Slashdot Mirror
Best IT-infrastructure For a Small Company?
DiniZuli writes "I've been employed by a small NGO to remake their entire IT-infrastructure from scratch. It's a small company with 20 employees.
I would like to ask the /.-crowd what worked out best for you and why? I came up with a small list:
Are there any must have books on building the IT infrastructure?
New desktops: should it be laptops (with dockingstations), regular desktop machines or thin clients? A special brand?
Servers: We need a server for authentication and user management. We also need an internal media server (we have thousands of big image and video files, and the archive grows bigger every year). Finally we would like to have our web server in house. Which hardware is good? Which setup, software and OS'es have worked the best for you?
Since we are remaking everything, this list is not exhaustive, so feel free to comment on anything important not on the list."
Media server? How about S3. Web server? How about EC2. Seriously, why spend time and $ on procuring, powering, cooling, backing up, and upgrading all that gear? Give everyone a laptop and a gmail account. Put the rest in a public cloud.
Yup, agreed. You could have your webserver in-house. You'll need a safe room to lock it away in, ideally with some aircon, maybe a halon fire suppression system. Plus an UPS, obviously. And you'll probably want to hire another cupboard, with the same systems, a few hundred miles away, for an off-site backup. Oh, and make sure your ISP provides you with a sufficiently fast uplink.
Alternatively, pay someone $50-$500 dollars a year for the same. It's a no-brainer unless you've got some really, really pressing reason.
Please consider this account deleted, I just can't be bothered with the spam anymore.
Keep the whole thing simple, the next person who comes in will thank you for it. Don't introduce any weird convoluted things into the system and make sure to make it so that the whole system is modular, easily upgradeable, and when the time comes and they need to expand that it's expansion friendly.
First off is keep it simple. The simpler the better. This is not an enterprise, they don't have a lot of people to call on for support. So don't build anything complex.
I probably wouldn't bother with central authentication unless there's a reason, just do it per computer. Ask yourself what it gains you to have. If the answer is just "simpler administration" then don't use it. 20 computers is not a problem to administer without it, particularly since not everyone logs in to all computers. However the central servers are a point of failure, a place for problems.
Also have someone else host all your servers unless a file server is needed. There are plenty of good server hosts out there. For the web, depends on what you want. Pair is a top notch web host I used for many years. Top flight quality in every regard. Hostgator is who I use now to save some money and I'm perfectly satisfied. It works well, is reasonably fast, and they don't bitch that I do like 100GB of traffic a month.
For an internal file server, something simple and reliable. A computer with RAID-5 or RAID-10. Make sure to do offsite backups. An easy option for that is Acronis Trueimage. Great backup program and they will do network backups for a fee. It can encrypt the backup so no security issues. If their service is too expensive, use the software to backup to external HDDs and lock them in a safe or something.
Thin clients: You must be joking. Don't do thin clients unless you understand it well and are willing out lay out a lot of cash to make it reliable. Remember that if a desktop crashes, gets corrupted, whatever one person can't work. If the tin client server goes down EVERYONE can't work. There are some situation where they make sense. If you aren't experienced enough to know when don't use them (yours isn't one BTW).
As for computers, get something from a major supplier. Dell or Lenovo are my recommendations. They don't have an in house IT department they can't really be faffing about with repairs. Get them from someone that'll do onsite service and get a nice long warranty (unless you are sure they'll be replaced sooner). Make sure that there is a company out there that backs up the hardware that people can just call to have shit fixed.
Desktops vs laptops depends on the usage. If the intent is that these are used in the office, then desktops. They are cheaper to purchase, cheaper to find repairs for out of warranty, and harder for someone to walk off with. Don't get a laptop unless there's a real need to get a laptop. If people are going to be walking around with them for work reasons then fine, though it still might be good to have a desktops as well in case they forget their laptops at home or lose them or something.
For OSes, depends on your needs. I'd say Windows unless you have a reason not to. Yes, yes I know MS evil and MS tax and all that jazz. Forget all that. These computers are tools to get a job done, the users don't care past that. Get them the best tools for the job. That will probably mean Windows for running MS Office, and for working with media since Linux tends to fall down in that department. Only do Linux if you are sure it will meet their needs (and by sure I mean you've tested it) and they can get the support they need.
In general I'd stay away from Macs. They cost more per unit, and they are not good with business support. Their idea of support is generally "Take the system to a store, we'll look at it and get it back to you." Fine for a consumer, not for a business. For a business you want "I call you and a tech shows up tomorrow with all the parts to fix it." Only go with Macs if you have a real reason and if you can't think of one, then you don't have one.
Remember to keep pragmatism in mind above all else. Get people the tools that do the job they need. That is all computers are to non computer people is tools. You are just being asked about expensive hammers or saws or the like. Your job is to figure out what they need, what will do the job the best, what can be th
Ask Slashdot: Why do your job when you can ask others to do it for you?
