Everybody knows they should have backups of their data. Many times hosting clients assume their host is keeping bulletproof backups of their site, their database data, etc… Often times hosting customers will not be vigilant in performing their own backups regularly. And time and time again, this comes back to bite them when they desperately need a restore…

There is no magic bullet

Although there are numerous ways for a web host to perform backups of client data, there is no magic formula to create reliable backups that are cheap and fast. Because of this, budget hosts (hosts who offer extremely low priced hosting) have to “cut corners” many times when it comes to backups.

With a low priced host, maybe they are skimping on customer support, using cheap hardware, using overcrowded servers, or sacrificing backups. Unfortunately, backups often take a “back seat” as they can be costly for a hosting company to perform backups correctly.

We’ve seen hosts that do no backups (your data is your responsibility), hosts that do backups to cheap drives and do not verify the data is intact, and hosts that perform backups only at random times. These type of practices can put your entire online business at risk.

Do your own backups

If you have a website, then you should regularly backup your own data. Even if your host has the best backup procedures in place, accidents do happen. If your website is only static html and images, it’s easy to keep a local copy of your webpages and graphics. You can use FTP or your web design software to transfer your files to your local computer. Or you could zip up your files and download the zip file.

But what if your website uses dynamic features like a blog, or you have an ecommerce store? Properly backing up this data is much more involved, and in most cases, more important than ever. Why? Because it is often this dynamic database driven data that is the least backed up by many hosts.

If you use MySQL databases, you have a few options:

• Use a web based program like phpmyadmin to export your data as an SQL dump file.

• Have your php script perform this sql dump to a file on the server that you FTP to your local computer.

• Setup a cron job (program that runs on a schedule) to perform an SQL dump via a server side program like “mysqldump”.

Note: If your database uses innoDB formatted tables, taking your own backups is doubly important, as most hosts cannot adequately backup innoDB tables/databases that allow for granular (i.e. specific table) restores.

For an ecommerce store, not only should you backup the database data (pages, products, coupons, etc…), but you should also have local copies of the templates and settings for a quick restore. If you use the latest version of ShopSite, you’re in luck, as ShopSite now makes this easy to do.

Restores can take a long time

When a host needs to restore a server from backups, we’ve seen some disaster stories emerge over the past few years. This is where the “rubber meets the road”…

1. The backups are corrupt
The host goes to restore the data, but finds that their backups are actually corrupt and not usable. If the host does not regularly test their data for integrity, this can happen, and often at the worst time. This results in either a partial restore, or a complete loss of all data… NOT GOOD.

2. The backup drive / server is SLOW
We’ve found that 50 GB an hour for restore speed is a good rule of thumb assuming the hardware where the backups are stored is reliable and fast. However, if your host is using slow old hard drives, or their network is slow for network based restores, this can drop to painstakingly slow rates, which will make the restore take quite a long time. How long? It depends on how much data needs to be restored…

3. There are Terabytes (TB) of data to restore
Some hosts love to offer large amounts of space, or unlimited space. They have servers with multiple TB drives that can store TB’s of data. If they need to restore these Terabytes of data, and the avg. speed is 50 GB/hour, you could be looking at 20-80 hours or more of time for the server to be back up and running. Yikes! We’ve seen some VPS providers take days to a week to recover from multiple drive failures when they need to restore Terabytes of data.

How LexiConn performs backups

For us, we use a variety of backup methods to insure:
1. Redundancy
2. Reliability
3. Speedy restores
4. Disaster Recovery

Redundancy

We achieve the first level of redundancy by utilizing hardware RAID-1 in all of our servers. This means that each server has two hard drives with mirrored data so that any one hard drive failure will not result in the server crashing, or any loss of data. Although RAID is not an actual backup method, it’s a great way to prevent data loss.

Our second level of redundancy is the 3 separate backup systems we employ:

1. Additional hard drive on the server
Each server has an additional Serially Attached SCSI hard drive that is used for backups. We perform a full backup each night, and keep 5 days worth of backups. We can restore any files or websites for a client upon request, and this backup can be used for a full server restore if our continuous backup system fails.

2. Continuous real-time backup servers
We have a few central backup servers that operate on a 4 hour backup schedule to perform backups of all of our servers. They utilize continuous backup software that can be used to “bare metal” restore an entire server. This means the most data that would be lost is up to 4 hours in the event of a complete server failure.

3. Remote offsite backup
Each day we perform a backup to a geographically separate datacenter. Each server’s data is backed up to this remote location.

Reliability

Our first line of defense for reliability is that our servers use Serially Attached SCSI (SAS) hard drives, which are known for their stability and longevity. They are enterprise grade drives which reduce the potential for hard drive failure, operate quite quickly, and do not contribute to the load on a server. Our backup drive on each server is also a SAS drive, which results in almost no load during backups, and results in a much quicker restore.

Additionally, our backup servers and remote offsite backup servers all utilize RAID-10 technology with a minimum of 6 drives per server, which allows for fast backups, data integrity, and faster restores.

This means that each client’s data is stored on a minimum of 15 hard drives in various backup locations.

Speedy Restores

We achieve fast restores for an entire server by using enterprise continuous backup software which can bare metal restore a server at the block level. Additionally, we limit the amount of data stored on each server (we do not overload or pack our servers and offer realistic data storage limits), so even at the avg. rate of 50 GB/hour, it will only take 2-4 hours at most for a full restore to complete. By using SAS drives, we maximize the speed at which a restore can complete (typically 2-3 times faster than cheaper SATA drives).

Disaster Recovery (DR)

Our DR plan includes the use of our remote, geographically diverse backup servers that backup each server daily. In the unlikely event of a meteor hitting our datacenter, we can restore data from this remote location, or even re-locate servers to another datacenter.

…

The bottom line… ask questions about backups before signing up for hosting, be informed about your hosts’ backup policies, and keep your own backups performed regularly to avoid disaster.

photo credit