Introduction
Disks fail, data centers go offline, but with Zadara’s new incremental backup to Amazon S3 (B2OS) solution, you can protect and archive your data, and then restore to any location as a new volume or block device. In this post, I cover our new B2OS capability and three other methods you can use to protect your data.
How VPSA Data Protection Works
Data protection services are just one of many features included in your Virtual Private Storage Array™ (VPSA™) to help with your day-to-day and long-term data management. A common thread to these services is the snapshot functionality provided with both block volumes and NAS shares.
Snapshots use a copy-on-write technology where after a snapshot is taken, data that is modified is copied to different blocks within the storage device so that the original data is preserved.
You define a policy of how fast you want to take the snapshot and any data written thereafter is marked “dirty” in the volume’s virtual segment table. When the next snapshot is scheduled, the original segment table is saved and blocks that were marked dirty are reassigned to unused blocks within the pool.
Very little space is required for the snapshot other than storing the segment table and metadata. Your data is never physically copied, making this a very efficient incremental backup process.
Pool Level Protection
When you clone a volume, a copy of the metadata from the snapshot is used to create the clone. The clone starts out as an identical image of the original because the data is shared between the original and clone. Any modifications to the original or clone use copy-on-write for completely isolated, yet shared data sets.
This level of protection works great if users delete files or a virus is introduced into your data, since all you have to do is clone the volume from a point in time prior to the issue. The problem is then fixed by restoring the data or uninfected programs.
Please remember that your backups and data coexist on the same media. That is, pool level protection is designed to survive a single RAID-1 drive or two RAID-6 drives simultaneously failing within the same RAID group. If you require higher levels of data protection, we offer local and remote mirroring.
Local and Remote Mirroring
Two other protection levels that use snapshots are mirroring to local and remote pools. For mirroring, your volume, represented by a snapshot, is copied block-by-block to a different pool. Although the pool resides on the same VPSA, the backups are on drives independent of your data. A common use for this is to use SSDs for production and 5- or 6TB SATA drives locally for mirroring. This provides an additional layer of protection and allows you to instantly access your data, without a restoration step.
However, if the data center goes down, you will not have access to your data. With remote mirroring, you have the same protection as local mirroring, but your data is offsite in a different data center or region.
There were several instances where AWS was offline as long as 14 hours, causing outages for their customers. For disaster recovery (DR), you can mirror your data to another VPSA in a different AWS region.
For example, we have clouds in both the East and West coast where a mirror would be configured between the two. If your production site goes down in the East, you can start operations back up in the West coast region using the latest snapshot.
Both local and remote mirroring are designed for rapid data and site recovery. But what if you have a requirement to archive your data for years? Our new Backup to S3 feature offers such a solution.
Backup to S3
With Backup to S3 (B2OS), snapshots are mirrored to AWS. All the metadata and data is stored compressed in S3 buckets where it can later be restored to any VPSA or copied to a block device. Since you can restore the data as a backup image, your data is vendor-neutral, allowing quick migration to other devices. S3 also has an aging feature that automatically moves data to Glacier after a preset period of time. This means your B2OS data can eventually (or instantly, if you set the aging parameter to zero) go to Glacier, for economical archiving.
Depending on the size of your volumes, this feature may be more economical than using local or remote mirroring; however, DR may take some time to configure if you do not have the pre-allocated resources up and running in the event of a disaster. And if you choose to age data to Glacier, keep in mind that Glacier restoration takes 4-5 hours.
A really cool feature is that when you restore your volume, you can restore it as a clone. This is a zero copy backup restore. When your servers access the blocks in a volume, a page fault is generated for missing data causing blocks to be copied from S3 on demand. This works wonders if you only need a few files from years ago and don’t want to wait for the entire restore job to complete.
Backup Type | Use Case | Considerations |
Pool | User deleted files, virus, and application errors. Short-term temporary backups. | Data and backups share same disks. Old backups are removed when pool space is limited. Not designed for DR.
|
Local Mirroring | Short and long-term archives. Additional layer of protection of data local to VPSA. | Not designed for DR solution. VPSA drive limitation reduces maximum production storage.
|
Remote Mirroring | Short and long-term archives, disaster recovery. Additional layer of protection remote from VPSA.
|
Requires second VPSA. |
Backup to S3 | Short, long and cold storage archives, disaster recovery; data compressed. Resides on third-party site.
|
DR response time slower. |
The table above summarizes a few use cases you may want to consider implementing when looking to protect your data on your VPSA. If you’re not using a minimum of local mirroring for data protection, you should be scheduling backup jobs from your servers using a folder/file copy or image backup. Otherwise, your data access may be at risk for reasons unrelated to your reliable, redundant VPSA.
Download the B2OS Data Sheet
Learn more about B2OS in our latest data sheet. Click here to download.