Web hosting is a very different marketplace to how it once was. Having a website does not automatically beget a hosting plan. Now you have to think about redundancy etc. What do you need to look for?
Introduction
I want a website hosted, and I don’t want it to go offline. Where do I start?” Yes, sometimes we still get phone calls that start like that at Hostway; and that is despite the substantial investment we have put into the content on our site, to help new users and old-timers find the answers they need.
For a company that deals with the Internet 24x7, and with staff that spend their lives learning about new technologies, understanding standards and protocols, and reading The Register in quieter periods, encounters with customers that are new to the internet are still a revelation to us.
But that is the perception that you have when you are so focused on your own industry. Just because I find it necessary to scan the BBC News site every day to get my fix of current events, it doesn’t mean that everyone follows the same pattern that I do – even if the media tells us that we are more connected than ever.
The real surprise now though is that we are not getting calls like that just from people who are new to the internet. We’re getting them from customers that have had a website since before the dotcom boom and bust, and companies that run complex ecommerce operations. The problem is that now there is almost too much choice. Shared hosting, Virtual Private Servers, Dedicated servers, redundancy, clustering, failover, geographic load-balancing, firewalls, VPN, DDOS, GPL, proprietary, UNIX, Linux, FreeBSD, Solaris, RedHat, Oracle, DB2, MySQL, Dell, HP, Cisco, Juniper and so on.
Just writing that list makes your hands hurt. How does it feel to consider each option and pare it down to the right solution for you and your company? And when you make a decision, how do you know that you have got it right?
Basically speaking…
To be fair, most companies actually only need a very basic hosting solution. This is because the majority of companies still use the Internet as a subset of their marketing campaigns – to provide information and act a repository for contact information.
In this respect, the type of hosting a company will generally need can generally be satisfied by the shared model; one of a number of sites sitting on a shared server, using the same resources, and using low bandwidth and processor/RAM cycles. The benefit of this model is that server management is taken care of for you by your hosting company and the price is going to be low.
However, it is also likely to mean that you will be at the mercy of the server load – if other sites use excessive resources, then your website is going to suffer, and if your site goes down, you haven’t got a lot of comeback; you won’t get a Service Level Agreement (SLA) on a low-cost hosting plan.
Middle of the Road…
No, not a soft rock band, although I could name many, but way of combining the shared model with the dedicated server model. A Virtual Private Server (VPS) allows you more space, guaranteed processor/RAM cycles, and partitioning from other users on the same system. At the back-end, it can also enable the movement of virtual environments from one piece of hardware to another without noticeable downtime.
More expensive and offering more features, the VPS model is especially good for resellers, marketing companies, and those that want to provide a white-label solution. Of course, you could still have problems, and only some companies will offer an SLA, but overall, service should be better.
One of the main benefits is the ability to interact with your server via a user interface (shell). With a VPS, you get root access, meaning that you can choose which services your VPS runs, configure Apache and Tomcat, and install your own applications. You’re also not tied into version control, meaning that if you want to test PHP & MySQL 5 in a production environment, it is all yours.
Dedication, that’s what you need…
Roy Castle sang it, and it still makes sense today. If you are a company that relies upon your website for any level of sales or revenue, then you need assurances that these revenue streams are not going to disappear from beneath your feet when you have just spent £10K on a multi-level marketing campaign.
Quite simply, if you make money from your website, you need to know that your website is going to be accessible every moment of every day. You will need to accept reboots, service upgrades, and occasional downtime. But the principle is the key thing to remember.
A dedicated server allows you to run your own website or applications, without the need to worry about other customers or users using any of the resources on that hardware. Essentially the server is yours to use, and though you may pay a higher price, you know that the only customer using that hardware is yourself.
At the same time, you are going to be able to sign a SLA with the hosting or co-location company that you use. A SLA will set down guidelines for response times, fix times, and key contacts in the company for you to speak to in the case of an issue affecting your website or server.
Though your only real tangible benefit with an SLA is an agreed percentage of recompense when you chosen company fails to meet target guidelines, an SLA will help to assure you that the company you work with has defined processes in place and recognises the need to respond to them correctly.
