Computers brought us the ability to make records, photos, videos etc.., super safe and super backed up, relatively cheap and easily.
Computers brought us the way to lose everything in a single power surge by not doing backups.
The owners of Maintenance Connection Canada have built up a couple successful companies through the years (and then, when they finished their growth stage, sold them to people who wanted to maintain a company instead of starting one.)
Back in about 1986 the company we owned a company that provided municipal accounting to the majority of English Canadian municipalities (75% of all new sales, and we were listed as one of the top 100 software companies in Canada … admittedly about position 98.)
At that time we won an award for the fact that we had a consistent, simple, reliable backup strategy for ourselves and our customers.
Fast forward 30 years … we still believe in having good solid backup strategies.
And … we find that most individuals and companies we question rely on luck (things not breaking) instead of backups.
Safety by location
Let's first look at the safety of your backups based on location.
Simple copy
A backup made to the same Hard drive, SSD, USB key etc.., as the original.
This is the easiest backup, it is usually the quickest to recover using, it is usually the easiest to set up.
Its biggest advantage is for human error. If you save a word document to another file before you edit it you can start over if you make a big mistake. In the database world these are great if you want to try something 2 or more times and want to start from the same point each time (and don't mind throwing away your changes each time.)
It has a 'single point of failure' meaning, it has the lowest safety value. If the drive it is stored on stops working, you have likely lost both copies.
Maintenance Connection databases usually only use this in a test environment, not a live data environment, though the test database may be made from a recent live database.
Second drive
This is where you have 2 drives, one is your backup. This is actually significantly safer than the above option. But in most cases it still has a single point of failure. If it is stolen, or in a fire, or a power surge fries both drives (yes this can easily happen) you have lost it.
But in terms of probability, unless theft is a high probability for you, this strategy is significantly better than the simple copy to the same drive.
A lot of database setups take advantage of this. They keep the live database on one drive, they keep the backup database and the log on a second drive. This way if the 1st drive fails, they can completely restore up to the minute.
There can be some performance issues with this and Maintenance Connection databases seldom are set up this way.
Off Server/Off Machine
This is where you copy it to another computer. Now, if one computer crashes, your data is safe on the other computer. It may take a few hours or a day or two to get up and running, but is relatively safe.
By definition, this is a 'second drive' and more.
Warning: If you are using Virtual computers at a service provider, being on a different virtual computer could easily mean that you are on the same PHYSICAL computer, so that is NOT an off server backup, that is a copy backup that MIGHT be an off server backup.
Off Site
First of all, all off site servers are also 'off servers' but not all off servers are off site.
This is where you have your data in a completely different location, perhaps on opposite coasts of the country.
One of the more dramatic (and hot topics in 2017) expressions I've heard is "If North Korea where to bomb our Cloud/SaaS/Computer provider, all data in that site would be lost. So an off site is necessary'.
But lesser ones that have really happened:
- A fire
- A flood (or a fire followed by a flood caused by the fire suppression)
Off Network
A true story, to illustrate the first reason you should have an off network backup
We, in the past, tried using cloud at cost servers. We never trusted them to the level to use them as the primary machines for any of our customers, but for a 2 year period they had been reliable enough to be worth it because they were extremely low cost. The 3 year they went badly down-hill.
In July 2017 they 'lost' all 30 of our servers except 1.
In August 2017 they lost the remaining 1 (and a bunch of the others that we had rebuilt).
In July, they informed us 'the disk array failed and we lost 1000's of servers permanently, please rebuild from your backup'. As this is being written, all these servers have been down for almost a week and they have yet to respond other than deleting our request for assistance. When we finally talked to someone they said 'Because there were thousands of reports of lost servers, we didn't have time to respond to all the submitted issues so we just deleted them all'.
Now, our development test data was not important enough to store 'off network' so we lost all our dev server backups as well because of these 2 failures close together.
Note; We have had many other problems with cloud at cost, they are low cost but they can never ever – even their most expensive options – be trusted for anything that you aren't willing to restore from backups on average every 2 months.
The moral of the story:
You must not ever trust ONE outside source with ALL the copies of your critical data. It DOES happen that you will likely lose everything someday.
A second reason: a Virus
While SaaS and similar companies strive hard to keep viruses at bay. Once in a while one slips through and can infect (or delete) everything including all of your data.
Keeping periodic backups with a different company on a different network gives you protection from this.
How many backups?
By location
1 different offsite location is better than none.
2 different offsite locations is considered safe. This gives you your data in 3 locations.
