Hardware people

Do you have a VMware or another virtual infrastructure in place ? Can you transfer it to your virtual environment ?
Do you have a SAN ? Instead of buying a local RAID controller and disks, you can connect the server to a SAN but you need to purchase an HBA card.
Do you have a KVM ? Can you re-use your old monitor ?
Can you transfer your software licenses from your old server to the new one ?

AlexNguyen wrote: »

Do you have a VMware or another virtual infrastructure in place ? Can you transfer it to your virtual environment ?
Do you have a SAN ? Instead of buying a local RAID controller and disks, you can connect the server to a SAN but you need to purchase an HBA card.
Do you have a KVM ? Can you re-use your old monitor ?
Can you transfer your software licenses from your old server to the new one ?

No VMware or virtual infrastructure
No SAN
Yes on KVM so monitor is not necessary..
Transfering software license is an interesting idea, right now everything is on Server 2003.. it is a non 64 bit version so we will lose a lot of upgrade ability as far as memory goes but would certainly drop the price down..

Krunchi

1-800-Buy-Dell

RTmarc

You biggest expense is probably the SQL server licenses and the User CALs; which you'll need a total of 15 by the way if you have 15 users. Shop around. There's no reason to take the first proposal you get. Check with Cisco, Dell, and HP. Personally, I like the Cisco UCS platform and it usually comes out the cheapest or comparable to what you would see with Dell.

it_consultant

$17000 is what I expect out of the BOM you supplied. Most of the cost there is software, you can skimp and go cheap and buy dell, but that has a habit of biting me in the butt so I suggest you stay with IBM or HP. You could cut the cost by going with one Xeon instead of two, it is unlikely that even with one Xeon, the processor will be a bottleneck considering the number of users you have. You could try a cheaper backup solution, Symantec gets pricey with their licensing. Again, I have seen this done and more often than not, it isn't pretty.

Maybe you should look at cloud based servers which run on a subscription cost as opposed to buying it outright. I would rather you go cloud then buy cruddy hardware because you are cost constrained.

The hardware definitely is a bit much for 15 users, but it's going to be, and yes, a lot of that cost is software.

I agree with IT_consultant: Go cloud. Either you're cutting corners, getting way more than you need, or you're going cloud. Fifteen users simply doesn't justify the cost of acquiring and maintaining server hardware, software, and backups. If you look at it as a monthly cost spread out over five years, you're easily going to spend $1000 a month on that server, probably closer to $2000. Go see what you can get for that on Azure (quite a bit).

Do you have a job security ? If you decided to go cloud, maybe your employer won't need you anymore to maintain that server.

I did a quick search with suggested price on the web:

1) https://www-304.ibm.com/shop/americas/content/home/store_IBMPublicUSA/en_US/evp/x3500m3.html
IBM x3500 M3 7380MC1, 2 x Xeon E5606 2.13GHz, 16GB RAM, 4 x 146GB 15K SAS,
SAS RAID Controller, Windows Server 2008 R2 Standard + 10 CAL = $7166
2) Microsoft SQL Server 2008 R2 Standard - Compare Prices and Reviews on Microsoft 22809176 Database Management Software at PriceGrabber
SQL Server 2008 R2 Standard = $725
3) Microsoft SQL Server 2008 R2 - Compare Prices and Reviews on Microsoft 35905353 Database Management Software at PriceGrabber
15 x SQL Server 2008 R2 CAL = 15 x $197 = $2955
4) Amazon.com: Symantec Backup Exec 2012 Agent for Applications: Software
Symantec Backup Exec 2012 Agent for Applications = $777.50

With fifteen users, I'm not under the impression OP's primary job responsibility is maintaining that server. If it is, that is irresponsible at best. It is a disservice to the organization to act as an FTE to maintain one server. It's simply much more economical to outsource it, and anyone qualified to actual maintain servers would be bored to tears. If OP is maintaining a whole bunch of other stuff, one server moving to the cloud won't be the end of his job. Since OP is asking for hardware assistance, I will assume that maintaining the hardware is not his primary concern, and outside of that servers in the cloud still need to be configured and maintained by their owning organization's IT staff to some extent.

Really, Cloud is the way to go here. The other cost to consider is downtime. One server, no clustering, no redundancy outside of component-level means more downtime, period.

Heck, I would rather do a MacGyver of consumer-grade and SOHO-grade hardware and software. If this is your only server and it only does this function, I just can't see you spending this much on it. I mean, not knowing your software, you could probably run this in a VM on a two-year old laptop with an SSD and get reasonable performance.

