starting comp biz, in charge of hardware purchases, advice?

bignasty said:
ok my .02 do you guys realize saying that home built machines arent good for servers is an all out lie??? look at the servers that hardocp and the hardforums sit on, and has anyone checked out rackspace...

Clearly you don't work in a corporate environment. In this type of environment, brand name will ALWAYS win. What happens 3 years from now, if jaqie's biz goes under, and the company needs parts desperately for their server that's totally down? These are things IT managers think of when even considering buying off-brand.
 
i think you've got a workable model in place.

It's the minute details that might be needing some work through and we'll try to be as much help here as possible.

here's what I think (and take it as you will):

Thinking long-term (and somewhat pessimistically): what does happen if your company goes under? liquidating your assests will be mandatory especially if you have to pay off your corporate loans. In that case, the ability of brand-name hardware (HP/compaq, Cisco, etc.) to maintain relatively high resale value will be more of an asset even though there may be a "steeper" buy-in.

cheers,

dave
 
Exactelly. Brand names sell. You will not get a single client to put their site at risk by putting their website on a home built machine.

I dont have anything against home built machines, in fact that is what I am typing on as we speak. What it comes down to is business sense. If the server goes down and your website goes down you could be loosing alot of money every hour. Who would you want working on it, the colective brain power of HP/Compaq, dell or a few people from the hosting company. I can tell you what I would want as a business owner.

I have and still work in a corporate IT dept, in the room next to me we have over 300 servers. Everyone of them is a brand name Compaq. In fact their are regulations and rules ragarding what type of servers (IE, they have to be from a company that can provide 24/7 HW and SW support and cannot be peiced together from off the shelf parts, it reads something like that).

There is nothing wrong with home built servers, but I would never risk my business on one.
 
The first thing that comes to mind is why have 3 different subnets? I agree that the gamers will suck up some bandwidth, but I don't see the thin clients and WiFi hotspot sucking up so much that they deserve to have their own subnets.

Secondly, about the switches. Granted, I know you need to watch your cash flow, but there is a REASON why these switches are designed like they are. You can manage your bandwidth utilization SO much better with one of these than a switch you can pick up at CompUSA that it's not even funny. Take a class in Cisco if you don't know how to set these up, but I would seriously consider them when you have enough capital to afford them.

Also, I don't understand why you have the T1 coming in on a PCI card, yet everything else coming in PCI-X. It's reasonable to assume that most of your bandwidth traffic is going right out of the building, so why have that possible bottleneck? I know you are using gigabit, but the outside traffic is still coming in on a 33MHz bus port.

The router. I have accepted the fact that you are dead set on FreeBSD, but the whole point of the router is to make routing rules, policies, and so forth. If you have a failover to some 40 dollar router, do you really think you will have that functionality?

What did you come up with as far as a UPS goes? Perhaps something like this: http://www.cdw.com/shop/products/default.aspx?EDC=073732
 
I'm just putting my .02 in quickly.

The marine battery thing that you were thinking about I have seen. The unit cost more then 500grand easily but not sure on that but if it can keep a full flight simulator up and running that has this as the hardware so yeah i need this for a whole house.

12 cabinets with 7 power supply pulling 300amp each( some 12volts DC output, 24volts DC output) (visual computer unit) computer
9 sun computers (host computers)
3 full fledge SEOS projectors (produces the 180 view)
countless I/O buckets (for switches and displays)
Electric control loading units.

The UPS unit if I remember right could keep this up for 45 minutes with everything running but I think it had 4 trays of batteries with 8 batteries per tray. I wish I could remember who made the dam thing but it wasn't APC. The batteries we could change ourselfs and we had to take voltage reading every six months on them and the unit was running at 55% capacity for load.

Thought I would add that the thing is real but not sure if there is a small unit size for these things.

Good luck on the business
 
jaqie said:
I agree about the webhost, scratch that idea. I also talked with a friend and decided that if I have anything to say about it we will be contracting him out for setup(freeBSD part I can do the rest) and consulting. He and I have already talked at length about it and thus the setup I have in mind has changed radically. Ill be editing the original post in a bit to reflect the new setup.
jaqie said:
after more consideration I am thinking of this UPS solution.