Of course the more places and the more copies you save the higher the cost, so at some point, additional backups do not provide enough benefit to be worth the cost.
Grandfather Father Son
The idea here is that each one is a little older than the next.
One backup is good.
Two backups are pretty good.
Three backups are typically considered as safe as you can get. Why would not 4 or 5 or 35 be considered better? In practical terms, unless you are doing things a few minutes apart, if you get to the 4th or 5th backup, chances are it is so outdated that it is near useless. This doesn't mean you should throw them away if you have the space, but realize each after 3 gives a very insignificant increase in safety. And as I said above, the more places and the more copies you save the higher the cost, so at some point, additional backups do not provide enough benefit to be worth the cost.
And What type?
There are various types. Some names you will hear thrown around:
- Image
- Full copy
- Incremental
Image
- This is where an image of the hard drive with all the info on the hard drive is saved.
Full copy
- This is where a copy of all the information is stored, but not necessarily in the same exact layout. A SQL Backup for example is stored in a 'more efficient' method (but slower to access) than when in the database. Things like indexes for example are ignored because they can be (and will need to be) rebuilt if you restore.
Incremental
- Just the stuff that changed.
- In a SQL Backup this is often the log file – a list of changes.
- Usually much smaller than a full copy.
- Often used on systems that backup every 15 minutes or hour or similar 'shorter than a few days'.
Annual
- This is often used in tax situations and for corporate financial records. It can also for a system like Maintenance Connection be a 'worst case emergency backup'. These are often kept for 10 or more years for audit purposes. But realize: data from 10 years ago may be extremely hard to access today.
- These TYPES of backups can be done monthly as well and kept for 13 or so months. You again have to decide the cost/benefits of how often.
And is it actually a backup?
One of the scariest things that happens in computers is when you go to restore your backup and find out it is invalid.
A real life story
We had one customer in the 1980's who ran a backup every night. They called us up one day because their hard drive had died and they were having trouble restoring from their tape backup.
When we work working with them the first thing the backup operator said was 'When we first installed, the tape backup took 2 to 3 hours, but for the last couple months it has been wonderful, it only took a few minutes'. The person then pulled out a stack of daily print sheets that had the backup report on it. The backup report said on every page almost nothing other than: Backup failed, 0 bytes copied.
They had of course reused every tape in the last couple months (they only had about 10 to cycle through.0
We looked at this in horror and had to tell them the bad news.
There WAS a silver lining. We found out the hard drive was 'seized' up. We found that by applying a higher amperage power supply to it we could force it to spin. So we started copying data off starting with the highest priority data.
Then we started copying the other things that would make restoring faster. In the middle of copying these the hard drive released a puff of smoke and never turned again.
But we had restored all their data.
The moral of the story:
A backup that has not been certified as restorable is not a backup.
Practical advice
For practical reasons you probably don't want to test every backup, but you want to test from time to time to make sure your backup strategy is working. And if you find that one of your backups is flawed, you should shift into panic mode and do whatever it takes to get a backup NOW … then go back and figure out what went wrong.
But I have a SaaS provider, it's in the cloud – isn't everything safe?
Personally I would never trust another company for EVERY copy of our company's data. I do trust those that have proven to be trustworthy for most of my backups, but never for every backup.
There will almost certainly be a cost associated with having a regular backup from your SaaS provider, but I hardly think it is worth not at least having your own copies monthly.
If you are using one of our SaaS systems you might think: My data is perfectly safe, Maintenance Connection Canada takes backups so seriously I don't need any other source.
I hope we never lose that trust, but I don't rely on Microsoft or Google or anyone, so for me to tell you to trust us would be hypocritical.
If you are using our SaaS system, yes there is a cost, but we do it very close to our cost – we do not consider it be part of our revenue stream, we want you to take active control of some of your backups. We make it easy for you, and we keep the cost as low as possible. We strongly recommend it.
So, are you willing to spend a little bit of money to make sure you have a good backup strategy?
If so, you can talk to us (even if you aren't using our SaaS servers), we provide experts – and yes we charge by the hour for this.
In addition, if you have your own servers or are on one of our SaaS we can help you (and yes, it does cost for this service and depending on what you need it may cost on an ongoing basis) to setup what you want if you want more than what we provide in the base monthly package price you are paying.
But whether you pay us to help you or not (and you shouldn't need as much now that you have read this document) please treat your backup strategy and tests of backups as a very important part of your computer systems – not just the Maintenance Connection ones.
As I said more or less at the top:
- Computers make it cheap and easy to backup
- Most people don't take advantage of it
- If you don't computers make it easy to lose EVERYTHING.