and average
values.
Table
View Thumbnail
A
thumbnail of the table can be displayed by clicking
on the icon button in the bottom left hand corner
of the table view. The thumbnail provides an
analysis
of numerical columns in the table. The
color selected for rendering of each individual
cell is derived from the ratio of the cell's
value to the maximum column value.
The thumbnail can be used to quickly navigate
the table identifying high values in a large
set of columns. This is especially useful in
correlating high column values across columns
as well as identifying outliner's not easily
identified by sorting because of the number
of columns required to be sorted. Within the
thumbnail the selected column and visible table
view portion are indicated to aid navigation.
William Louth, JXInsight's Product Architect
