I have called support and chatted with TechNet Concierge, and can't get an answer there, so this goes out to anyone who can help... I want to set up a NON production isolated testing lab (where at most I might have one other user via a laptop laptop for end user testing, so user traffic is not a consideration in resource allocation, where too many app servers or VM's concurrently installed on the same physical server would be), where I can concurrently install ALL microsoft application servers over time. The big question is regarding hardware and VM's and knowing how many to have right off the bat before I being the process and what effects this. The tricky part to this, is following the general server guidelines on each server product page outlines *best practices* for production environments, not for lab testing. My situation is squeezing the most software on as few physical boxes and VM's as possible for learning. A few specific considerations/questions about this: 1. There are certain servers that are required to be on two physical machines (or 2 VM's or a combination of both), like DC's, or the ability to test a server farm or clustering. So my biggest concern is how many bare metal servers I need or if literally everything can be VM's? 2. Many servers have additional server requirements and dependancies, like SQL server. It needs Active Directory, where many other servers will use SQL as it's back end database. 3. Many application servers use IIS. WIll IIS be on each individual VM or will each app server use a single installation of IIS, or will it be a combination of this. 4. The biggest question will be what should the initial physical server use as it's base bare metal OS and configuration? My initial thought was the first DC, but then I realized maybe that should be ISA or Forefront? But then I'm stuck with the chicken and the egg... If ISA/Forefront on baremetal, no DC to join to, but if I install the DC first then there is less security...? 5. My biggest sticking point is not knowing enough to do all this (which is why I'm doing it in the first place) but then also not wanting to re-do servers over and over and over to correct mistakes in how I set things up. And maybe I'm confused... maybe you CAN install every app server with every feature on one big beefed up server... someone please educate me! But remember to keep in mind, this is for testing and learning not production. I have a limited budget and I'm trying to balance doing as much as possible on as many VM's as possible to limit how many complete physical servers I need to buy... and by servers I mean loaded workstations I'm going to build from my local computer superstore or get from newegg.com, not actual real server hardware, I can;t afford that for personal use...~Michael

I threw up a Dell T710x64 w/ 2 quad core 2.2G and 16 G of Ram for testing. Right now have ESXI 4.1 with one 2008R2 DC, one 2008R2 with SQL 2008x64 w/analysis server, one Win7x64, and one 2008x86 with SQL 2005 w/analysis server. I'm using a little over half the cpu capacity and about 3/4 of the mem capacity. You can also ask around here. http://communities.vmware.com/home.jspa Regards, Dave Patrick .... Microsoft Certified Professional -Microsoft MVP [Windows]

Hi, Thank you for your post here. If you want to create a lab environment other than a best practice guideline for production environments, you can combine all roles of a product on a single VM. For those roles/services that will conflict each other, separate them in different VMs. It will maximize the utilization of the resources on your physical server. 1. There are certain servers that are required to be on two physical machines (or 2 VM's or a combination of both), like DC's, or the ability to test a server farm or clustering. So my biggest concern is how many bare metal servers I need or if literally everything can be VM's? How many services/products would you like to test in your environment? Will you concurrently test different services/products? Or will you test one services/products (exclude some infrastructure services such as AD) in a period of time? I think the number of bare metal servers you will need is based on your requirement. 2. Many servers have additional server requirements and dependancies, like SQL server. It needs Active Directory, where many other servers will use SQL as it's back end database. Make SQL as VM. Performance is not a big issue since it is test lab. Is that right? 3. Many application servers use IIS. WIll IIS be on each individual VM or will each app server use a single installation of IIS, or will it be a combination of this. If you connect the lab network and the productive network with NAT router, a central single point IIS server from which you can access all lab apps would be a good idea. Based on your requirement and the support of the application that you will test. 4. The biggest question will be what should the initial physical server use as it's base bare metal OS and configuration? My initial thought was the first DC, but then I realized maybe that should be ISA or Forefront? But then I'm stuck with the chicken and the egg... If ISA/Forefront on baremetal, no DC to join to, but if I install the DC first then there is less security...? Forefront ISA/TMG can work in the workgroup environment. 5. My biggest sticking point is not knowing enough to do all this (which is why I'm doing it in the first place) but then also not wanting to re-do servers over and over and over to correct mistakes in how I set things up. I believe that a testing lab is always dynamic environment. You may test one/some products today and test some other products several month later. Those products will have their specific roles and requirements on the infrastructure services. You cannot cover all testes with one-time created VM environment.