Web Hosting Blog :: Ubiquity Hosting Solutions

As a sales rep, I regularly get asked by clients for a cost estimate of moving their in house datacenter into a real datacenter like one of ours.  Sometimes we can save a customer a lot of money with the move, sometimes the break even, and sometimes they spend a little more.  Ultimately, anyone that moves usually ends up with a better IT solution for their dollar, even if they’re spending a little more to do it.  Here’s a few bullet points to consider:

1) Power - You’re likely paying for power at your office and with average nationwide power pricing, you can expect to spend $35 per server per month just to keep them powered on 24/7/365.  With a server farm of 10 servers, you’re looking at about $4,200/year on your power bill.  It’s something a lot of offices never think to consider, but if you don’t really like that $4,200, you can PayPal it to me.  Oh? Now you like it? That’s what I thought.

2) Redundancy – Is your office prepared to handle a power outage?  Can you operate with your servers down?  Do you have redundant air conditioning?  If you’re in a pretty standard office complex, you probably don’t have favorable answers to any of those questions.  Datacenters are traditionally placed with access to two or more power grids.  Should power fail, all of our datacenters are on a battery backup.  While the battery is going and if power is not immediately restored, a generator kicks on and several days of fuel is already on site (plus fuel companies are on contract to refill the generator tanks before the fuel runs out.)  Your office definitely doesn’t have that.  Does your office air conditioning go out every once in a while?  If it does, you risk having to shut down your servers or power them down to protect them from heat damage.  Our datacenters all have redundant cooling – and no just so we’re up if one fails, but so we’re up if one needs to be repaired.

3) Connectivity - This touches a bit on redundancy, but do you have two connections to your office from two different bandwidth providers?  How often does your business grade connection go down?  If you’re anything like a typical office, it’s a few times a year that you’ll lose internet connectivity.  If that takes down your website, your phone system, your email system, etc. it’s costing you business, productivity, or both.  How much is that internet connection costing to the office?  I had a client paying $600/month for business grade DSL and business grade cable internet (for redundancy.)  When he moved in with us, he had full redundancy from 4 providers instead of 2 and he was paying less than $100/month.

There are a whole host of other reasons to outsource your datacenter needs to a real datacenter, such as support (someone can watch your servers 24/7/365 for you cheaper than you can hire someone to do it), capital reduction (we can rent you servers cheaper than you can buy them), and expandability (really? you’re going to put another rack in that already hot server closet at the office?)

So next time you’re looking at your IT budget, consider all of the hidden costs of hosting your services in house.  Even if it’s a break even move, get out of there so you save yourself a headache!

With some places in the United States regularly well over 100 degrees Fahrenheit, it’s a remind how how incredibly hot it can be outside this time of year.  Your servers are in datacenters with high end, industrial grade cooling systems designed to keep them much closer to (or right at) 68 degrees Fahrenheit.

Why so cool?  Well, it’s not because we love the electrical bill.  First and foremost, keeping the datacenter at a reasonable temperature means that servers don’t overheat.  By why *that* cool?  Well, that’s about the optimal temperature for not only keeping the server from overheating, but keeping it at full performance and preserving the life of the hardware.  If we were to keep the datacenter in the 80s or 90s, the hardware would probably still run, but you’d experience a lot more hardware failure.

When it comes to performance and reliability, cooler is always better.

The Nobis Technology Group, parent company of Ubiquity Server Solutions (www.ubiquityservers.com) and Ubiquity Hosting Solutions (www.ubiquityhosting.com) has officially moved its headquarters from Bloomington, Illinois to Phoenix, Arizona.  This move should have been be completely transparent to customers, although will immediately result in a major reduction of overhead costs, allowing all of our brands to be more competitive.  So ultimately this will result in a savings that is passed on to our loyal customers!

In 2006, concurrent with a merger with DarkStar Communications to form the Nobis Technology Group, Ubiquity moved into its first office space in Bloomington which was an old highschool that had been converted into office space.  Despite expanding into an adjacent office suite in the building, Ubiquity’s growth outpaced the ability to expand on site and Ubiquity moved across town to a larger, more upscale office.  Ubiquity has operated out of that office space since 2007 an has experienced tremendous growth both in staff and overall size.  We’ve added a POP, doubled (or maybe tripled?) our staff, and we’ve taken a strong focus on improving our product.

