I had to give myself a reality check today. I have always wanted to have my own personal enterprise class network. Let me explain what I mean. I enjoy learning new things or learning new ways to do old things. I have been planning my dream network that I want to have when I actually get a house, but today it hit me that the cost of such an enterprise network wouldn't be justified if the resources were never even utilized above 1%. Heck even .05% at best. Might as well change my name to that resembles how crazy I am for thinking and wanting this in my house.

I enjoy playing around with and setting up apache web servers, mysql database servers, mail servers, dns servers, routers, firewalls while integrating centralized authentication where possible. If it includes Linux in the mix I am even more apt to peruse it. In an enterprise class network, you often have so many users accessing a resource that it requires a dedicated host or dedicated redundant hosts. An example..

X Crop has close to 10,000 and 30,000 users at one site. To ensure that information is able to accessed and critical services are available they implement the following. The client machines are windows systems while the server systems are Unix based.

2 Active Directory Servers configured in Master/Slave Configuration
2 Linux Bind DNS Servers in master/slave Configuration
2 Mysql Servers with replication and fail over
3 Load Balanced Apache Webservers
2 NetApp Storage Appliances with 120 TB Capacity Each
4 Linux Servers for dedicated to compiling and testing source code
2 Subversion servers
2 Linux Mail Servers in master/slave configuration
1 LDAP Server (All Unix systems utilize LDAP Authentication to Active Directory)

You get the idea, basically I wanted to build an enterprise class network in my house to have as my playground with dedicated hosts for common enterprise applications. Before today you couldn't have convinced me otherwise to change my dream network. Today though that changed and I actually asked myself aloud, "What is the reality of my need for such a network and can the cost be justified?". I started thinking about how much electrical power would be required and how much cooling would be required just so I would have an enterprise network of my own to play around with. I already work in an enterprise environment, but I can't break things for the fun of it and try to fix it just to hone my skills. My servers would not have been powerful but the idea behind it all was dedicated enterprise services.

The other thing I wanted to do was be able to bring all of my website hosting and database hosting in house, because I hate the idea that my sites are shared with other sites whom may be very insecure and allow the host to be compromised. However until I can find a cheap place to co-locate a set of servers or pay the extreme prices for dedicated hosting I am forced to continue to use shared hosting. I also recently heard that Verizon frowns upon using their home FIOS to host and serve your content over their network, weather or not this is true I am not sure. I have yet to read to TOS.

Here is what I came up with for my amended dream network .

1 Bind DNS Server
1 LTSP Server
1 OpenFiler
1 VMWare ESX Server
1 Bacula Backup Server

Currently thinking about my need for an LDAP Server locally and if I would want to have and Active Directory Server to play around with as well. Micro$oft in my home :barf: I couldn't fathom it but I need to keep up on both my Windows Server Skills and Unix/Linux Skills. If I do I likely want to/try convince myself that I "need" two more servers.

The other thing I am starting to think about and ask myself is how much storage do I really need? Could I really utilize a 50 TB System without filling up a data storage system with movies. I don't own or watch television so my need for a place to store downloaded movies isn't a worry. I do know that if I had a large data storage location in house I would have a few local Linux mirrors to make installing net installs much faster. I don't even game anymore, which thinking about that is kind of sad. It's like I grew out of gaming.

Has anyone else here had to give themselves a reality check when trying to achieve their dream home network.

Been there, done that! I would advise to set up the network if you are interested, just don't leave it all running It's good to teach yourself new stuff and to play with, but can get expensive to keep running

Just build a quad core computer and create virtual servers for most of what you wanted. You can even install additional network cards into that system so that each one has its own card.

This is pretty much what I did to solve this issue. That and I use a atom cpu mini computer as a NAS and misc server that is running all the time. It uses a lot less power then a normal computer or my monster quad core machine.

Its kinda better to get familiar with all the virtual stuff anyway since that's where most companies are going these days.