When you put that much money into a server and backup for a small need like this, you're essentially spending 95% of that money for the data assurance and reliability of a server. It's not performance you need. A geographically dispersed cluster with rock-solid backups, guaranteed performance, operated and maintained by an organization staffed by the people that developed 95% of the software in use is simply going to be a better platform for your server, and in this case it will be a cheaper one, too.

As you said, we don't know what kind of software the OP has. We don't know if the OP has to be compliant with a standard. For example, we have some applications used by a small group of SCADA operators but we can't put it in a cloud, because of NERC CIP compliance and confidential datas.

paulgswanson

For the SQl have you ever tried NaviCat or MySQL? Its WAY cheaper that Microsoft SQL and most likely has more features. That should shave a significant chunk of the cash off the block.

First of all thanks for all the replies everyone. Job security is not an issue, this is only a small part of what I do.. The server that is currently handling this is a IBM 236 that we bought 7 years ago and just threw in the closet and is honestly still running fine but we are just concerned with the age so the CFO want's it replaced before we start having any issues.

Moving it to the cloud - I have been avoiding this due to the fact that I do not want everyone dead in the water if/when our internet connection goes down. Also we are only running a single T1 connection with around 40 users (only 15 use the server in question). At peak times our internet connection is already pretty slow, which doesn't really matter since none of our work requires the internet for anything other then email. I realize I could setup something like OpenDNS to try to regulate but would much rather have something in house.

I guess what I am trying to figure out is if the software company that we purchased our software from is just trying to get us to buy all this hardware as overkill or do you think it is really necessary? The items in the original post is what they recommend we purchase through them and they will set it up for us. The other option is if we build a server in house (or get it straight through IBM, HP, Dell) we can send it to them and they will configure it.

Does it sound like they are just trying to get us to purchase overkill hardware? Would the below hardware be sufficient for a server running a program that uses about a 5gb sql database with 15 concurrent users?

Intel Xeon E5620 Westmere 2.4GHz 12MB L3 Cache LGA 1366 80W Quad-Core Server Processor BX80614E5620 - $400

ASUS P6T WS PRO LGA 1366 Intel X58 ATX Intel Core i7 / Xeon Intel Motherboard $200

PNY 16GB (4 x 4GB) 240-Pin DDR3 SDRAM ECC Registered DDR3 1333 Server Memory Model MD8192KD3-1333-ECC - $140

(4 in RAID 10) HP Raptor EW222UT 160GB 10000 RPM SATA 3.0Gb/s 3.5" Internal Hard Drive - $720

Microsoft SQL Server 2008 R2 Standard with 10 CALs - $2000

SQL Server 10 additional CALS - $950.00

Windows Server 2008 R2 Standard (OEM) x64 - Server & 5 CALS - $550

Microsoft Windows Server 2008 10-Client Add On User CALs - $300

Total - $5260

cyberguypr

Wait, did they actually say that if you buy a box you'll have to physically ship it to them?

paulgswanson wrote: »

For the SQl have you ever tried NaviCat or MySQL? Its WAY cheaper that Microsoft SQL and most likely has more features. That should shave a significant chunk of the cash off the block.

Not really an option, this is a 'boxed' software which is basically a java application on top of a SQL database

Two options, they can configure it remotely or we can ship it to them, they will preconfigure it and send it back. If we ship to them the cost is less for setup then if they login remotely.

I don't think the company is trying to screw you around. The solution provided is fairly appropriate as an entry-level SMB solution. The hardware is definitely overkill in terms of performance, but good servers aren't cheap even with the weakest hardware available. Something with reasonable component-level redundancy, a good warranty, and reliable parts simply isn't cheap.

As far as your custom build, that all would work, but I would really buy an HP, rather than part a server out. Plus, you don't have backup in there. An HP server meeting the need will be around $6,000, add in software and backup and you're at $10,000 or so. Still cheaper than what you were advertised, but I'm sure they take a margin. If you really want to go the consumer-grade route, I would go even cheaper than that setup and build a desktop:
Quad-core Core i5 with cheapest stable motherboard available (~$350)
Dual SSDs in RAID1 ($350)
8GB of RAM ($~140)

Not huge savings, but if you're going to part it out you might as well go cheap. A server CPU on an ASUS motherboard is no better than a desktop CPU. Any solution other than HP/IBM is essentially a cheap one, so there's really no reason to pay for server-grade equipment, because it's not really server-grade. Those are consumer solutions using server parts.

