Generic ESX Designs

[drewdat]

I was wondering if anyone knew of a good place to look for generic ESX infrastructure design scenarios? I've been spending a lot of time looking through threads in the Enterprise Strategy and Planning section, but I'm having difficulty assimilating the information. Apparently there are many ways to skin this cat and for each way there are dozens of vendors to sell you different knives, if you will. I was hoping to find some sort of document/article/post/etc. that would give an overview of the basic directions you could go, the approximate budget needed for each, and the different vendor solutions that fall under each. I've seen a few posts that have broken down pieces of a design into generic options and have found those very helpful (e.g. these are your basic options for replicated offsite storage: external DR tool like vReplicator, offsite virtual storage like VSA, etc.). Does this exist? Could someone point me in the right direction?

[Texiwill]

Hello,

THere is not a 'one design fits all' scenario. The best I can suggest is let us know what you want to do, then we can assist, in the meantime check out the Virtualization Bookshelf for some reading material.

Best regards,

Edward L. Haletky

VMware Communities User Moderator

====

Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.

Blue Gears and SearchVMware Pro Blogs: http://www.astroarch.com/wiki/index.php/Blog_Roll

Top Virtualization Security Links: http://www.astroarch.com/wiki/index.php/Top_Virtualization_Security_Links

[drewdat]

I realize that virtualization is not a "one size fits all" deal. I guess I'm looking for info on some high-level design options. It seems like most of the info out there assumes you aleady know this and delves right into the nitty-gritty. I think this high leve generic design info would be especially helpful for the SMB market (who often don't have the resources to be or hire VM experts). With an understanding of the high-level design concepts then I can start making more sense of all the nitty-gritty. I'll take a closer look at the link you sent me. I'll open another thread to ask for recommendations based on the specifics of my environment. Thanks for your help!

[Texiwill]

Hello,

High Level for SMBs generally boils down to:

How much can we get for free and how much Redundancy can I afford. With Backups thrown into the design as an after thought, and generally absolutely no thought towards security because it costs too much. However, that aside, their are plenty of secure SMB designs available. The general SMB design for a company depends clearly on the number of VMs to host and the utilization of those VMs. For example:

1 SMB had 7 hosts, one of which would get pummeled regularly with LARGE file transfers.....

Another SMB had 50 TS users + 2 databases that had to be up 100% of the time during the day and perform very very well.

Another wanted redundancy without paying for it for just 4 VMs

So you can see, each of these has a different design goal and truly depends on what they wish to do, how much they wish to spend. Also in each of these we designed security into the picture from the beginning with very little in the way of extra expense.

Designs for SMBs usually start with hardware on the HCL, but not necessarily fully loaded, they can upgrade later, usually 2 nodes minimally with maybe some form of shared storage, or the other node is the recipient of some form of replication just in case one node fails.

Others want full VMotion capability with more than enough nodes. It boils down to budget.

Best regards,

Edward L. Haletky

VMware Communities User Moderator

====

Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.

Blue Gears and SearchVMware Pro Blogs: http://www.astroarch.com/wiki/index.php/Blog_Roll

Top Virtualization Security Links: http://www.astroarch.com/wiki/index.php/Top_Virtualization_Security_Links

[drewdat]

My situation is this:

We have about 70 workstations in our main office and 10 more in a branch location. I have around 50 guys in the field with tablet PCs. We have been running VMware Server for the past 3 years running 11 VMs. Right now we have two Dell 2850s (2 X 2.8GHz Dual Core XEON) connected to a PowerVault 220s with 2 seperate banks of hard drives. My current backup strategy is to backup the network with 1 LTO3 tape everyday and take it offsite.

I'd like to move to ESX or ESXi and build in some redundancy and load-balancing. How much downtime we can allow depends a lot on what it is going to cost (obviously no downtime would be ideal). I'd like to shoot for a maximum downtime of 4 hours or so. I'm hoping to build a solid infrastructure that will allow us to scale and expand capabilities (e.g. adding offsite DR down the road). Performance-wise, almost everyone in the company is using one custom CRM-type Pervasive DB application all day long. This app has a lot of links to files on our file server. Those in the branch office are hitting this app via Terminal Server over a P2P T1. Those in the field with tablet PCs are passing back and forth small bits of info with the occasional larger picture or pdf to this same Pervasive DB.

I'm a little unsure of our budget at this point, but I'm thinking around $30k. I think the budget can be somewhat flexible if needed.

Any advice would be great.

[meistermn]

Get the facts . Measure this environment. Expect different results in a Virtual Environment.

What is the workload of workstations? Where is the bottleneck. CPU, MEM, NETIO, Disk IO and/or Graphic.

Disk IO is the slowest component in physical and virtual world. Keep an extra eye on it.

[Texiwill]

Hello,

What you describe is not impossible and you may be able to do it within your budget but as meistermm has stated run some analysis on your existing TS, database, and other servers to determine if they are really good candidates as they are or will you need more VMs to make TS function properly.

It is possible.... Depends on the level of redundancy you really require.

Best regards,

Edward L. Haletky

VMware Communities User Moderator

====

Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.

Blue Gears and SearchVMware Pro Blogs: http://www.astroarch.com/wiki/index.php/Blog_Roll

Top Virtualization Security Links: http://www.astroarch.com/wiki/index.php/Top_Virtualization_Security_Links

[drewdat]

All of my servers are already in a virtual environment (VMware Server) so I shouldn't have to worry about that (unless it will vary a lot compared to ESX). Any reccomendations on the tools to use to take these measurements (other than perfmon)?

[Texiwill]

Hello,

If they are all VMs already then its not really an issue. The key now will be how much redundancy do you want? Minimally you may want:

2 ESX hosts w/shared storage + enough local storage to run your most important VMs and minimally 6 pNICs if you are using IP Storage. 8 would be better.

2 Quad Cores would be sufficient with enough memory to handle the load.....

Storage is the big ticket item..... iSCSI (should have its own dedicated network), FC needs its own infrastructure..... Either way its expensive.

Best regards,

Edward L. Haletky

VMware Communities User Moderator

====

Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.

Blue Gears and SearchVMware Pro Blogs: http://www.astroarch.com/wiki/index.php/Blog_Roll

Top Virtualization Security Links: http://www.astroarch.com/wiki/index.php/Top_Virtualization_Security_Links