APC Smart-UPS XL 3000VA RM 3U 120V
Part Number: SU3000RMXL3U
Estimated Resale Price*: $1,425.00

APC Smart-UPS 48V RM 3U External Battery Pack
Part Number: SU48R3XLBP
Estimated Resale Price*: $679.00

should provide at least 15 minutes full cabinet runtime, and a lot less expensive then 10k.
I want to config the game servers to shut down immediately upon a switch to battery, that should extend runtime quite a bit.
information was already in the thread, guys. ;)
thanks a lot for the advice. and oh yeah you misread it about how Im configuring the T1 coming in. it IS on a PCI-X card.
 
jaqie said:
information was already in the thread, guys. ;)
thanks a lot for the advice. and oh yeah you misread it about how Im configuring the T1 coming in. it IS on a PCI-X card.

OK. Not how you made it sound originally, but as long as I'm clear on it.
 
btw, just to follow-up a bit more:

there's no real disadvantage of going the Newisys/Celestica route either. I do pre-configured servers using their barebones stuff all the time and frankly, it's rather damn good....and not too expensive to boot...

beyond that, they're normally validated for all the Opteron families within weeks of release, versus, the months it takes for HP, et al. to go through the process.

cheers,

dave
 
EEPS their (celestica) smallest server seems to be a dual opteron! way way way way more then Im planning on spending... Im going with socket A solutions because they are very very inexpensive...
the same 'problem' with newisys. ouch!
 
i do realise that (heck, i sell 'em...of course i know that) but given the applications you were referring to, i thought it might be a "best fit."

cheers,

dave
 
dave_graham said:
btw, just to follow-up a bit more:

there's no real disadvantage of going the Newisys/Celestica route either. I do pre-configured servers using their barebones stuff all the time and frankly, it's rather damn good....and not too expensive to boot...

beyond that, they're normally validated for all the Opteron families within weeks of release, versus, the months it takes for HP, et al. to go through the process.

cheers,

dave

Personally, if getting something mission critical, I would wait the extra few weeks to ensure that all the bugs are out of the chipset and processor before I would lay down 10K or whatever for a Opteron server. That's just me though.
 
dave_graham said:
i second the opinion regarding the Cisco PIX and Catalyst switches...With lower grade hardware, you always run the risk of problems.

If you need Cisco hardware, i'm more than happy to help....remember also, Cisco does not transfer warraties on software if you purchase the routers/firewalls second hand. you have to buy it new in order to get the full warranty. Also, their support packs are top-notch and would be worth purchasing along with the router.

cheers,

dave

Can a brother get a 515 PIX for cheap? ;)
 
Cheech said:
The first thing that comes to mind is why have 3 different subnets? I agree that the gamers will suck up some bandwidth, but I don't see the thin clients and WiFi hotspot sucking up so much that they deserve to have their own subnets.
Its not for bandwidth, its for security.
Secondly, about the switches. Granted, I know you need to watch your cash flow, but there is a REASON why these switches are designed like they are. You can manage your bandwidth utilization SO much better with one of these than a switch you can pick up at CompUSA that it's not even funny. Take a class in Cisco if you don't know how to set these up, but I would seriously consider them when you have enough capital to afford them.
you can actually do that with some of the better D link and netgear switches. I am considering running a layer 2 switch for these but frankly I dont see it as much of a problem... I mean when the bandwidth will really be used is when there is a LAN party... and Ill be limiting the segment they are on via the router anyways...
Also, I don't understand why you have the T1 coming in on a PCI card, yet everything else coming in PCI-X. It's reasonable to assume that most of your bandwidth traffic is going right out of the building, so why have that possible bottleneck? I know you are using gigabit, but the outside traffic is still coming in on a 33MHz bus port.
you misread me. it's actually coming in on one of the dual PCI-X NICs.
The router. I have accepted the fact that you are dead set on FreeBSD, but the whole point of the router is to make routing rules, policies, and so forth. If you have a failover to some 40 dollar router, do you really think you will have that functionality?
no, I know I wont. however, the features in my current DI-614+ are quite impressive for a low dollar SOHO use. site filtering, and the like. I plan on using them as a temp thing only as I put the server back together when it fails. few days wont hurt.
What did you come up with as far as a UPS goes? Perhaps something like this: http://www.cdw.com/shop/products/default.aspx?EDC=073732
ick. nothing but APC. no way im going with anything else. I was thinking (and this was earlier in the thread) of this:
APC Smart-UPS XL 3000VA RM 3U 120V
Part Number: SU3000RMXL3U
Estimated Resale Price*: $1,425.00

APC Smart-UPS 48V RM 3U External Battery Pack
Part Number: SU48R3XLBP
Estimated Resale Price*: $679.00
 
I would also recommend using Dell or HP for your servers. I would definately recommend trying to resell as opposed to selling your own whitebox brand.