The downtime thing is a legitimate concern, but easily alleviated by adding in a cable line and doing a failover or load-balanced setup. The cost of redundant Internet plus cloud services is generally going to be much cheaper than buying a real server, backing it up, and maintaining it. You need Internet anyway.

You also have to consider server downtime. Even the $17,000 solution is going to go down. It's one server, not a cluster. Sure, it might make it seven years without an outage, but I'd say that's unlikely (you lucked out, if the old one did). The difference between the Internet going down and the server going down is that you can cheaply provide redundancy to the Internet, and it's unlikely to be down for an extended period of time or cost you anything to fix. If that server goes down, it's all but guaranteed you lose at least half a day, and possibly have to pay to have it fixed (at the very least, there's the opportunity cost of you fixing it yourself).

I recommend cloud in any scenario like this because it is cheaper and more reliable than a real server solution. If you had a full infrastructure, ie five or more servers providing a variety of services, it might be very different, because we could cost-justify VMs on a small cluster of host machines. One server is hard to cost-justify without cutting corners either on the quality of the hardware or the backup solution.

We're talking about maybe $400 /month on Azure or similar, maybe less, plus another $100 or so on a second Internet line and a small up-front setup cost and router/firewall setup. For $500/month you get a more reliable, faster solution with better backups that you don't have to manage.

bwillford wrote: »

If we ship to them the cost is less for setup then if they login remotely.

Wow. That's, just, absurd. I would actually charge more for in-person configuration. Unless they are installing the OS from scratch over iLO, there's simply no added cost to them configuring it remotely. If they're just configuring the software over RDP or similar, this doesn't add up

If something like you posted would provide sufficient performance then why not just do that and replace it every year or two? At cost like that we could even have an extra one just sitting around in case anything happens.. The upfront cost of around 4000$ for all the software/license then just transfer license once a year to a new $800-$900 machine?

Is the $1000 backup software really worth it? On most of my servers I normally just run the free version of Cobian backup and back everything up to a usb drive every night, once a week ftp it to a remote location.. I should be able to use that to grab the database file each night and ftp it on the weekends also like I do my file servers right? It also emails me logs of the backup each morning to confirm there wasn't any issues..

Your SQL DB is 5GB. How much does it grow per year ? Or it's always 5GB ? Does your software can run on the free SQL Server Express edition ? SQL Server Express 2012 can support up to 10GB of DB.

In our department, the "standard" for a standalone physical server is to have 5 disks minimum: 2 disks in mirror (RAID-1) for the OS and 3 disks in RAID-5 for the data.

As others said, you should stick with a brandname (like HP, IBM or Dell) server where you can get easier support (e.g. drivers) and replacement parts.

That's an option, but you have to consider the cost of downtime and labor, even if it's your own labor. A solution like that is very likely to go down, even in a rapid replacement cycle. Take it over five years:
Up-front of $5000 for software, hardware, and backup
Replacement cost of $1500/year in hardware and labor

That's $12,500, or $200/mo. Does it really save anything over a $10,000 or $15,000 server solution or $300-$500/month in cloud services plus Internet redundancy? That's assuming no maintenance or downtime costs, which is an unfair assumption. Both will be higher than the server solution and much higher than cloud services.

Heck, this runs acceptably on a seven-year-old server? I mean, you could run it as a VM on a $500 PC. Two cores of a modern CPU are more powerful than a quad-CPU server from 2005, easily. Consumer-grade SSDs easily outperform small RAIDs of seven-year old SCSI drives. Today's software is a little more demanding, but not that much. If performance is really the only reason you're looking at a server, and you don't mind the hassle of having to fix it when it breaks, consumer-grade all the way.

Just don't forget about backup. Actually having a reliable backup system you can test and that works in a pinch can get expensive and messy.

Edit: Can you do a bare-metal restore in under four-hours with your backup system? Or do you have to reinstall the OS, reconfigure everything, then recover your databases? Again, the downtime cost has to be considered. If you don't mind being down longer to save money on a good solution, then free backup solutions are fine. For most small business, however, I've found the cost of downtime makes having a versatile backup solution easily justified.

bwillford wrote: »

On most of my servers I normally just run the free version of Cobian backup and back everything up to a usb drive every night, once a week ftp it to a remote location.. I should be able to use that to grab the database file each night and ftp it on the weekends also like I do my file servers right?

Yes, you can schedule a task in SQL Server to export your DB every night to an USB drive and do what ever you want with it.

What I did in one of our servers is to export our SQL DB to a file server which already have a backup software agent. So I don't need to install an agent on our SQL server.

Realize I am probably thinking/going about this the wrong way but knowing that the $15000 hardware is more about reliability over performance requirements I am really considering the "powerful pc" option.. I could buy two $1000ish machines, get one all configured up and take an image of it and put it into production.. Load the image on the second one and just put it on a shelf.. When the first one goes down I could grab the last backup of the database, throw in on the one on the shelf and keep going.. down time shouldn't be more then an hour? Fix the one that went down by replacing whatever need be and then put that one on the shelf..

I realize that this is probably not the correct way of doing it but what are the main pitfalls of doing it this way?

Also with consumer/soho grade atleast I can pick up replacement parts easily and locally.. I had a 4 year old IBM server whos power supply had failed and we were down 24 hours while waiting to have a replacement overnighted..

If you are honestly okay with having more downtime, more work, and a clunkier solution for the sake of cost savings, go for it. Downtime would still be close to half a business day in that circumstance, but it would work. If it's your money (ie you have an ownership stake in the business) and therefore your extra work is paid for by the savings, it's hard to argue.

If you're just trying to save the company money, I wouldn't do it. You might not be there forever, and the cost of someone else inheriting that solution will be far more than you ever saved with it.

What was the previous server? A quad-CPU monstrosity with six to ten SCSI disks, all circa 2005? A $1000 desktop could easily better similar performance with SSDs.

sratakhin

ptilsen, what's wrong with Dell servers? We have a bunch of HP and Dell servers where I work and none of them have had any problems that would cause an outage.

OP, just virtualize the old server and run it on something like VMware Workstation. It will be easier to back it up too.

Dell is okay, but it's kind of meh. HP quality, reliability, even support has been much better in my experience. I'm not crazy about any of the hardware providers, though. They all have their problems, both in terms of products and support.

Also, keep in mind, I come from an MSP environment where Dell is widely known to compete with its partners for their customers, so I have some bias there.

EV42TMAN

Dell HP and IBM are all rip offs if your want a real server that isn't going to break the bank go to Nor-Tech | Custom-built PCs and Services for IT Solutions Providers they have prebuilt ready to ship servers around $1000+OS options and can customize it for what you need.

sratakhin

A real server is going to provide 99.999% uptime. I don't think that a $1000 server will have redundant PSUs, NICs and hardware RAID controllers, unless it's a refurbished or used unit.

...or go to Newegg and build it that way.

I stand by what I said. Anything outside of a enterprise-grade server is essentially a piece of crap. It might perform similarly (hint: it won't if you do RAID 5), but the experience working with it is very different and it will not be as reliable. Maybe Nor-Tech or Supermicro or whatever is better than Asus which is better than Biostar. At the end of they day, they're all very different from using an HP, Dell, or IBM.

Again, that's not to say it's never justified. For many businesses, the costs of using HP can ultimately be lower than using a budget solution when it comes to downtime, maintenance, and implementation. Others can get by quite nicely with cheaper solutions. Google, for instance, doesn't use off-the-shelf enterprise-grade hardware. Google, however, has fault tolerance up the wazoo and can afford to be replacing crappy systems all the time.

Edit:

sratakhin wrote: »

A real server is going to provide 99.999% uptime. I don't think that a $1000 server will have redundant PSUs, NICs and hardware RAID controllers, unless it's a refurbished or used unit.

I agree in principle, but I don't agree with the number. Virtually no physical server will provide 99.999% uptime. It has components that cannot be made redundant, and the combined reliability of those components is less than five nines. That is a big part of why I push for cloud at this level. The cost of implementing true server redundancy for a single system, even using extremely cheap components, is very high due to the complexity of the setup alone.

The difference here is that a good server might have, say, 3-and-a-half nines, while the cheapo server probably provides two or so. If you're compromising to save money but still want to limit downtime, the three-to-four nines range is a much better spot to target, in my opinion. The cost of getting from four to give nines is generally higher than getting from two to three, which makes three to four a better spot.

The reason why I think OP is seriously considering consumer-grade is because he is willing to tolerate the low reliability. It is easier/cheaper for him to just fix/replace the system more often than to buy reliable equipment.

You got what you paid for...

Like cars, you can buy a Lada, Ford, Toyota, or BMW. Each can get you from point A to point B. But which one will break first ? Which one is more fun to drive ? Which one will get you more chicks ?

sratakhin

Lada will beat anything hands down

You'll get noticed immediately. I'm sure everyone will ask "wtf is that?"