Since that time, Ubiquity has realized the importance of relocating to a city that can better handle its growth, something that Bloomington has been unable to do.  The move to Phoenix is ideal for a host of reasons, but predominately a reduction of costs in overhead.  We’ll keep you updated, but I think now it’s time to go jump in a pool and celebrate the move!

1) Our network. We run our entire network over Brocade XMR-4000 routers, not a Cisco Switch that could just as easily be a rack level switch like a lot of our competitors do.

2) Our service. Our support team is here 24/7/365.  We can be reached by phone, livechat, or email any time of the day or year and you’ll always get a timely response.

3) We care. You’re not just another number with us, while we’re certainly not small, we’re still not too big – we establish relationships with our customers and for this reason most of our customers stick around for a long time.

4) We back up what we say. We offer a 100% uptime SLA on power and network as well as a 1 hour hardware replacement SLA.  If we can’t meet that, you get a credit on your account.

5) Our locations. Six locations mean your project can expand across the entire US.  It may not apply directly to you on just one server, but we do offer services in six cities across the US, so if you need more than one server, there’s no need to try new vendors – you can go with us for almost, if not all, of your USA server needs.  Even if it’s not relevant now, it’s useful for future expansion.

Choosing where to put your servers is often underestimated.  Ubiquity offers services out of six geographic locations, covering the continental United States.  We offer services out of Atlanta, Chicago, Dallas, Los Angeles, New York City, and Seattle.

Why do you need to consider this? Well, ultimately data has to travel from either yourself or your users to the server and back to you or your users again.  Even though, for the most part, it’s traveling at the speed of light, there is a noticable difference between when data travels 2000 miles and when it travels 20 miles. It’s particularly noticeable on with streaming content and less noticable with static pages that are mostly text, but it’s never, never, never bad to reduce latency to your server.

When considering where to put your server, put it as close as possible to the population center of your users.  If you’re a Chicago Cubs fan stranded in Dallas, it’s probably wise to put your Chicago Cubs fan site in our Chicago data center instead of our Dallas datacenter.  It’s a bit of an extreme illustration, but hopefully it illustrates what we’re going for here – you want to get your server as close to your users as possible and sometimes that means further from you.

If your users are everywhere, the general wisdom is to get as close to major population centers as possible.  Lets take the example of a site that is visited predominately by Americans.  You’ll obviously want to pick something close to the center of the United States so you can reach all of your users, but which center? Dallas or Chicago?  Consider the three largest cities in the United States, New York City, Chicago, and Los Angeles.  Placing your server in Chicago puts it close two two of those, so Chicago might be wise.

All in all though, for most web hosting content, anywhere within a few thousand miles is pretty darn good.  With game servers and voice servers, a few hundred miles is much, much better.

While some colocation is fairly static and remains the same year after year, most colocation involves careful planning of future growth.  There are a lot of things to consider when planning this growth.  Here are some things to think through.

1)  Realistic Growth Speed. I can’t count how many startups come to us asking about a full cabinet for their 1u server “so we have room to grow.”  Believe it or not, sometimes they fill in the cabinet, sometimes they go bankrupt.  The important point here is that it’s very easy to overbuy colocation so that you can grow into it.  Sure, colocating 42 devices is cheaper per device than 5 devices are per device.  But is it cheaper to have a little bit of downtime later on (to move your equipment) or is it cheaper to swallow the larger monthly recurring cost to allow yourself room.  There isn’t a right answer for everyone, but there is a right answer for you.

2) Know when to rent. Keep in mind that if you have to ship servers to and from the datacenter regularly, it may make more sense to rent them in the first place.  Shipping adds up and it’s just money out the window.  Sometimes customers try to get the most out of 1u by slowly upgrading it from a low end server to a medium server, to a high end server, all the while the shipping costs are adding up.  Maybe it makes sense for you, but don’t forget to factor that in.

