A colleague sent me the following link the other day: The Case Against Tape Backup. The link gets you to a white paper written by Double Take Software, which argues that using tape for remote office backup is a bad idea.
Their conclusion? That instead of tape, we should employ CDP software, and that this is much more cost effective.
Unfortunately the paper is deeply flawed. Don't get me wrong, I would be very happy if the general conclusions they drew were true. I would be just as happy if they were content to conclude only that tape was bad--meaning too expensive and too unreliable.
However, they are not content to reach such a conclusion, and in their attempt to make the leap from "tape is bad" to "CDP is good" they go dramatically awry in two respects. First, they posit a false dichotomy. It is not like the only alternative to tape is CDP for remote backup. And, in fact, some of the solutions that they (conveniently) ignore, are substantially less expensive than their proposed CDP solution. Second, their financial reasoning is pretty badly flawed. And I have never been a fan of make your own solution look good by falsely making the alternative solution look bad.
However, I do think that following through the line of reasoning in the white paper is pretty interesting--and by doing so we can expose some better data about the costs of remote office backup and the alternatives to tape.
So were does Double Take start? They say: "There was a time when tape-based backup was widely believed to be the only feasible backup solution for remote offices. But tape backup alone is no longer the only realistic solution and centralized backup solutions are coming to the forefront. As a result of this shift, enterprises have re-evaluated tape-only backup solutions and don't like what they have found. According to a report by ESG , nearly one-quarter (24%) of companies say that twenty percent (20%) or more of their tape-based backups fail. As such, depending on tape backup alone creates an unacceptable level of risk."
OK, so far, so good. Sort of. I would be a lot happier if the statement was 20% of tape based restores fail. Because honestly, if 20% of your tape based backups are failing, you are doing something wrong. As much as I feel very strongly that tape is not the right target for backup any more, I still believe that done right, tape should provide a 98-99% reliability for backup and restore. If you aren't getting that, you need to try Data Protection Advisor to understand why you aren't.
That is more of a quibble than anything though. From their, the authors write: "Fortunately, advances in technology have made centralized backup easier to manage and less of a drain on the WAN. A centralized backup solution takes the tape-only backup responsibilities out of the hands of non-technical resources in each remote office, and puts them in the hands of experts back at a central data center. Replicating data to a central location from branch offices using Double-Take can reduce the per-location costs associated with tape-only solutions and provide a higher level of recoverability for business-critical systems and data."
And this is where the false dichotomy comes in. What the paper says is true, but it is not complete. There are alternatives to Double Take's CDP that are even better at reducing the costs per location associated with data protection and recovery.
For argument's sake, I am going to focus on one of those alternatives: Avamar. I will compare and contrast the costs associated with using Avamar for remote office protection. Yes, there are other alternatives, such as using small deduplication appliances at remote sites and replicating those to a central office. And yes those alternatives make financial sense too. But to avoid too much complexity in the cost discussion, I will stick to Avamar.
So let's take the cost format provided by Double Take, but add Avamar to the equation:
|Tape Backup||Double Take||Avamar|
|Hardware, Software, Setup||$120,000||$31,750||$800|
|Off-site pickup and storage||$36,000||$0||$0|
|Year 1 Total||$360,000||$52,884||$8,500|
A quick explanation. The $800 charge for Avamar is based on a total hardware and software cost for Avamar Virtual Edition of $20,000 per TB ($20 per GB). I am not really sure if that is exactly accurate, but it is close enough. And according to Double Take's reasoning, we are only trying to protect 20 MB of changed data per day per office. From an Avamar perspective, that would normally mean about 2 GB of source data per office, or 40 GB total. From a salary perspective, if we assume 20 offices, that would normally mean 20 servers. In my experience, each backup administrator can take care of 400 servers (best case). So we assign 1/20th of the burdened cost of a backup administrator, then we have a cost of $7,500. (And for all you backup admins who are wondering where you can get a job that pays $150,000 per year--note that I said burdened cost. This is the cost I normally use in business cases, and includes all the costs associated with a role, such as salary, benefits, infrastructure, office space, and so on.)
Therefore, the cost of an Avamar solution is 83% less than the Double Take alternative--and more than 97% less than the cost of tape!
How about the cost of recovery? Well Double Take takes a crack at that too. Once again, I will use their approach, but include Avamar. I am also going to make one or two very significant corrections. First, the table as they present it:
|Tape Backup||Double Take||Avamar|
|Cost of unrecoverable data ($10,000/MB)||$200,000||$694.44||?|
|Cost of downtime ($42,000/hr)||$336,000||$0||?|
|Year 1 Total||$536,000||$694.44||?|
The assumption here is that with tape, I will recover to the previous days' data. Meaning that you lose 1 business day's worth of data. Secondarily, we are assuming that we will take 8 hours to recover from tape. (That is also explains how I derived the amount of source data for the first table--if there is $200,000 worth of data loss, then that is 20 MB. If we have a 0.1% daily change rate, then that means there is 2 GB of source data.)
Now I have no idea how Double Take derived the $694.44 figure. They claim that is 5 minutes of data loss. If each office generates 20 MB per 8 hour day, then 5 minutes would be about 210 KB of data. But that error is trivial compared to the real problem here: Double Take is assuming we are doing data protection at a central site. That means there is no local recovery. That means that I have to send the entire recovery over the WAN (or do a local restore at the central site and physically ship the server to the remote location). To ship 2 GB over a WAN connection that has been sized for 20 MB is going to take a very long time. Days or longer--this is more than 1000x the amount of data the link is designed for. So my suspicion is that the real case is that it is going to take 1 week or more to recover the data with this approach. If I make that correction, I find that despite saying that I have $200,000 worth of lost data, that amount pales in comparison to the cost of downtime with the Double Take approach. (And as an aside--I think $10,000/MB, or $10 billion/TB is absurd. That would mean the cost of losing data for most companies is many many times their market capitalization.)
Finally, lets look at Avamar. I will be generous and use the ridiculous figure of $10,000/MB. Downtime however, is very short. If I use Avamar Virtual Edition, I will have a local backup from which I can restore. So lets suppose that takes 4 hours. We now have the following chart:
|Tape Backup||Double Take||Avamar|
|Cost of unrecoverable data ($10,000/MB)||$200,000||$694.44||$200,000|
|Cost of downtime ($42,000/hr)||$336,000||$1,680,000||$168,000|
|Year 1 Total||$536,000||$1,680,694.44||$368,000|
Now we see that Avamar costs approximately 40% less than tape, and 80% less than Double Take. If we take a more reasonable approach to the cost of unrecoverable data--say $100/MB--then the difference becomes even more remarkable. In that case, the technology with the overwhelmingly best cost basis is Avamar. Tape becomes about twice as expensive, and Double Take's approach is approximately 10 times more expensive.
TCO can be a very valuable tool that can help us understand the relative benefits of different technology alternatives. But it needs to be employed responsibly, and with some common sense. And it does no good to set up a false dichotomy: rarely is it the case that you can only do A or B, but have no alternative. In this case, we can see that the alternative, Avamar, is significantly less expensive in both the case of cost of backup, and the case of cost of recovery.