Intermittently slow performance staging data from DLP after migrating to virtual machines

Products

• Information Centric Analytics

Issue/Introduction

After migrating the server hosting Information Centric Analytics' (ICA) RiskFabric relational database from physical hardware to a virtual machine (VM), the RiskFabric Intraday Processing job's staging step began running much longer than its historical average. The nightly RiskFabric Processing exhibits the same behavior.

Environment

Version : 6.x

Component : Symantec DLP Integration Pack

Cause

When a VM is configured to use the Balanced power plan in Windows, Windows will idle communications between the SQL Server database engine and the Oracle Client (Oracle OLEDB provider) during times when the RiskFabric database is not being actively queried by the application. This occurs despite the fact that SQL Server is actively waiting on the result set from Oracle provided by the Oracle client.

Resolution

Per the Server Requirements for Symantec ICA section of the Symantec ICA Administrator Guide, Windows' High Performance power plan is required for use on all servers hosting ICA's components (IIS, SQL Server, Analysis Services).

Additional Information

This condition affects staging queries that return large data sets, such as the active incident staging query that populates the table Stg_SymantecDLP_ActiveIncidentList. This can be seen in the Log_DataTransformation table by executing the following query:

USE RiskFabric;
GO

SELECT   TOP 10
         StartDate,
         EndDate,
         RunMinutes,
         RowsAffected
FROM     dbo.Log_DataTransformation WITH (NOLOCK)
WHERE    LogName = N'Stage Stg_SymantecDLP_ActiveIncidentList'
ORDER BY 1 DESC;