3) Give yourself time. A huge mistake we see nearly every day (seriously!) with colocation is waiting until the last minute to pull the trigger on upgrades.  Not all upgrades are instant. Seriously.  If you need a cross connect, it can take up to two months for all of the necessary parties (telcos are notorious for slow times like this) to pass off all the paperwork and complete all the labor.  Plan well ahead for major upgrades and at least know the lead times.  Sorry, we simply can’t deploy 100 cabinets overnight.

Those three tips should set you well on your way.  Mistakes in planning colocation can be very costly, so definitely plan the project ahead of time.

Sometimes it doesn’t make sense to rent the server or servers that you need to operate your IT infrastructure.  When this is the case, what customers frequently need is called “colocation” or “collocation.”  Colocation is essentially the concept of renting datacenter resources and providing your own hardware.  The typical colocation project involves renting space, power, and bandwidth.  While sometimes other things are necessary (private circuits, ramped services, etc.) we’ll just cover a basic colocation experience in the interest of answering the great majority of questions about the process.

Step 1: Purchasing

The first step is to contact our sales department and let us know what you need.  We’ll send you an official quote.  If the quote looks good, we’ll go ahead and begin the process.  If the quote doesn’t look good to you, let us know what we need to tweak to meet your needs and we’ll provide another quote.

Step 2: Setting up Invoicing

The next step is to set up billing for the service.  We’ll take down your contact information, generate an invoice in our billing center, and provide login information so you can sign on and pay it.  If your colocation involved a contract longer than month to month or was over $500/month, we’ll also typically have a written, signed contract by both parties – don’t worry, it’s a very painless process.  We generally skip this formality on very low dollar colocation arrangements and month-to-month colocation arrangements, but if you’d like the MSA (Master Service Agreement) on a smaller colocation agreement, we’re more than happy to sign one.  Once the invoice is paid, the lead time for the service begins.  Leads times are typically brought up in step 1, but can range from 3 to 30 days on most projects depending on size and complexity.

Step 3: Equipment setup

Once your invoice is paid, Ubiquity will contact you with IPs and other provisioning information.  If you’re going to ship your server(s) to us, just let us know the tracking number(s) as soon as it’s available.  If you’re going to bring your equipment in in person, contact us letting us know what time you’d like to bring it by. Once you’ve brought in your server(s), we start the billing cycle from that day.

If you have any questions, please let us know!

As your site grows, your hardware might need to as well.  In fact, with any sizable growth, upgrading the hardware that is dishing out content to your users is inevitable.  Over the next few weeks we’ll be looking at some general strategy in upgrading hardware so you get the best bang for your buck.  After all, not in your interest or our interest to waste time, money, and resources on upgrades that don’t make sense.  If this is your first Linux Dedicated Server, you might want to check out last week’s post about renting your first dedicated server.

Choosing and upgrading a bandwidth plan

By default, our plans come with 6TB of data transfer on a  100mbit port.  There are a whole host of changes you can make to this plan.  You can upgrade the port speed ($35/month).  This essentially allows you to do a lot more bursting, but doesn’t let you move more traffic over the course of the month.  In other worse, it lets you “spend” your traffic allotment faster, but doesn’t make it bigger.

Our 12 TB plan is a $70 upgrade, but allows you twice the bandwidth.  It lets you transfer more bits and bytes over the course of the month.

Alternatively, you can get a dedicated port.  The nice thing about dedicated ports is that you cannot get an overage.  In other words, if you have a 6TB plan (6000GB) and you use 6001GB, you’ll be responsible for a $1 overage.  It doesn’t sound like much, but it can add up fast.  To ensure you don’t have overage bills, get a dedicated port.  This is frequently called “dedicated unmetered.”  It’s essentially saying that you can use the full 10 mbps, 100mbps, or 1000mbps (or custom port size) without any chance of overage.  Changing the port speed greatly changes the price.

If you have any questions, let us know at sales (at) ubiquityhosting (dot) com!

