2

I am a computational biologist working for a small startup tasked with setting up a local implementation of a small tower server (if it matters, I'm looking into getting a PowerEdge T620 or similar) for use as a compute system & small MySQL implementation.

My concerns lay in data security, more specifically, how I handle data backup. We would like flexibility when it comes to overall storage, however our starting point for storage is at the 6TB mark. Therefore, I'm envisioning the use of 2 x 3TB drives.

I'd like to have a strategy in which I always have a backup copy of my data (updated every evening) accessible to the server & an off-site copy (which is physically brought in and updated over the weekend).

It looks like the server has several RAID options (through their PERC H310 Integrated RAID Controller), three that seem particularly useful in my case are

  • RAID 10
  • RAID 0+RAID 0
  • RAID 1+RAID 10

My question: is it safe/advisable to swap out the two hard drives that make up the backup copy for the two that make up the off-site copy, and do that backing-up that way?

Or should I look into buying a RAID enclosure which will act as the off-site backup and connect it through something like USB3.0?

Note: I'd like to physically disconnect (at the least power down) the backup while updating the off-site (that is, I never want all drives attached to the server). Therefore, I'm wondering how feasible it is to pull drives out and swap them with others.

Secondary question: Dell's RAID 0+RAID 0 setup is described like this -

RAID 0+RAID 0 for H710P/H710/H310 (1 SATA + 1-31 SAS HDDs) add $0.00

Is it even possible to use four drives split into two RAID 0 setups?

  • Anything shopping-like is not on topic here, but I just want to mention that the H310 does not have any kind of write cache whatsoever. Your performance will suffer greatly. I highly recommend an upgrade to the H710 adapter with NV cache. – pauska Sep 29 '14 at 20:09
1

Backups are hard. And absolutely required.

As marc99 pointed out, disconnecting drives physically from the enclosure is not a safe option.

First, you need to decide what data is actually important. You mention MySQL and it's taking around 6gb. Is anything ELSE on the server valuable?

Have you looked into compression? Does that save you anything? Depending on the dataset, it might, or might not.

Have you looked into hot vs cold backups of mysql? mysqldump works well, but it's not a hot backup tool. you need to stop all write transactions.

If you've tapped out the capacity of the server running mysql, you might have to do mysqldump over an SSH connection to a different server... this may have other benefits, since it could then send the freshly completed backup off to somewhere else (cloud storage like S3 or Google Compute Storage comes to mind for this) while your 'main' server continues to work.

  • most of the data size comes from sequencing data files (standard txt). the database is filled with metadata surrounding those files; besides that, there won't be much else that i'm interested in backing up. i haven't looked into compression, no; given what I said above (in this comment), is it an option? the database is not write heavy at all. in fact, since i'm the only one that's effectively accessing it (mostly for read purposes) doing a mysqldump would be a good idea – Constantino Sep 29 '14 at 21:45
0

You should make a decision on the storage strategy based on needed capacity/performance. Backup solution should be based on time to retrieve your backup/time window for the backup to run.

About storage and performance, you should consider:

  • disk type (SATA/SAS vs SSD): capacity vs performance
  • number of disks: more disks, more performance
  • raid type

About backup strategy, swaping disks won't be the safest solution. If you have a small backup time window, snapshot is probably the best solution here. Hardware snapshot might worth it depending on the budget.

If you a seek a proper off-site solution, you should have a look at LTO library. It will allow a 6TB full backup over 4 maybe 5 cartridges. Be careful with the backup time window again.

Marc

  • our backup time isn't small (i'm the only one access this server) – Constantino Sep 29 '14 at 21:43
  • So you can just skip the snapshot part. You should aim at a server with 2 disks of 3TB or 4TB + 2 small harddrives for the operating system + 1 library (including 1 LTO drive and few cartridges). – marc99 Sep 30 '14 at 8:09
  • do you mean to say that the two small HDs for the OS would be in RAID0? – Constantino Sep 30 '14 at 14:58
  • No you should have 2 disks in RAID1 for the OS and minimum 2 big disks for the datas in RAID1. RAID0+0 doesn't mean anything to me since one RAID0 can handle more than 2 disks in a single RAID0 group. RAID0 is not an option since loosing one disk means loosing all datas. You better look at RAID5 or RAID10. You should have a look at this: [en.wikipedia.org/wiki/RAID] – marc99 Sep 30 '14 at 18:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.