Using Multiple Workstations as Servers Instead of Virtualizing

BigMac

Member
Reaction score
3
Location
North Carolina
Don't get me wrong, I fully understand the importance of having quality equipment for running a server. I just don't see the purpose of having a $5000+ server for doing it.

Scenario: 50 employee business needs in-house hosted email, web, file, Asterisk, and terminal server. Instead of getting one or two high end servers with dual Xeons, 32GB RAM, RAID 10, etc, why not get 5 Dell Optiplex 780s with Core2Duo's, 4GB of RAM, and install single 10,000RPM hard drives. All servers will be running Linux except for the terminal server. Do daily images of the hard drives to a NAS box or similar and keep two 780s on standby for spares. You can get refurb 780's for around $200-$300. The chances of one of these dying is more likely than a high end server but I can't see by much. You save lots of money and when one dies you lose one service instead of everything.

Thoughts?
 
This is NOT a solution and if your customer values their business then you should not even be flirting with this. Plain and simple.
 
While what SAG said is true in the spirit of best practices, let's talk it out.

Even though the OP's scenario is a 50 employee company which should easily be able to afford a "real" server, many businesses face this decision every day.

So I'll jump in on the opposing side.

If the following were true:

1. budget is truly tight - not just a cheap decisionmaker
2. client can accept/absorb/doesn't care about the downtime required to replace a failed workstation-server
3. daily imaging actually takes place
4. the client CLEARLY understands the risk/reward of the decision

Then why not?
 
Thanks for your opinions. I'm assuming that reliability is the big factor here as none of the applications I listed are very power hungry except for terminal services.

How much more reliable are "real" servers vs workstations? I've seen both have hardware problems. I've sold dozens of Optiplex's over the last few years and never have seen an issue other than lightening. Also done many server repairs with issues from hard drives to motherboards. Everything has the possibility of failing. Some things are just better engineered than others. Are the higher end product really worth buying if you have an IT dept. and have a hot spare on hand?
 
My first post is always my stance. If a business owner care about his business' productivity, then they should not completely "cheap out".

Of course, mraikes has a point - it can be done. And if pains are taken it can be successful. Let's have a look at your OP and break it down shall we, I'll play devil's advocate to myself here.

Don't get me wrong, I fully understand the importance of having quality equipment for running a server. I just don't see the purpose of having a $5000+ server for doing it.

If they wanted to be cheap and still get decent performance. It can be done for half that using server grade hardware. Albeit, not the best of hardware though.
Chasis: http://www.newegg.com/Product/Product.aspx?Item=N82E16811152170
Motherboard: http://www.newegg.com/Product/Product.aspx?Item=N82E16813182240
CPU: http://www.newegg.com/Product/Product.aspx?Item=N82E16819113316
Memory: http://www.newegg.com/Product/Product.aspx?Item=20-139-140&Tpk=N82E16820139140

Building that with 32GB of the memory I posted would cost about $1000 in hardware. ESX on a USB drive is another $5. All that is left is figuring out a datastore. You can buy/build a NAS to act as an NFS mounted datastore, or...
For onboard storage you can use a Dell Perc6 RAID controller from ebay for $80 and 3 10K Sata drives in a RAID5 for datastore at $450
http://www.ebay.com/itm/DELL-POWERE...sk_Controllers_RAID_Cards&hash=item1c31af85ea

http://www.newegg.com/Product/Product.aspx?Item=N82E16822236244

Now your total is about $1500 (give or take) and you have an ESX host with 8 physical cores, 32GB memory, and a *decent* datastore.

That took me about 8 minutes to figure out. But I work so much with ESX I know off the top of my head alot of what will work and what won't. For instance, I know the PERC5 RAID controller won't work with the mobo I linked unless you block off the SMBus connectors.

Of course, this is all just me rambling here. You very well could get an awesome used Dell 2900 series with plenty of 146 SAS drives from ebay for under $1k.

That takes care of the first sentence of your OP. :)
 
I don't even know where to begin, other than "Go back to IT 101 class".

But to start scratching the surface at least...here are a few of at least a hundred reasons..

$5,000.00 isn't a "high end server"......you start getting into high end after you get 2x digits in front of the comma in the price tag. $5,000.00 is where a decent entry to almost mid level server starts....once configured.

However for most people in this forum, IT consultants, catering to the SMB market, the 5-7k range for a single server is common. You typically don't need a huge honkin high end server in this market.