As your site grows, your hardware might need to as well.  In fact, with any sizable growth, upgrading the hardware that is dishing out content to your users is inevitable.  Over the next few weeks we’ll be looking at some general strategy in upgrading hardware so you get the best bang for your buck.  After all, not in your interest or our interest to waste time, money, and resources on upgrades that don’t make sense.  If this is your first Linux Dedicated Server, you might want to check out last week’s post about renting your first dedicated server.

Picking out and upgrading your HDD configuration

Since hard drive (often shorted to HDD for Hard Disk Drive) size is pretty straight forward, we’ll simply cover the other upgrades that are possible.  Any of the upgrades below are possible in varying sizes, although the that variation itself may be limited.

RAID – There are three main types of RAID used in Web Servers.  There are a few options beyond these, but here are the best options for web hosting, particularly when you factor in cost.

1. RAID0 – In this RAID array, every other bit is written to a different disk, thus increasing speed by the number of disks in the array.  If anyone one disk fails, all of your data is lost.  Most people do not put any more than two disks in a RAID0 array because the more disks you have, the more likely it is to fail.  With two drives, you’ll have approximately the combined total storage space of both drives.
2. RAID1 – In this RAID array, the disks are mirrored in realtime.  It is only done with two disks.  With two drives, you’ll only have the total storage space of the smaller drive.
3. RAID10 – This has the benefits of the speed of RAID0 and the redundancy of RAID1.  Actually, it’s implemented (more or less) as a RAID array of RAID arrays.  With four drives, you’ll have the storage space of 2x the smallest drive.

HDD Speed – Most Ubiquity servers come with 7200RPM drives.  These can be upgraded to 10k RPM drives or 15k RPM drives.  As you may have guessed, if RAID0 or RAID10 isn’t enough speed, you can put these in your server to really kick it up a notch.  These involve custom quotes, so contact us at sales (at) ubiquityservers (dot) com for more information about upgrading to a higher RPM hard drive.

Spare HDD – Lets say you want backups, but you only want to run them once a day.  A good way to save money is to add another drive to the server, but leave it unRAIDed (you save money by not using a RAID card.)  Then, simply use it to back up your main drive during off hours.

This blog post isn’t even close to a complete analysis of each type of upgrade, but hopefully it was able to give you an overview of what your options are.

As your site grows, your hardware might need to as well.  In fact, with any sizable growth, upgrading the hardware that is dishing out content to your users is inevitable.  Over the next few weeks we’ll be looking at some general strategy in upgrading hardware so you get the best bang for your buck.  After all, not in your interest or our interest to waste time, money, and resources on upgrades that don’t make sense.  If this is your first Linux Dedicated Server, you might want to check out last week’s post about renting your first dedicated server.

Choosing and Upgrading the Server Memory

Memory is something you need to think about before you pick which processor you want to go with, because that will dictate what rules you play by when it comes time to consider memory.  There are two ways to upgrade your memory – upgrade to faster memory (i.e. upgrade from DDR to DDR2 or from DDR2 to DDR3) or add more of whatever you’re using.

More memory never hurts.  It doesn’t necessarily help, either (if you upgrade well beyond what you need.)  Motherboards (the main component that ties your whole server together) have a limited number of memory slots and a maximum amount of memory supported.  Eventually you’ll hit your server’s maximum amount of memory.  If this ever happens, consider upgrading in the near future.

Not all memory is equal.  As the years go by, memory, just like a computer processor, evolves and operates at higher speeds.  We currently offer three types of memory – DDR, DDR2, and DDR3.  DDR2 is twice as fast as DDR.  DDR3 is twice as fast as DDR2.  Our Celerons support DDR memory, our Xeon 3220s support DDR2 memory, and our Xeon 5520/dual Xeon 5520 support DDR3 memory.  If a tech tells you your memory is inadequate, it’s possible that upgrading processors will have the nice side effect of upgrading your memory speed.

Lastly, upgrade memory in quantities that keep it going at full speed.  Based on the ways the DIMMS (memory slots) are set up in our servers, you’ll generally want to upgrade your DDR memory in multiples of 1gb, DDR2 memory in multiples of 2gb, and DDR3 memory in multiples of 6gb.  Ask our sales team at sales (at) ubiquityhosting (dot) com for more information about your server, because it will vary from server to server.