Gimme Fast Fiber!
RPO and RTO, what are they, and how do they relate to protecting your data? Many think RPO and RTO have an inverse relationship. That’s not necesssarily true, cost should also be a factor.
It’s important to think about RPOs and RTOs, especially when writing your SLAs or “Service Level Agreements” for customers to assure your data recovery methods will match their expectations. Why and how does RPO affect RTO and vice-versa? Is one more important than the other? Does one cost more? Can I have the best levels of both? Read on…
RPO – Recovery Point Objective
RPO represents the point or time that you want to recover your data from. The gap between RPO and current time is how much data you are willing to lose from an elapsed time perspective. For example if your RPO is 15 minutes, you are willing to lose up to 15 minutes of data. This means you will need a method to backup or replicate your data every 15 minutes. A more common backup RPO might be each night or “from midnight”. This may be a less desirable RPO, but could help a lot with improving RTO.
RTO – Recovery Time Objective
RTO is how much time it takes to achieve the recovery and meet the RPO. In real-time replication scenarios or continuous protection servers, the RPO is great (1 minute or less) but it may take longer to restore from it. Why? Because the backup or replication technology is not quiescing the data or waiting for idle time since that would undermine its RPO. As a result you may have very recent data, but it may be corrupt or in a unrestorable state. This means it will take longer to recover and fix these problems, and you will have a longer RTO.
Cost
Let’s not forget the third factor – COST. For a price, you can have a good RPO and RTO. The more money you spend, the better the technology gets. However if a file is always open and being changed, its very difficult to get a stable version of that file, but you can get really close. Database mirroring is one flavor, and although it’s built into the database products, to get it to work well with a short RTO you need to run in HA (high-availability) mode, which over a WAN means spending big bucks for fast, redundant WAN links.
The best way is to achieve short RTOs with traditional backups, snapshots, copies, during a quiet time and add a near-time replication solution for your most critical data.
-Mike
(originally published in 2008)