Why indeed?What reasonable motivation could he have to poll a well-established base of computer experts for advise? Could it be that an infrastructure is a hard thing to get perfectly right? Maybe up-front decisions made right will negate hours of work and wasted productivity down the line? Remember those security and infrastructure failings we've been so critical about all these years? Those clueless IT guys who screwed up royally and condemned employees and management to countless hardships? Maybe he doesn't want to end up in that position... maybe he wants to do things right?
That lazy bastard!
I get the whole "he should do his job thing," but I'd agrue that he is. His job is to improve/develop that company's infrastructure. It doesn't matter that he doesn't have ALL the knowledge in his brain to do this from scratch. He's researching using the tools he has avaiable and one of those tools is the knowledge base at slashdot. Except unfortunately it seems everytime someone asks the slashdot "community" for help with anything. They immediately get thrown under the bus for asking the question in the first place. So much for the "community" and helping colleagues in the field.
Great idea, except:
1) S3 performance is poor. You've got to pay a LOT for performance.
2) Non-hardware (administration) costs are still going to be the same.
3) Cloud services are dependent upon connectivity. Which do you trust more: no link failure in thousands of miles of cables, fiber, and networking equipment, -or- the volatility of your local network and attached storage systems? You will need at least 2Mbit of low-latency throughput for something like this.
4) You will need redundant outside-network links. This may not even be possible in his locale, and if it is, there's no guarantee something upstream won't die (and in many places, the certainty of something failing upstream is fairly high due to shared carrier).
5) Are connections of sufficient throughput and latency even locally available? There's no mention of things like: mail use, type of work performed, etc. What if they do CAD work? What if they do a lot of email with attached documents? Graphic or sound work? These are use cases which are horrible for cloud computing.
That's just a starter list. It's suitable for some purposes, but for most day-in and day-out stuff, it is not good as a primary source of IT infrastructure.
For general purpose daily cloud computing, S3 isn't even a good/best option.
As for the OP... this guy should obviously not be in IT. The most notable thing missing from his list is: competent and experienced IT personnel. Obviously this was not considered as a requirement by those paying the bills, but it is important.
Hint: use requirements are the first thing to consider. Everything is based off of that. The vendors picked depend on experience and available purchase agreements. What I do for 90% of my customers will likely be a poor fit for many of your customers. And so on.
Fucking amateurs. They make us MSPs look bad.
~/ssh slashdot.org ssh: connect to host slashdot.org port 22: too many beers
Baloney. Use SME Server or Zentyal. I run a nearly identical organisation and my headaches have been significantly reduced since we stopped relying on Windows servers.
And to all those who derided the OP for asking others to do his job for him: This is why you ask others' opinions: because sometimes what you think you know isn't always true.
Crumb's Corollary: Never bring a knife to a bun fight.
OK, seriously, I've done a couple dozen of these 10 to 50 user installations. Half the time is spent at the beginning to determine what the customer needs and wants, and what the budgeting will be. Things invariably cost a lot more than the customer anticipated so your goal is to manage expectations. If you don't do that, your life will either become a living hell (if you will be providing long-term support) or you will leave behind an unhappy customer.
Some of the basic things that were not considered when customers brought me on:
Are there remote employees? Will they need VPN access? What platforms are they using to connect? Can you verify that the endpoints are secure?
What is the anticipated volume of mail? In this day, it's often much cheaper to outsource to Google for smaller installations, but in some cases it makes a lot of sense to keep in-house.
When hosting your own web server how much downtime is acceptable? Do you need 24/7 uptime or will you have maintenance windows? What if your primary site burns to the ground? Do you have the floor space and adequate cooling? How much traffic is anticipated at the beginning of the project? How much do you expect to grow?
What applications do you need in-house? Accounting packages? Company intranet? Database? How will you separate your LAN for security purposes? Do you take credit cards as part of business?
What infrastructure applications do you need? Can you afford downtime on these? How many ports/switches do you need? Wireless? Separate backup LAN? OOB management for your servers?
Before you even start pricing hardware, find out what your customer needs and wants and willing to pay for.
This is a public forum, we're all volunteers here.
Personally, I'm okay with the occasional "Help me with best practices" post. I wouldn't want to read that stuff all the time, but it adds to the mix, when taken in small measures. Keeps me in touch with developments outside my immediate interests. Sometimes generates lively debate. Maybe helps other readers in the process, benefits the general welfare.
If you want to blame anyone, blame Slashdot editors for publishing this kind of thing.
-kgj
You make a great point. If I am hiring someone to achieve a goal for me, the absolute last thing I want them to do is research the possibilities and find out what experiences and approaches others have taken in the past. I want someone like the people posting in this sub-thread. I want the kind of person who knows that research and due diligence are a complete waste of time. I mean what is there to know? Just do it, and worry about what "it" is, and whether the approach was a good idea later, after you've done the first 90% and it is time to do the other 90%.
Guns don't kill people; Physics kills people! - John Lithgow as Dick Solomon on Third Rock From The Sun