Opinion: Online Backup Services, too good to be true?

I recently shared views about Digital Asset Management (part 1 and part 2) and I made mention of online backup as an additional security layer.

There are a number of challenges with online backups, 3 are particularly obvious

  1. Your data is in the cloud, as such it might be vulnerable to unauthorised access by third party. Solution: encrypt your data before sending for backup, do use super strong keys (as we now know as a fact that unless you are using very strong keys, you are vulnerable to attacks such as from the NSA)
  2. The backup company you are using might go bankrupt, not only don’t you know what will happen then with your data but your safety net is just gone. Solution: as for point 1 encryption, and use of multiple online backup to diversify your risks
  3. The backup company might change its pricing policy and what used to be affordable is becoming unreasonable financially however you remain hostage of a company that is holding your backup. Solution: use multiple online backup to diversify your risks and if prices are becoming unreasonable, you can just drop one provider

There is however one issue more important than these 3 and which is much less obvious:

What bandwidth have you got to actually transfer the data to be backed up?

Effectively many companies are offering premium unlimited plans which are practically not unlimited at all. The most affordable of them is CrashPlan which offers a unlimited plan for a single computer for USD 60 a year. Unlimited capacity? yes on paper, effectively people are reporting in forums very low transfer speed and when being questioned, the company answers that they try to do their best to allow user to transfer 10Gb of data per day.

Now let’s do some simple math. 10Gb per day represent a speed of less than 1Mbps. Many people are reporting that their maximum speed with CrashPlan is around 2Mbps, so according to the company line they should be happy with double what the company sees as the floor bandwidth, right?

Well, not quite. We need to look at it from a different angle. Advanced amateur photographers will use memory card typically of the order of 32 or 64Gb and might be filling for a day a couple of them. Say they come back with 60Gb of keepers (including the various PSD files they may derive from the keepers) per day, which at 2Mbps takes about 3 days to backup. As amateur you might shoot only during weekends, that’s twice that amount, thus you’ll need 6 days to backup your file from 2 days of shooting, that’s fine. But if you shoot any more, if you and your wife or children are shooting with you and you want to backup, your dataset will grow faster than what you can backup online… simple as that. If the actual bandwidth decreases you might not even be able to ever finish your backup…

In my personal case, I just upgraded from cable to fibre and I have now a sumptuous 200Mbps upload of which I am using around 4Mbps for backup to CrashPlan and to Amazon Glacier… I will certainly monitor the situation and approach my providers to see how this can be improved but it is sounding less and less promising.

Additionally there is a pattern I detect as well: online backup services are increasing their prices without increasing their quality. If you keep in mind that Moore’s law also applies to computer storage, i.e. storage capacity will double every 18 months or so for the same or even lower price, there is something wrong again here.

In a nutshell the leading online backup company with which I had experience are as follow

  • Mozy: Mozy has a rather slow bandwidth and dropped their unlimited plans which is replaced by a per Gb price. For many people a few hundred bucks a year became a few thousands for their data
  • Carbonite: Same as Mozy with a very slow bandwidth
  • CrashPlan: in appearance a good option with an inexpensive unlimited plan, however much too slow speed to complete the initial backup unless you have a small data set and it is not growing too much
  • Amazon Glacier: this is an interesting plan, at about USD 0.01 per Gb/month it is cheap but not high speed either (though I have to monitor this one since my connection upgrade). It is more expensive to restore data though but a backup is an insurance, you hope you won’t ever need it.

I would surely wish to see more transparency from this emerging industry. In particular I would love to see a graphical distribution of their measured speed (from their end) viewed by user location (even ISP), by Backup set size (there are doubts out there that the biggest set receive lower priority to satisfy the biggest number of customers), by plans to see how much money really buys… Now of course it won’t prevent them from lying about their true performance but it would give some needed transparency for customers to make choice and competition to push price down and quality up.

But this might be too much to ask: like most telecom companies, they orchestrate their lack of transparency to prevent customer from being able to compare. Unfortunately, we see the same trend with online backup services.

I would be quite interested if you can share your experience with these service providers or if you know other offering good performance and are reliable as company.

Opinion: Online Backup Services, too good to be true?Christian C. Berclaz

9 comments on "Opinion: Online Backup Services, too good to be true?"

  1. Stephen Ermann

    My (short) opinion is that if you produce 20GB/day of data worth protecting, you should probably not be using an “unlimited backup for $1.99/month” provider – it is not possible for them to produce the kind of service needed at that price.

    • Christian C. Berclaz

      Stephen you might be right there, it might be the same issue as the old unlimited data plan from ATT on cell phones. The fact remains: should you collect a fee for a service you qualify as unlimited when it is, in fact, limited and you are not transparent about how it is limited?

    • Stephen Ermann

      I would argue that theses services are effectively “unlimited” for the vast majority of users, but are not guaranteed to work perfectly in every conditions, in every case.

      A few years ago, a friend of mine tried launching an online backup service for pro-am photographers. I don’t think he was very successful in that project.

      Providing that kind of service requires a significant investment in storage space, bandwith and management. In addition, services like CrashPlan are usually based in the USA, so you are dependent on the international Internet links. Providing a service that could handle 30GB/day/user would probably need to be somewhat local.

      Given Singapore’s love for photography, it might make sense to find or launch a local provider…

    • Christian C. Berclaz

      In first approximation the question of profitability is a matter of playing on Moore’s law and on usage distribution. Like many other things usage will vary and many users with low usage will finance a few users with higher usage (that’s how fidelity program works, that’s how “all you can eat” businesses operate, etc). Now as the user base expands over time also the storage cost decrease as well as your $/bps. You have to share your overall bandwidth with all users and as bigger users need higher speed, this is where they get relatively lower speed than the lower users (10Gb per day is super high if your backup size is 30Gb, but the same 10Gb is ridiculous when your backup size is 1Tb). My point is that the service should be marketed in a fair way but I’m convinced this is very profitable.

      As to the question of the speed, achieving around 4-5Mbps would be already a great improvement and remains very reasonable for intercontinental internet traffic.

      Let’s see if there is a local alternative and let’s discuss a launch 😉

    • Stephen Ermann

      The issue is that “unlimited” services tend to attract the very heavy users, potentially destroying the business model for everyone.

      Which is why every recent “unlimited” service that I know has some kind of built-in throttling – all-you-can-eat-buffets limited to 3 hours per session, unlimited mobile calling limited to 100 different numbers per period, unlimited music download to 5 devices, …

      And “unlimited” means the provider does not limit your usage, but does not mean that he has to build something special for you. You don’t expect CrashPlan to pull a dedicated fiber-optics line from Atlanta to your home for $4/month, do you?

    • Christian C. Berclaz

      The matter is how we practically define unlimited storage. Nowadays the standard storage measure for non-tablet format is the terabyte. At 10Gb per day it means 100 days to backup which is unpractical. Why? because it represents a timeframe where high-end technologies are more likely to become mainstream. What is today extreme (like 4K TV) will rapidly become more common and storage need will increase, thus online backup must adjust accordingly.

  2. Christian C. Berclaz

    I promised to report back on Amazon Glacier. Unlike what I indicated in the post, Amazon Glacier is not low speed and rather performing well in the area of 20+ Gb a day. Interestingly it seems to backup by bursts: at times transfer are slow as a snail and at other times it runs at speed around 20-30Mbps. In average Amazon Glacier provides very acceptable speed.

    I am currently exploring alternatives with BackBlaze.com, which seems to provide speed around 3-4Mbps in Singapore (my ideal target for low cost backup) and Backup4U.ch a Swiss based company.

Leave a Reply