I also think you should get a better ide raid card for your router. You said you're going to be configuring the freebsd router, right? I just want to make sure that you're the one doing it instead of having someone else come in. That way, if there's a problem, you can take care of it yourself. I also would thinking building another lower spec pc to use as the backup router with a identical freebsd config makes more sense than using a crap router as your backup. I personally switched to m0n0wall which is freebsd based but runs off of a cdrom and a floppy. I have had bad experiences with the hard drives in pc's I've used as firewalls/routers in the past. I've also moved to using a embedded pc from soekris and that's worked well.

Once you get this going, definately show us some pics and give us a update. This is a very interesting project.
 
rosco said:
I would also recommend using Dell or HP for your servers. I would definately recommend trying to resell as opposed to selling your own whitebox brand.

I also think you should get a better ide raid card for your router. You said you're going to be configuring the freebsd router, right? I just want to make sure that you're the one doing it instead of having someone else come in. That way, if there's a problem, you can take care of it yourself. I also would thinking building another lower spec pc to use as the backup router with a identical freebsd config makes more sense than using a crap router as your backup. I personally switched to m0n0wall which is freebsd based but runs off of a cdrom and a floppy. I have had bad experiences with the hard drives in pc's I've used as firewalls/routers in the past. I've also moved to using a embedded pc from soekris and that's worked well.

Once you get this going, definately show us some pics and give us a update. This is a very interested project.

Agreed on many points. Outsourcing at this point is a REALLY BAD idea. I cannot stress to you enough how much of a REALLY BAD idea this is. I currently work for a consulting firm that does outsourcing for larger companies, and I am an IT manager for other small businesses on the side. I can tell you with absolute certainty if I leave some of my clients, they are going to have a hell of a time training another IT manager. You will have the same issue, and especially since he BUILDING YOUR CORE INFRASTRUCTURE, I would want to be there every step of the way. Now, don't get me wrong, I am all for a small business to outsouce their IT needs, especially if they have someone grossly underqualified trying to manage them, but in your case you ARE qualified, and the biz hasn't even started yet! Hiring someone, ESPECIALLY if it's a W-2, costs TONS on the backend (read: your bottom line) than what shows up on his pay stub. Bottom line before I beat this too much, you are a technology company, and you seriously need to watch your cash flow, don't blow it on this. Read some man pages, do some homework, pull some all-nighters, and eat a fuckload of pizza and figure that shit out :)

Second point was exactly what I was thinking for the backup router too. Take all the surplus parts you are going to have laying around, whip together a PC, and use that for the backup. Don't sell yourself short and use a 40 dollar Best Buy router.

Yes, we most definitely need some updates. I am very intrigued at the progression of this little project of yours.
 
its not quite outsourcing AND its not going to hurt our bottom line because it will be a cash thing, not the kinda thing that goes on the record. We will pay him for it out of our own pockets... and we plan on learning FreeBSD very well ourselves to boot. Sorry I have been gone for a bit but my system went ape. vidcard problems, and DVDRW problems. I have it sorted now thank goodness. had to update the bios/firmware in both to solve them.

Yes I plan on keeping everyone well informed on the goingson, that's for sure.

And again, thanks for all the help!

*sigh* now to reinstall all my programs...
 
jaqie said:
its not quite outsourcing AND its not going to hurt our bottom line because it will be a cash thing, not the kinda thing that goes on the record. We will pay him for it out of our own pockets... and we plan on learning FreeBSD very well ourselves to boot. Sorry I have been gone for a bit but my system went ape. vidcard problems, and DVDRW problems. I have it sorted now thank goodness. had to update the bios/firmware in both to solve them.

Yes I plan on keeping everyone well informed on the goingson, that's for sure.

And again, thanks for all the help!

*sigh* now to reinstall all my programs...

Not going to hurt your bottom line because it's a cash thing? It's still an expense, and expenses hurt your bottom line. The reality is you better make this guy real happy because I can almost guarantee you if he leaves there will be stuff that he did that you will have no idea to do, thus forcing you to hire another consultant (probably at a higher rate) to do the work.
 
Him and I have been friends for many years and have been through a LOT of crap together. He knows my darkest secret among many man y other things, and we have done business deals (though not of this magnitude) in the past. He is very laid back and we get along swimmingly.

Honestly. I look at it like this:
Business cash flow
Home cash flow

I am going to pay him out of our OTHER income, our home cash flow. the business, for the first year at least, will only feed itself, we wont take any out for us. So, no, I dont think it will hurt our business' bottom line.
 
Back
Top