First...good servers with hardware RAID and hot swap drives...you will not experience downtime if a hard drive fails. Swap the failed drive with a good one, end users won't even notice a skip in the heartbeat. Using a desktop PC...you'll at least have downtime...from the time it crashes (all work stops right then)...until some IT person arrives and does a restore and/or clone from some backup drive. I'm counting at least several hours of downtime.

Performance. Desktop computers are built with desktop components. Good servers are built with...well....server components. Yes there is a difference in performance when you compare a desktop NIC against a server NIC once you start factoring in heavy loads under many concurrent users. HDDs...once you start having 20, 40, 60 or more people working with many files in shared folders....even the mighty WD Raptor HDD can't begin to hold a candle to good enterprise SAS drives on a good RAID controller.

How about line of business applications? Most businesses run at least one application shared by staff. Often several. Most of these applications are designed to run on Windows. Many of these applications are designed to run on a database engine. Whoops...that stinks, you only quoted the client a bunch of 3 year old Optiplex 780 models with a meager C2D CPU and only 4 gigs of RAM. Oh shoot...this line of business application is supported only on Windows Server. And it's crippled by a single spindle HDD setup. Yeowch why does it take forever to run something? A business of 50 employees is going to run quite a few different applications. Management/Administrative will have their own, Accounting and HR will have their own, Marketing will have their own, Sales and Production...and then you have at least one or two that everyone uses.

Easier to manage the backup when you keep all the company data on as few of servers as possible.

Time and experience have produced a proven track record for us in the IT field that illustrates.... "Once a business crosses the line of having 10x staff, 10x PCs...it becomes beneficial to simplify the network and structure it AROUND a server(s)."
 
I have a t610 running in my office, 64 GB ram, 2 x hex-core xeons with hyper-threading, 4 x 600 GB 15k drives, 4 x 1 Tb near-line drives, 2 x Intel pro dual-port nics, running 6 to 8 vms at any one time. The thing draws about 250 Watts under load. The san box with 12 x 10k drives draws another 100. 10 x 780s under load are going to draw well north of 1000 Watts. Light bill is going to get nasty. You should make sure they don't have a little 15-amp circuit feeding the server room. You'll also be generating a lot of heat. Better get one of those Mitsubishi AC units in there. They run about 2 to 3 grand. Let's not forget about the time you spend maintaining this monstrosity. Tell me: have you examined how much this setup is truly going to cost them? In a year? 2 years? 3 years?

I'm not sure of your experience. No smb consultant I know would even entertain this idea. Not just the cost, but the ridiculous complexity involved. 2 good servers with the right tech company will cost far less and be far more reliable than what you've proposed.
 
Stonecat and angry_geek pretty much hit this nail on the head already. I have been doing this a long time and cannot think of a single viable reason for choosing the suggested setup versus doing it the right way. Solutions aren't tagged as being "best practice" for no reason.
 
*raises hand*

server n00b here. I was just curious why people still go the high rpm sas drive route...in theory shouldnt a high end SSD's low access time prove far superior for constant small reads/writes?
 
Last edited:
*raises hand*

server n00b here. I was just curious why people still go the high rpm sas drive route...in theory shouldnt a high end SSD's low access time prove far superior for constant small reads/writes?

Yes . . . in theory. They also have a finite lifespan measured in read/write cycles. They are astronomically expensive (server class SSDs, not cheap sammys or crucial for your laptop). Performance is not that much better than 15k sas drives especially in RAID arrays with good controllers. They have a much higher failure rate than enterprise mechanical drives. If you have an environment with heavy database usage, the SSD can hit the end of its lifespan in less than 2 years. Expensive proposition when talking big data like we're seeing all over the place. It's not uncommon for clients to have DBs well north of 200 gb plus big file shares. 600 GB 15k drive is around $300 retail. 300 GB enterprise SSD is north of $1,200. I won't be justifying those costs combined with a required replacement cycle any time soon.
 
*raises hand*

server n00b here. I was just curious why people still go the high rpm sas drive route...in theory shouldnt a high end SSD's low access time prove far superior for constant small reads/writes?

*Still prohibitive cost compared to spindle
*My personal experience with the track record of SSD's is a failure rate near 50%. Now granted....this isn't on Intel or Crucial M4 SSD's...typically other brands like Transcend, Kingston, A-Data.
*With servers...typically the majority of the crowd stays with tried true and tested technology. The latest bleeding edge stuff is saved for less important things until substantial time has gone by to mature the technology.
 
My 2 cents:

"All servers will be running Linux except for the terminal server."

