Backup Best Practices 2

[mfc133]

Does anyone know of any good resources (website, whitepaper, etc.) for best practices regarding backup procedures. Mostly having to do with when to choose Full/Incremental/Differential, how often to take media offsite, when to archive data, etc. I have a pretty good idea of what needs to be done, but my basically manager wants to see best practices/documentation that says "in this case, do exactly this". I don't really agree, but unfortunately I've been tasked with finding this stuff.

Any ideas?

Thanks

 

[comtec17]

The term "Best Practices" is taken out of context and is an industry used word by vendors. When a vendor states best practice, it usually means that the vendor has tested the software/hardware with certain settings that seemed to be best during a test.

It is up to you and your manager to decide what your business needs. A consulting firm, white paper, etc can only point you in a direction. Look for backup case studies on google for some hints.

Typically in the field, I see customers doing full backups 1 time per week and either differential or incremental during the remainder of the week.

The best approach is to figure out your RTO (recovery time objective) and RPO (recovery point objective) for your business critical applications, also known as a gap analysis. Then look at your bandwidth, how much data you need to backup, and your budget to figure out how much money you can spend each year on tapes and management.

 

[birky]

Comtec is correct that there are no real 'best practices' it really depends on your environment and requirements such as the RTO/RPO.

You need to look at the downtime that you could experience for each of your servers/apps without it affecting the business too much..then you can prioritise the business critical boxes before the others, maybe even looking at replication to minimise downtime.....this will help you to decide whether to use diffs/incr, the fewer boxes you use diffs on, the less media you are going to use and also a shorter backup window if you have a high data change rate.

If you have remote sites with slow links then you may look at implementing Synthetic Full backups so that you don't have to worry about doing the Fulls so often.

For archiving data it all depends on business requirements again....do you have compliance regulations that you have to adhere to? ...will users come to you in say 12 months and expect an email to be restored? (if so then you want to keep copies but, not online on disk or in the library.

Offsiting media - think about how far back your organisation could go back if you lost your site (fire/flood etc) where the onsite data is held...could the company afford to lose a few days data because you only offsite media once a week so you can only restore back to last weekends full backup?
Are your backups located in the same building as your production servers? ....if so do you have another building to relocate your backup storage to?

A lot to think about when designing an effective DP solution but, like has already been said, you have to tailor it to your requirements and continually update it by doing regular risk assessments by talking/scheduling meetings with the various depts to ensure you are meeting their requirements.

+ above all...

TEST TEST TEST TEST TEST TEST

your backed up data for restorability.

 

[mfc133]

Oh I completely agree with you guys about best practices,
unfortunately the powers-that-be like seeing the documentation, best practices, whitepapers, etc. I've been able to find a good amount of stuff on google that should hopefully keep them happy.

If my job were to actually sit down, look at what data we're backing up, how fast it needs to be restored, how long to keep it, etc. and propose a backup/archive strategy I would be set. Unfortunately there's just way too much bureaucracy here(especially for a small company). The mentality here is you have to have a printed doc that explains exactly how to do something INCASE(which in reality never happens) someone questions your decision you can say "Well I followed this 'best practice' when I did it". It's ridiculous how much politics can get in the way of getting real work done.

But anyway, what I will probably recommend to them is using a variation of the GFS method, with full backups weekly and differentials daily. Right now we're doing fulls daily for quick restore time, but it's probably overkill since about 40% of our data is written once and never changes.