Google announced synchronous replication for Google apps. Of course, Google has a multi-million dollar marketing budget – so they didn’t announce replication, they announced “backup and disaster recovery” for Google apps and used terms such as RPO and RTO.
When you’ve been around a while, you see that the technology business cycles. One of my favorite services is Gartner’s “hype cycle” various analyses, simply because of the name (note: Gartner’s hype cycles aren’t really cyclical, but they are pretty useful for technology strategists.) In the mid-1990’s, cloud computing was going to be the end-all and be-all with computing delivered as a service. Cloud-based computing has gotten hot and cold since then – and I’m hoping it has more staying power this time. Actually, I’m more than hoping – the company in which I work is investing a lot in cloud-based services.
In any case, I think that the Google announcement should help with a major issue that people have with the cloud – the fact that their data is “out of their reach and control” in the case of an issue. Now – the truth is that there are far fewer “disasters” on an on-premise basis than there are “disasters” on a cloud architectural basis – so this will only go so far – but it’s certainly a step in the right direction.
However, what I really want from Google apps is a true disaster recovery service that handles my concern with errors and attacks on the Google infrastructure overall. What that means is that I want an on-premise appliance with DR handled in the cloud. That way I can get to my data regardless of the status of the cloud. That’s the reason I’m such a believer in on-premise/off-premise backup and disaster recovery architectures.
Talking about RPO and RTO (recovery point and recovery time objectives), as Google does, is wonderful – but I noticed that the RPO/RTO discussion doesn’t take into account either logical failures or the fundamental availability of the data when Internet disruptions take place.
In any case, the Google announcement – as you’d expect – got a ton of press. Some of the better analysis I saw can be found at the URLs below: