my thoughts, musings, rants and discoveries . . .

With an increasing number of backup solutions available, I’d like ask a question. Why are so many people reluctant to backup their data?

I see this problem all the time. Even in corporate environments, where backup is pretty much mandated, people are reluctant to use the solutions. Some go out of their way to avoid mandated solutions. But really, how many of us have been bitten over the years about losing data because our system crashed?

The bottom line is there is almost no excuse for not backing up ones data on a fairly regular basis. Having said that, I’d like to actually list some cool ideas and why I wouldn’t rely on them.

1. Apple’s Time Machine: is really a pretty cool solution, for the most part. If you’ve never tried it, and own a Mac, you should. Having said that, the problem really isn’t in using it the first time. It’s the 2nd time and after that. About the best thing Time Machine is good for is when it comes to restoration of a Mac. It will capture and restore every little thing including cookies, passwords, etc. You don’t have to worry about re-installing all of your applications either.

   The reason I say it’s not so great the 2nd time is that depending on how much you actually have to backup, the time required to determine what has changed can be quite long. While the actual backup part can be quite fast, it’s this generation of a list that can be intolerable. I suppose it’s a small price to pay for the ability to restore everything so easily.

2. Dropbox: is currently a pretty hot item in the reviews. It’s a pure software widget that allows you to upload all of your files, share them with others and it’s dead simple to install and use. The service is also pretty cheap at just $9.99 per month for 50GB.

   The problem with the Dropbox solution is time. You have to have lots and lots of time depending on the speed of your internet access and how much data you really wish to backup online. If you are with a cable provider, you may not think twice about uploads. However, if you’re in select areas where AT&T DSL is your only option, you may be in deep kimchee. This has certainly been the case for several friends of mine who only have 384kbps upload capabilities. In this scenario, backing up 50GB of data can take almost a week.

   Even if upload speeds aren’t a problem, you may find a problem when it comes to restoration of your computer from an online backup. Imagine downloading your apps, data, settings, etc. just so you can get back to functional.

3. Then there are the hybrid solutions, which place a locally installed appliance to help mediate this whole process and to provide local restoration capabilities. The trade off with this approach is cost of the service and management of some extra hardware.

   So far, while I believe this is one of the best approaches, the key question in my mind is whether or not this is worth a premium monthly service. Just to illustrate the differences in prices, let’s compare the Dropbox service with service from someone like Barracuda Networks. If we take a look at the costs for backing up 50GB of storage, we’re looking at a tenfold difference– $0.20/GB to $2.00/GB.

   Some will argue that the additional costs are to subsidize the hardware that is deployed. And that may indeed be a fair argument for some. But I know in the case of Barracuda, they charge for their appliances. Even at a reduced rate, they’re by no means losing money. Still, despite the costs, they are clearly selling their offerings to corporate accounts.

So with these types of solutions, is backing up still just too much of a pain in the butt for you? Do any of these sound at all attractive?

I fundamentally believe that there is no real reason not to protect all of your data. But for whatever reason, backup is just not a very interesting application. Perhaps we don’t backup simply because in a disconnected world of laptops usage, there’s not enough to remind us to do such a simple task.

It seems as though the subject of de-duplication is becoming more commonplace as it’s getting difficult to avoid conversations related to it. Just the other day, I had another customer ask me for a recommendation concerning it. While their question was interesting to me, their assumptions for considering it were more so.

It’s amazing the assumptions some people will make in my industry. I can remember a while back, I had a CEO for a company state to me they wanted to get into the storage industry but wanted to do so without the use of disk drives because “We all know you can’t make any margin on storage”. Well, that’s just not the case.

Anyway, the presumption on why de-duplication could be a good infrastructure solution is that they’d been told they could reduce storage consumption by as much as 50%. Really? 50%? I suppose if this were high school and plagiarism was pervasive, this might be so. But it’s not likely that people are replicating that many files. Having said that, de-duplication technology can be beneficial and I provided them an example of how it can help.

In my example, I took a fairly typical scenario in my company where I might generate a presentation that is roughly 25MB in size. It’s a corporate presentation so I might actually shoot a copy of it to all of the employees. We’ll round things to 20 just to make this rather easy. Assuming each employee is using an email reader that downloads attachments to a local folder, we’re looking at 500MB of the same file being stored onto laptops, desktops, etc.

In this day and age, 500MBs isn’t very much. So is the issue the actual storage consumption? After all, if we’re talking about 20 unique individuals, on laptops, and working from different remote offices, we’re looking at 25MB of consumption each. But if we throw all of these people into a central office and decide we want to provide backup for all of them, the problem starts to take shape. On the local area network, backing up this much information to a file server should take roughly 5 seconds on a gigabit network if there’s nothing else going on. That’s hardly anything to be concerned about. It’s not until we wish to back things up remotely that we actually see the problem.

Let’s suppose our company has a T1 connection to the internet and we can upload files at 1.2Mb/s (megabits, not megabytes). All of a sudden we’re looking at backing up all instances of this file to the remote site at fairly slow bandwidth. In this example, assuming we can get 100% utilization, we’re looking at a transit time of 3,333 seconds or roughly 56 minutes. Gee, nearly an hour just to backup all of the instances of one file over the internet. That’s a lot of utilization.

If we simply backed up the first instance and could replace all replicated objects with just hardlinks, the total job would only take 2.8 minutes. If only things were this easy. Actually, for the most part, they are.

There is a definitely a case for de-duplication when it comes to cutting down on WAN transit times for remote backups. But does this actually warrant a product purchase when the products typically cost much more than the storage that will be housing the data? Perhaps, if you look at it from a network bandwidth perspective. But when you consider emerging technologies such as WiMAX, you should re-check making expensive purchases.

There has been much hype about rolling out large WiMAX fabrics in key metro areas. With speeds as high as 72Mb/s, we’re talking about a backup window of possibly as low as 76 seconds. How real is this? Well, just in the last couple of weeks, we’ve read about Charter Communications rolling out support for their cable subscribers in the realm of 160Mb/s. The service is available in very select regions and at a premium. But that’s not the point. The point is that network bandwidth to the internet is increasing at a dramatic rate (no pun intended) and at very affordable prices.

When you consider these things, where does de-duplication really fall? Storage is getting cheaper. Network bandwidth is getting much faster and more affordable. And more and more companies are beginning to make de-duplication scripts or applications available free of charge. With these things coming and affecting much more than just backups, it’s hard to justify such capital expenses.

The reason I got on the subject is that I just read a news article on yet another FREE de-duplication tool called Dupe Manager. It’s only a 0.1 release, but it’s not the only one out there. And, it’s free. Did I mention that? Well, it is. This is yet another reason I love working with open source. Open source lives and breathes to solve simple, everyday problems like these, which makes a lot more sense considering the relative nature of the problem. It’s a bit problem for many today. But will it continue to be as big a problem in 6 months? 12 months?

If you're using Suse Linux Enterprise Server (SLES) 10 SP2 and want the latest version of the iSCSI Enterprise Target (IET), aka iscsitarget, you'll want this patch set to basically back out a lot of the compatibility stuff that was placed in the patch's Makefile.

The reason this special patch is necessary is due to the backporting of numerous functions from 2.6.25 kernel into Novell's 2.6.16 kernel used in SLES10SP2. This version update is especially necessary if you wish to make use of blockio.

To apply this patch, use the following procedures:

1. copy the iet-0.4.17 to your system: 'svn checkout svn://svn.berlios.de/iscsitarget/trunk'
2. cd trunk/
3. patch -p1 < ../[this patch]
4. make sure openssl-devl is installed
5. make&&make install

Enjoy!