Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Designing and Deploying a Scalable EPM Solution Ken Toole Platform Test Manager MS Project Microsoft Agenda Top 5 reasons a deployment doesn’t scale Understanding the EPM Design process Build the implementation team Logical Design of Project Server Key architecture factors for Scalability Top 5 Scalability Killers 1. Inadequate network infrastructure 2. Inadequate database hardware 3. No Pilot 4. Undefined project management practices 5. Unrestricted access to data EPM Initiative Development Develop vision for EPM Stake-holder identification and buy-in Collect Business Requirements EPM strategy development THEN … EPM solution design EPM Design and Planning Implementation Team Large scale Project Server deployments are complex and require varied expertise System Administrator Server Install, Clustering, Network Configuration, Shared Storage, Active Directory, Domain Name Resolution, Security Database Administrator DBMS installation, Drive Configuration, Tuning, Backup and Recovery, Security Implementation Team Network Engineer Router and LAN configuration, loadbalancing hardware, Firewall configuration, security Project Server Deployment Specialist? New training and certification exam Capacity Planning Metrics Scalability Factors Configuration Options Scalability Metrics Users per hour Max number of times a specified set of user operations (workflow) can be completed per hour. Focus on peak times like weekly time reporting. Size and complexity of data will effect this metric. Not a generalized or generic metric Cube Build Time Views Publishing throughput Key Scalability Factors Network performance Database server performance Project Server security usage Project Server data usage Network Performance Most frequent bottleneck for most installations Indicated by poor performance or throughput with low cpu utilization on all tiers of the application Primarily the result of latency rather than bandwidth, and the overhead associated with each network “roundtrip” Network Performance corrective actions Network infrastructure 100mb minimum between servers 1Gb Ethernet or fiber between server components option. Network Interface Card settings Avoid ‘hardware default’ On some systems consider 100mb/full setting instead of ‘auto-detect’ Test and consult your hardware vendor Hardware Options ‘NIC Teaming’ configurations Unisys style ‘virtual machine’ configuration (only suitable on very large hardware platforms) DB Server Performance Typical DB server performance analysis and tuning. Often will present as a bottleneck after resolving any network issues Solutions are simple but can be costly in terms of hardware. DB Server Performance corrective actions Data file configuration Separate transaction log and data files on separate logical drives Use db optimization scripts to separate Project Server data and indexes into multiple file groups Place index and views file groups on separate logical drives than other Project Server file groups Disk I/O hardware Use a larger number of physical disks per logical drive to ensure disk read/write activity is not bottlenecked on few physical drives Use hardware RAID controllers instead of Software RAID controllers Consult you hardware vendor DB Server Performance corrective actions Server hardware Carefully test hyper-threading configurations and SQL Server settings related to threading and fibers Consider faster processors and increased number of processors Partition Project Server data onto multiple DB Servers Project Server Security What data a given user has access to impacts the amount of data retrieved and transported throughout the system Use of restrictive permissions to limit the amount of data available to Project Professional users is recommended. RBS based security is optimal Project Server Data Usage Project Management process and standards impact the way data is grouped in the system. Find a balance between the number and size of projects. When project sizes exceed ~ 2000 tasks consider phasing these into separate projects. Use ‘View in MS Project Professional’ feature from Project Center and enterprise outline codes to work with ‘partitioned’ projects efficiently. Avoid PM Practices that require a large number of project baseline saves. Consolidate non-project or Administrative tasks into a small number of projects Front-End Configuration Clustering Options No Cluster Single Cluster Project Server WSS Dual Clusters WSS Project Server WSS PS Project Server WSS WSS WSS Project Server Project Server Clustering Factors WSS and Project usage balance Availability requirements Future Growth Front-End Hardware RAM and Network 1 GB of RAM recommended 2 GB if using Portfolio Modeling features 2 to 4 processor commodity web servers Hyper-threading is ok HW accelerator cards can help in SSL implementations Middle-Tier Configuration View Processing Options All on One Dedicated Project Server Views Processor DB-Based Project Server Views Processor Project Server DB Server Views Processor View Processing Factors Size of typical projects Number of Project Managers Frequency of Publish Frequency of OLAP Cube Build View Processing Hardware CPU and Network Fast single processor speed Disable Hyper-threading High-speed link to DB server Session Manager Options All on One Dedicated Project Server Session Mgr Multiple Project Server Session Mgr Project Server Session Mgr Session Mgr Session Manager Factors Number of concurrent users Front-end clustering Availability Session Manager Hardware RAM 1 GB + recommended Minor impact on processor utilization Database Configuration Database Options (Split) All on One Two-Way Split PS Database Three-Way Split PS Core Database PS Views Database PS Web Database PS Views Database PS Projects Database Database - Factors Size of typical project Number of project managers Frequency of publish and cube build LOB or Custom Integration Database Hardware Typical SQL Server behavior CPU, RAM, Disk I/O speed 1 GB RAM per CPU for rule of thumb Hyper-threading is ok Memory configuration Maximize memory available to SQL Avoid large amount of memory configured for fail-over © 2004 Microsoft Corporation. All rights reserved. This presentation is for informational purposes only. Microsoft makes no warranties, express or implied, in this summary.