The Client
Fortune 500 packaging supplier
The Challenge
This client’s IT department implemented several IT network performance monitoring tools for tactical performance troubleshooting but wanted to use this same data strategically to:
- Determine upgrade requirements by answering questions on server utilization and end-to-end user application–response time trends in summary and detail by location and device
- Ensure that established internal system Service Level Agreements are maintained so that users can perform their jobs efficiently
- Provide proactive maintenance and improvements to end-user response times on applications by affecting the appropriate mechanism—the network, server, or both
- Identify when external network support vendors are noncompliant with Service Level Agreements because noncompliance could justify service cost reimbursements
- Determine end-to-end IT hardware utilization for capacity resource planning
Business goals
- Create a single repository of historical network and server performance measures of data extracted from multiple real-time performance monitoring applications
- Allow the IT department to be more proactive in strategic planning and decision-making for the organization’s infrastructure by:
- Having the ability to interactively, in one tool, analyze end-to-end network performance metrics and the multiple factors affecting it
- Monitor performance trends and substantiate them against system Service Level Agreements (SLAs)
- Provide administrators with the ability to:
- Monitor network and server performance for adverse trends
- Respond quickly to issues adversely affecting performance
- Manage the utilization of servers and network devices before they adversely affect performance
The Haverstick Approach
The Haverstick team designed and implemented a data mart using Microsoft’s OLAP Services and SQL Server. The approach required the extraction of data from two data sources: NetIQ’s AppManager SQL database and Packeteer’s PacketShaper Backend data repository. The AppManager tool provided memory, CPU, and disk space utilization metrics. The PacketShaper repository contributed metrics on bandwidth utilization, traffic class protocol activity and TCP efficiencies from outlying locations to the corporate data center.
The solution used Data Transformation Services (DTS) to directly pull data from AppManager into the data mart utilizing a star schema design. First, the data is pulled from a decentralized backend repository for each location and transformed into the SQL server staging database. Then DTS is used to pull the data in the staging database into the data mart utilizing a star schema design. The fact tables for both data sources are then transformed into multiple OLAP cubes using Microsoft OLAP services.
The Performance Data Mart is refreshed with current data from 30+ locations on a daily basis. End-user access is provided using the ProClarity OLAP tool.
The Result
Now IT managers and administrators can be more effective at managing performance issues and improving application response times by using these applications and tools to answer questions about historical date such as:
- What are server CPU, memory, and disk space utilization trends for each location by month, by operating system, for the past year?
- What servers exceeded daily performance utilization thresholds more than 25% of the month? Is this a trend or just a one-time situation?
- What application classes are consuming the most bandwidth at a particular plant?
With the capability to answer these questions, the IT department is more proactive in maintaining and improving end-user response times as applications and systems are implemented corporate-wide. By monitoring traffic class activity for a location with limited bandwidth, they were able to identify and de-allocate low-priority classes and re-allocate bandwidth to high-priority protocols without spending nonbudgeted dollars to increase overall bandwidth.
In another instance, technical support used the data mart to evaluate server metrics over time and determine what metric trends increased capacity requirements and what metrics were exceptions. Using this information, they were able to identify and replace only the pieces of the server that needed upgrade to handle the additional capacity requirements.
In addition, the client not only can interactively analyze the metrics in the cubes by multiple dimensions at their workstation, they can also create active Web-page views of the data and communicate the information to others in the organization by posting them on the corporate intranet. | 