Risky Business…
For every hour that your website is not available, how much traffic are you going to lose? And how does that lost traffic convert into lost sales? If you have 24,000 visitors every day, you could lose, on average, 1,000 visitors if your website is offline for just one hour. If you have a conversion ratio of 10% on your site, you are going to lose 100 customers. Can you afford to lose 100 sales? Of those 1,000 customers that cannot reach your website during that hour, how many will try again another time? You may lose 10% just because they think your service is unreliable.
But would even your own server, and a SLA stop the worst from happening? Not at all. In fact, I have known companies that have had a dedicated server and SLA, and a business-critical website, who have been disappeared from the internet map for up to 48 hours at a time.
But it does minimise risk, and as you scale up, your risk falls exponentially.
Risk is a key concept in the world of hosting right now. Events such as 9/11 and 7/7, the Asian tsunami and Hurricane Katrina, have highlighted the need for disaster recovery and stability. Many companies are now reconsidering their system design and deployment, in the knowledge that a terrorist incident or natural phenomenon could bring their services to a halt.
So we’re in a period where clients are willing to explore new technologies as a way of mitigating risk to an acceptable level. And that is where the confusion sets in. Up until now, what we have been discussing has almost become de rigeur in the hosting marketplace. Anyone with access to the right tools and with time on their side could put into place a hosting solution that would fall into one of the models described above.
Powerful options…
The only way to drive risk down is to find single points of failure in a system and remove them. In practice, you need to provide a backup wherever there is a perceived failure point that could affect an operation.
For instance, you may already have a dedicated server and an SLA. That’s excellent; your company is very pleased to be able to dedicate all the resources of that hardware into making your website responsive and quick to generate code and pages for numerous users.
But if the power supply fails in your server (they happen more often than you may realise), then your server is offline and only a replacement PSU is going to get it back up and running again. To do that, you need to unrack the server, open it up, replace the faulty part, re-rack, and switch back on again. And, of course, wait to see if the Operating System (OS) comes back up again.
So this is a single point of failure. If your power supply unit (PSU) fails, your server fails. Now, by adding a second PSU, you server could continue to run even if the first one stops working. You would still need to schedule a time to fit a replacement, but this could be done at night, or during one of your quieter periods, and you can tell customers when that is going to be. This is much better than unexpected downtime in the middle of your busiest part of the day.
Think Carefully…
This principle applies to almost all areas of your hosting model. You can have RAID arrays setup to mirror your hard drive content, or replicate files to a backup server that just sits in a data centre waiting to be used if the first one fails.
You can load-balance your hardware, sending requests to different servers based on algorithms. If one fails, the load-balancer just sends all the requests to the server that is still in operation.
Or you could host servers in different geographic locations in case that wave ever comes up the Thames and breaks through the barrier, flooding Docklands and the home of the majority of internet operations in the UK.
Alternatively, you could do something as simple as hosting with a company that has offices above ground-level. And yes, and that is a serious concern right now!
Costly, Costly…
There are options out there that can make your systems as safe and secure as a traditional bricks and mortar operation, and a good hosting company will work with you to identify where they feel your online systems are at risk.
However the trade-off is always going to be cost. How much can you afford to spend? And at the opposite end of the spectrum, how much can you afford to lose? In this respect, it is important to come to an understanding of where you can afford to spend less, and the impact that it will have on your business.
For instance, you may feel that if you have a website hosted on a two load-balanced servers, then you don’t have a need for redundant power supplies. Or you may decide that you would rather have RAID 1 HDD and a second power supply rather than the cost of introducing a load-balancer, which could in itself, be a point of failure.
Tie Me Up, Tie Me Down…
Of course, you do have one over-riding advantage. Any readers of this article are going to be looking at a *NIX based solution, which is not only going to be easier to administer in the long-term, but will also save money in terms of hardware provision and application licensing. And that is going to prevent you from being tied in to suppliers with proprietary solutions that have high minimum requirements and do not offer all the features you need. With Open-Source, if it doesn’t exist, you at least have the option to create or change it.
The End is Near…
So when a customer phones us and asks us what type of hosting he needs, we tend to find that our conversations last longer than they did a couple of years ago; and they are much more complicated. We try to explain all of the above and fit the right solution to the needs of the customer, and the more options there are, the more you have to talk about.
But how do you know that at the end of the day, you got it right? Well, hopefully this article will help you understand the trade-offs that you may need to make, and which decision is right for your company. If not, then remember that there are companies out there that can give you the advice that you need.
_________________________________________
|
Title
Click here to print our PDF pricelist.