That was the scenario backup expert W. Curtis Preston laid out this week in his New York Storage Decisions conference keynote, "Stop backing up, but restore instantly," for hundreds of storage administrators.
Preston, founder of Truth in IT and backupcentral.com, described how "nonbackup" data protection methods are the future. "All the backup methods I talk about today don't have a traditional restore," he said.
Preston said the data protection status quo no longer works because RAID rebuilds take too long, data windows can't handle all the data for many organizations and restoring takes too much time to put the data back.
A big problem with RAID today is that it takes days, if not weeks, to restore hard drives that are already at 4 TB and growing.
"Seagate just announced it will have a 5 TB drive next year," Preston said. "All I can say is, 'Holy crap.'"
Preston said there is too much data on a server, too many files in a file system and too many virtual machines for traditional backup to work well. Backup systems can spend more time finding files than backing them up, and there's not enough time to restore millions of files.
"We're starting to look at exabyte data centers," Preston said. "How do you handle that?"
These problems often make it impossible to meet recovery time objectives (RTOs) and recovery point objectives (RPOs). He suggests erasure coding as a superior way to protect data. He also told the audience to change the conversation around data protection to focus on RPO (how much data can you afford to lose) and RTO (how fast do you need to get the data back).
"If you spend all your time talking about backup windows and if you don't use the terms RTO and RPO in everyday backup conversations, something is wrong and you need to change that," he said.
He said one of the best ways to deal with short RTOs and RPOs is self-healing object storage -- which uses erasure coding -- as well as LTFS and continuous incremental backups using CDP and near-CDP software that allow instant recovery.
Object storage, which breaks data into chunks and usually includes associated metadata, can be protected with erasure coding that allows for the failure of several storage nodes without losing data.
"Erasure coding is a beautiful thing," Preston said. "Erasure coding is what we should've done when we did RAID."
Preston described CDP as replication with rollback and said it can reduce RPO to zero and RTO to anywhere from seconds to minutes. Near-CDP, which is snapshots with replication, can reduce RTOs from seconds to minutes and RPOs of around an hour.
"Stop restoring," he said, "and recover instantly."