Reliable storage solution, need advice

SaleB

Distinguished
Aug 19, 2012
19
0
18,510
Hello,

I have a big data library. At the moment the library contains three 2tb Samsung 204UI, Seagate ST3000DM001 3tb, Seagate ST4000DM001 4tb, WD Green EZRX 4tb and WD Red EFRX 4tb. All of it is divided in two computers. One is my main pc and the other is there just for storing the drives. The system is working without redundancy or backup. In last eight years the only protection I have used is changing the drives before they get three years old. I had a few close calls over the years, but in most cases I did not use any data (SMART saved me). There was also one occurrence of bad sata cable that lead to destruction of partition table (in that case I retrieved about 90% of data on that disk).

I have considered some redundancy many times in past, but most reliable solutions are too expensive for me. I have given much thought about ZFS. At some moment I wanted to take an ASRock 12xsata board with Avoton C2550, 32 GB ECC RAM and make two Z-Raid 2 arrays, with an intention to add a third one and a LSI controller at some moment. Then I learned that is any array in a pool dies, you lose the whole pool. So, I concluded that it is not very reliable and very expensive.

Recently I have lost one drive again (no data loss) so I am again thinking how to add some redundancy. In recent months some bigger drives became available and the presumption that 10TB consumer class hard drive may become available during the next year simplify some things. I have been thinking about NAS devices, but the simple truth is that it is cheaper to make a NAS from a PC, learning curve can be steep, but performance is much better. I have been also thinking about some simple raid solution, e.g. MDADM Raid 1. There is also a question, is ECC RAM important to MDADM in the sam way as it is important to ZFS? I know, ZFS functions almost entirely in RAM and ECC is essential. How much is ECC important for a MDADM? I also asume that MDADM needs much less RAM, am I right? How likely is it for MDADM array to disintegrate (to be lost)?

I have been thinking about dedicating a machine for RAID and the buying drives and adding a data drive and an empty drive as new array in the RAID machine until I have RAID 1 of every disk. In this way I can do it sequentially over a year and will not be forced to buy six or ten drives at one time. Can that be done?

If you have some better solution for my situation, tell me please.

Some other information that may be important: There is about 4tb of very important data. The increase is about 1tb/yr. There is 16tb of not so important data with increase of about 4tb/yr. I am most of the time a sole user, mostly from inside the network, sometimes from outside. There are other occasional users, 4-5 of them, mostly not at the same time. They are using some media streaming most of the time (music, video). I would like to create a solution that will keep up with data increase for some time, by the means of only adding drives and a controller when needed.

I am sorry for a longer post, but I have tried to be as detail oriented as possible.

Thank you
 
Solution
No data solution is 100%. You will need a combination of on site and off site storage to try and keep this all.

For the 4 TB of important data, I would use an external USB hard drive such as the WD my passport attached to the desktop. I would also recommend a NAS.

For the 16 TB, go with a home built NAS.

Also, you should get offsite backup. There are 100 TB plans with an annual fee.
No data solution is 100%. You will need a combination of on site and off site storage to try and keep this all.

For the 4 TB of important data, I would use an external USB hard drive such as the WD my passport attached to the desktop. I would also recommend a NAS.

For the 16 TB, go with a home built NAS.

Also, you should get offsite backup. There are 100 TB plans with an annual fee.
 
Solution