I believe the industry needs to formulate what defines backup and what are the essential elements of a modern backup solution! How will we be able to address the ever shrinking “backup window,” exponential data growth, and the resource-optimized infrastructure as a result of virtualization? Most of all, how are we going to map backup and recovery processes to business objectives? The definition of this space is becoming more pertinent than ever. Just last week, FalconStor CSO Jim McNiel was discussing data protection and recovery in TheCube at VMworld 2010 in San Francisco; in addition, Kent Langley, CTO of cloud service provider nScaled talked about RPO and RTO requirements for backup services in the cloud.
But before jumping on a really slippery and very long slope all the way from the top, I’ll stop and get back to the subject of today’s blog. Why snapshots for backup? I would think that we all agree that by using the snapshot approach you eliminate the backup window – which happens to be the biggest pain of traditional backup – so now you can take more frequent backups without greatly affecting production systems. That gives us more recovery points, thereby improving recovery point objectives (RPO) tremendously. On the other hand, Curtis talks about clean snapshots that would improve recovery time objectives (RTO) – a clean snapshot transforms recovery from a restore process to a simple mount that takes minutes instead of hours.
I also agree that not all snapshots are created equal. In most cases, snapshots were designed primarily for data protection, but were not necessarily optimized for data recovery. This is why snapshots are designed to operate like a traditional backup process, with a base copy and incremental changes; therefore, to recover from the latest snapshot – which would be a natural thing to do in a case of data loss or a system failure – you would have to factor in the calculation of the deltas back to the base and then mount the disk image. Curtis gives the two most common examples of traditional snapshot methodology, copy-on-write and redirect-on-write. However, at FalconStor we use a different, patented approach that is referred to as “reverse-delta snapshots,” where the image of the disk is always current and the last snapshot is always ready to be mounted.
As I heard Curtis say before “it’s all about recovery stupid,” since the primary goal of any backup product is to be able to recover and to recover fast.
When FalconStor’s founders were at Cheyenne, they designed backup to work on the file server because that is where the data was at that time – it was a simple matter of efficient operations. Today data lives in the SAN or NAS fabric, and this is where the backup processes should take place for the exact same reason. But using snapshots as a backup alternative needs some backup backbone, and I’m not referring to the company here!