You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: docs/dea/database-experimentation-assistant-view-report.md
+36-38Lines changed: 36 additions & 38 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -2,7 +2,7 @@
2
2
title: View analysis reports for SQL Server upgrades
3
3
description: View analysis reports in Database Experimentation Assistant
4
4
ms.custom: "seo-lt-2019"
5
-
ms.date: 01/24/2020
5
+
ms.date: 02/04/2020
6
6
ms.prod: sql
7
7
ms.prod_service: dea
8
8
ms.suite: sql
@@ -16,69 +16,67 @@ ms.reviewer: mathoma
16
16
17
17
# View analysis reports in Database Experimentation Assistant
18
18
19
-
After you use Database Experimentation Assistant (DEA) to [create an analysis report](database-experimentation-assistant-create-report.md), use the steps below to review the report for performance insights based on your A/B test.
19
+
After you use Database Experimentation Assistant (DEA) to [create an analysis report](database-experimentation-assistant-create-report.md), you can review the report for performance insights based on the A/B test you performed.
20
20
21
-
## Select a server
21
+
## Open an existing analysis report
22
22
23
-
1. In DEA, select the menu icon. In the expanded menu, select **Analysis Reports**next to the checklist icon to open the Analysis Reports window.
23
+
1. In DEA, select the list icon, specify the server name and authentication type, select or deselect the **Encrypt connection**and **Trust server certificate** check boxes as appropriate for your scenario, and then select **Connect**.
24
24
25
-
2. Under **Analysis Reports**, enter the name of a computer running SQL Server that has an analysis database, and then select **Connect**.
25
+
26
26
27
-
27
+
2. On the **Analysis Reports** screen, on the left-hand side, select the entry for the report you want to view.
28
28
29
-
If you're missing any dependencies, the **Prerequisites** page prompts you with links to install them. If necessary, install the prerequisites, and then select**Try again**.
29
+
This section walks you through the analysis report.
38
34
39
-
You can get insights into how well your workload is represented, as shown in this example chart:
35
+
On the first page of your report, information about the version and build information for the target servers on which the experiment was run appears. Threshold allows you to adjust the sensitivity or tolerance of your A/B Test analysis. By default, threshold is set at 5%; any improvement in performance that is >= 5% is categorized as ‘Improved’. The dropdown allows you to evaluate the report with different performance thresholds.
You can export the data in the report to a CSV file be selecting **Export** button. On any page of the analysis report, you can select **Print** to print what is visible on the screen at that moment.
42
38
43
-
##View and understand the analysis report
39
+
### Query distribution
44
40
45
-
This section walks you through the analysis report.
41
+
- Select different slices of the pie charts to show only the queries that belong to that category.
46
42
47
-
### Query categories
43
+
48
44
49
-
1. Select different slices of the left pie chart to show only the queries that fall under that category.
45
+
-**Degraded**: Queries that performed worse on Target 2 than on Target 1.
46
+
-**Errors**: Queries that showed errors at least one time on at least one of the targets.
47
+
-**Improved**: Queries that performed better on Target 2 than on Target 1.
48
+
-**Cannot Evaluate**: Queries that had a sample size too small for statistical analysis. For A/B testing analysis, DEA requires the same queries to have at least 15 executions on each target.
49
+
-**Same**: Queries that have no statistical difference between Target 1 and Target 2.
50
50
51
-
51
+
Error queries, if any, are shown in separate charts; a bar chart classifying errors by type and a pie chart classifying errors by Error ID.
52
52
53
-
-**Degraded queries**: Queries that performed better in A than in B.
54
-
-**Errors**: Queries that show errors in instance B but not in instance A.
55
-
-**Improved queries**: Queries that ran better in instance B than in instance A.
56
-
-**Indeterminate queries**: Queries that had an indeterminate performance change.
57
-
-**Same**: Queries in which performance stayed the same across instances A and B.
Clicking on any bar or pie section in the charts drills down into the category and shows performance metrics, even for the **Cannot Evaluate** category.
64
63
65
-
- Select a specific query to open the related comparison summary.
64
+
In addition, the dashboard shows the top five improved and degraded queries to provide a quick performance overview.
You can see the A and B instances on which the query ran. You can also see a template of what the query might look like. A table displays query information that is specific to instances A and B.
68
+
You can select query template links for more detailed information about specific queries.
70
69
71
-
### Error queries
70
+
72
71
73
-
The comparison summary report has expandable **Error Information** and **Query Plan Information** sections. The sections show the errors and plan information for both instances.
72
+
- Select a specific query to open the related comparison summary.
74
73
75
-
- Select the error (red) pie to show these types of errors:
-**Resolved errors**: Errors that were in A but not in B.
76
+
You can find summary statistics for that query, such as the number of executions, mean duration, mean CPU, mean reads/writes, and error count. If the query is an error query, the **Error Information** tab shows you more detail about the error. On the **Query Plan Information** tab, you can find information about the query plans used for the query on Target 1 and Target 2.
0 commit comments