Total Cost of Ownership of BPPM - Part 12: Best practice - Fail to plan means plan to fail

I recently took a client project of BMC TrueSight/BPPM 9.6 implementation.  The client insisted to do the installation themselves and had me involved after the installation was completed.  When I was finally involved with the project after they spent weeks to finish the installation, I found out that many things were installed incorrectly: The BPPM server was under capacity, BPPM and Entuity were sharing the same server, the 2 BPPM Integration Service servers and 2 BPPM Cell servers were each configured as application cluster in an Windows OS cluster.  On top of that, there was no written architecture plan.

Needless to say, the installation needed to be re-done.  Had they had an architecture drawing or written plan before starting installation, and let me (or someone with BPPM architecture experience) review and confirm it, they could have saved themselves a lot of time in installation.

So what need to be planned before you start installation?

1) Server capacity: BPPM server is a heavy duty server.  It includes web server, app server, analytic engine, BPPM cell, and optional Sybase database.  All these components are memory "hungry".  Unless you are installing for a proof-of-concept environment, go for 32GB RAM.  If you are implementing both BPPM and Entuity, install Entuity on a separate server.  With a MySQL database on the same server, Entuity needs to have 32GB RAM too.  Virtual machines work OK though physical machines would give you more power.  BPPM Integration Service server is much less demanding and would work OK with 8GB RAM.  Each Integration Service (pair) can connect to 900 PATROL agents.  BPPM cell server needs 4GB RAM for one cell or 8GB RAM for 2-4 cells.

2) High availability: There are three different options for high availability - application cluster, VMWare cluster, and OS cluster (such as Microsoft Windows Cluster).  Application cluster can fail over at individual process level when a process fails.  OS cluster can fail over at host level when a process fails.  VMWare cluster can fail over at host level when a host fails.  For BPPM Integration Service servers and BPPM cell servers, configure them as application clusters if you have enough dedicated hardware.  Do not mix an application cluster with OS cluster.  For BPPM servers, since they don't support application cluster, use OS cluster if you have enough dedicated hardware.  If you don't have enough dedicated hardware and run everything on virtual machines, you can configure VMWare cluster on all servers. 

3) Integration: BPPM server can be integrated with many other BMC products such as BMC Sever Automation, Atrium CMDB, and Single Sign On.  If you plan to configure any of these integrations during BPPM server installation, you need to have detailed integration information ready.  However you do have an option to configure your integrations later.

4) Name convention: In additional to installation directory path, you need to have a systematic way to determine the name for each BPPM cell, BPPM integration service, and BPPM integration service cluster.  More on name convention planning was discussed in this earlier post: BPPM 9.5 Quick Start - Part 3: Name convention