Just because its Linux doesn't mean it won't need a lot of resources,

Email: They will probably want exchange not some random Linux system. Although I have seen really good open source alternatives to exchange even those required server grade hardware not desktop computers.

File sharing: Yes you can stick free NAS on one of the systems but I doubt it'll be able to handle 50 people transferring files back and forth. Its a great system that works well but it only works well when the hardware can handle it.

Asterisk:Why would you put a cheep refurbished computer to handle the phone calls of a company?

That actually reminds me of a time when a new taxi dispatch company opened up they decided to get a nice office and all new stuff but when it came time to get VoIP they tried getting $20 phones off eBay and said that paying $30 per month for VoIP was expensive so they went with some random company that charged 15 per month. How long do you think they lasted? not very.

So yes you can have Linux but remember that all Linux alternatives and open source alternatives to Microsoft products still require the same type of hardware, maybe they use a little bit less resources but you still need a good system to be able to share files with 50 people.
 
Thanks for the info but personally I think you are all wrong. I think my idea is very good and I am going to give it a go.


No no I kid. But really, thanks for setting me straight here. It was just an idea and I wanted to see some commentary. I started looking into used server systems and surprised at how affordable they can be snagged up for, notably the PE2900 series that SAG mentioned.

The reason I posted is because I have been doing residential and SMB work for about 8 years. The SMBs consisted of 5-10 employees and I have sold a couple of entry level servers for SBS but recently I have set them up with cloud services such as 365 or Google apps, and used a business class QNAP NAS for file sharing which seems to work great. I would like to be prepared if a bigger client comes along that needs an database running and I need to present a solution. I have been approached by a few bigger businesses in the area with 20-25 employees and quoted a T620 for both situations but they both thought it was too pricy and moved on to the next guy.

Again thanks for the responses. I value your comments.
 
......... I would like to be prepared if a bigger client comes along that needs an database running and I need to present a solution. I have been approached by a few bigger businesses in the area with 20-25 employees and quoted a T620 for both situations but they both thought it was too pricey and moved on to the next guy.

When we lose a bid, we ask why they went with the other guy. price, approach, reputation? Even if they won't tell us the price, we we try to find what approach and hardware they are using.
 
I have been approached by a few bigger businesses in the area with 20-25 employees and quoted a T620 for both situations but they both thought it was too pricy and moved on to the next guy.

Again thanks for the responses. I value your comments.

I've been doing SMB networks for a long time...pretty much since Win3.1 and NT 3.5 days.

For "smaller" SMBs...we'll do lots of quotes....and find out they went with the cheaper guy.

Some clients will be cheap. And some of those clients will learn from that mistake...they will learn "you get what you pay for"...and they will be back, asking you to come and "fix things, make things better and faster".

I refuse to quote an entry level sub-par server just to get a job. I don't want to support that piece of crap. I don't want to be a "Linksys ranger" and cheap home grade equipment in a business.

Yup, we quote and use true servers....SAS drives, hardware RAID controllers, 300 series and up, HP Procurve switches, good biz grade workstations like Dell Opti or HP Elite.

If a client wants home grade workstations and sub 3000 dollar SATA/fakeraid based servers and sub 100 dollar switches...let some other IT guy get stuck supporting that pile of junk and constant complaints of slowness. I won't associate my name with a pile of junk like that.

Your goal should be to groom your base of clients......to be a bunch of clients that pretty much give you an open checkbook. They won't question you when you recommend this or that. THAT'S where you want to be.
 
I've been doing break-fix work for an OBGYN for about a year now. They are switching EMR systems and their current server (only 3 years old) won't support it. When the client asked about a new server I quoted them on a Dell T420- 6C Xeon, 12GB Memory, Raid1/Raid10 SAS, Server 2012/SQL 2012.

They were surprised how expensive it was. When I pointed out that their current server which was sold to them for only $4k by the VAR that sold them their EMR was obviously pirated software and that Windows/SQL on it alone was over half the price they paid, plus the fact that it was an Asus whitebox with 4GB memory, 3 Sata in RAID5, and that a majority of the performance issues they have been having was due to this they took deeper interest. It took me about 3 weeks to get them well informed.

Once well informed they understood why the new server cost so much. It is set to be delivered to my office on Saturday for me to setup their new EMR system on.

Sometimes customers don't understand. Sometimes money is a big concern until they are educated, then they realize that they need to spend the money. When all is said and done I know I will have a good server onsite there to support. The customer will be calling me less with performance related issues, and they will be happy. And a happy customer makes for a profitable relationship... both ways.
 
Back
Top