Now I just have to figure out what to use to build each one. For the filer I am thinking about Dual Quad maybe, or possibly a single Intel i7-980x Gulftown Six core system. It's overkill for a Filer I know but I want to make sure it a future proofed, and I mean really future proofed, 10 years should be enough. Heck the Single Six-Core would do just fine I think. The filer isn't going to be doing much but serving up blocks up data.

The other systems I don't now about yet. Tempted to go all out on the ESX System, because it will be doing a lot of VM work. And by All out I mean a Quad 1207(F) Six-Core AMD system, giving me a total of 24 Cores.

I think that reality check is wearing off!

Unless you are exploring a system in which multiple cores are essential (perhaps exploring complex high-transaction rate databases) then all those cores would not do anything for you.

The key concept is "modelling". I worked on the development of a total solution for ISPs for one of the biggest international systems houses. In the development environment we used the smallest rack-mount Sun "servers" because the had as much capacity as we needed to develop the system.

Historical hardware performance/cost figures show that "future proofing" is a marketing phantasm, designed to get customers to spend more than they need. Hardware is still evolving fast. The best long-term performance/cost strategy is to buy "sweet spot" hardware and replace it with higher capacity hardware when a) there is a need and b) such higher capacity has become the norm and so is priced as a commodity and not exotically.

Just my . If you get a kick out of having the latest and greatest then go ahead, buy it and enjoy!

Catkin,

This is true. It must be wearing off. I'm still looking into hardware and ESX. Looking to build a system that will Maximize ESX's use of the hardware. I've never built and ESX System so I am still in the research stages as to what I will need vs what I would want. Sure 24 cores is a dream system but if ESX can't fully utilize the hardware and all the cores then there really is not point to such a system being built other then to say you spent a lot of money.. But one can dream...

But seriously I am thinking about building these systems stated above but I want to make sure the that hardware will be properly utilized by the software running on them. But at the same time I can dream about the perfect 24 core system where the hardware is perfectly utilized and everything is in harmony. Sounds like I been toking the reefer to much, eh? Actually never touched the stuff, but people often wonder about me. LOL

I'm also currently at the point of trying to figure out how much disk I should buy. I would like to build a 50TB System, but do I really NEED 50TB. I would like to have a mirror copy of a few Linux Distro's that I do net installs with so I don't have to use up precious bandwidth. Like I said before I don't download movies to my storage doesn't need to massively huge. Currently trying to calculate how much data storage space I should get. 10TB seems like quite a lot, but I guess I could start there and always expand if needed. I hate reality some times.

Today is a sad sad day. I was bidding on a 4GB EMC2 Fibre Channel Switch and was the winning bid at 112 dollars with my bid limit set at 175. At the last second and I mean the last second some IDIOT came in and swiped it from me for 177 dollars.

I was wanting it for when I go to build my open filer. I wanted to have a fibre channel network so that I could have data transfers at block level rather then file level and also have 4GB of bandwidth without bonding nics.

Now I have to give myself another reality check and ask myself if I should continue the network route of storage of just build a new desktop and have direct attached storage through the use of a RAID Card. But if I do that I won't really be able to have normal desktop case as I know the case I would want would need to have hotswap bays.

The cost of building my filer will be the same is if I build a new desktop. To network my storage or not to network. That is the question. I'm single so I don't have a family unit who needs to share data.

I do want to have a VM Server and my thought was to have an LTSP server and use Thin Clients as the desktop unit since I don't need a powerful system. Then I was going to have storage shared between the VM Server and LTSP Server using the OpenFiler box.

I am thinking about scrapping my whole filer project and just building the LTSP Server with all the storage I would ever need, or I could just build myself a desktop with parts I don't really need such as gaming graphics and such just because I can.

Today is a sad sad day.

Pity about the Fibre Channel Switch. For that sort of price ($118! Or could the buyer's bid have been automatic?) it would have been money well spent on not just something bigger but a different technology to become adept at using. Maybe organisations that have a business need for that sort of technology are up-scaling or updating and will be offloading more so you will have another chance

I call ebay the dog effect.
Bidding is like 10 dogs going after 1 bone.