I posted recently about web site hosting, but the question came up of how you measure the reliability of a web host. This is a pretty deep subject that you could write a book about.
Of course, to start off, you have to define some things, like what reliability means to you and how much you really need it. Are you running an ecommerce site that gets millions of hits and sales per day, with a minute or two of downtime or sluggish page loading leading to thousands in lost revenue? Are you running a private family wiki to plan your family reunion next year that probably no one will notice if it's down a day or two? How much is reliability really worth to you? You can find plans ranging anywhere from $4 to $400 per month for hosting services, all with different levels of guaranteed service and different amounts of storage space and file transfer or bandwidth.
Going back to what reliability means, we might ask what are the causes of unreliable data or data loss? Well, there's user error, such as deleting or saving over the top of a file. Then there is hardware failure where a drive actually crashes and everything on it is lost or the network is disrupted in some way and your data is temporarily unavailable. Chances are most hosts will have something in place to keep their hardware running, with some form of RAID and maybe even clustering. RAID is a redundant set of hard drives where one can die and the others keep running. Clustering is a similar concept, but instead of just redundant disks, you have redundant servers. Throwing a couple more disks in a machine is a lot cheaper than setting up a completely separate server, so be prepared to pay a lot more for clustering.
WestHost claims to have 99.9% uptime, which is about 40 minutes of downtime per month. Site5 is another host who has a 99.9% uptime guarantee, where you will get a prorated credit if there are unscheduled outages over 45 minutes in a month. How do you verify that? Well, you can look to someone like Netcraft who monitors the web and tells you all sorts of things like what operating systems web servers are running and how long they have been running since they rebooted. They also have a page that shows hosting providers and shows how quickly they respond and how much downtime they have experienced. That chart is nice, if your host is one that has paid to be included on it.
In order to recover from PEBKAC errors (problem exists between keyboard and chair), look for a host that provides the ability to create cron jobs, which are scheduled tasks you can use to back up your files, so in case you delete or save over something, you can recover it. You'll probably have to set these up yourself. If you have enough disk space on your host's server, you can backup to that same server, but you'll want to backup to another location as well. Don't schedule so many offsite backups that you use up your monthly bandwidth quota, though.
Many companies will have a 30 or 60 day money back guarantee or a free trial period. Talk to people you know about how they like their host. Look at what kind of tech support availability they have. Although many companies will let you get by without a contract, most will give you a pretty decent discount for prepaying a year or two in advance, so you'll want to check to see how much of that is refundable if you want out after the guarantee period.
So, when choosing a host and a plan, it all depends on what you need and what you are willing to pay. No matter how much you pay and how foolproof your host's backup plan seems, never trust a single layer of backup. The key word is redundancy. How redundant? At least one layer deeper than you think you'll ever need.
No comments:
Post a Comment