Storage Switzerland works with IT professionals to design data protection and disaster recovery plans. Those conversations always start with a discussion about service level objectives (SLO). Each application or data set should have an SLO, that comes from several very specific objectives that include Recovery Time Objective (RTO), Recovery Point Objectives (RPO), Version Retention Objective (VRO) and Geographic Recovery Objective (GRO). As organizational and user expectations continue to increase, we are adding a fifth objective to the SLO matrix; The Recovery Performance Expectation (RPE).
What is RPE?
A RPE is an articulation of what the user can expect in terms of performance of their application in the recovered state. For example, consider the common scenario of the combination of a purpose-built backup appliance and backup that can perform recovery in-place. If that appliance leverages deduplication and compression to drive cost down, the performance of that setup is likely to be nowhere near the performance of the primary storage system – especially if that primary storage system is flash-assisted in some way. This performance gap means the users may be very disappointed with our recovery efforts no matter how fast we return the application to service.
Like RTO and RPO, determining your RPE is a give and take with the business units. They will start with an unreasonable performance expectation, just as they tend to ask for an RTO and RPO of 0. You can start the process of determining your RPE by articulating what performance will be during a failure of the primary storage system – using your current recovery system. If they are happy with the RPE then the process is done. If not, you need to create alternatives so performance will meet expectations or so they can adapt expectations to the realities of the budget.
Either way, discussing the RPE up front helps make sure that users are not complaining about performance in the middle of a disaster, distracting IT from concentrating on getting the organization through the disaster.
RPE and the Cloud
The RPE is also critical as more organizations take advantage of the cloud, leveraging services like DRaaS (disaster recovery as a service). While most DRaaS providers articulate how long it will be to get you back up and running after disaster declaration, most do not discuss what your performance will look like during the disaster. Remember, you may be in the cloud for months while you rebuild your new data center, so understanding the performance is a critical conversation to have. The RPE forces that conversation.
Our SLO strategy is all about setting expectations, and making sure that you can hit them. It is not about buying new software or adopting the latest technology, although sometimes that is a necessary outcome. SLOs are also easier to adapt to the changing reality of the data center, RPEs are the latest example of that.
Did you know that Storage Switzerland writes a dozen SLO plans a year for organizations at no charge? We do that to make sure that what we are teaching in our workshops actually works in the real world. Contact us if you’d like us to work with you on such a plan. But know that if you attend one of our workshops, what we teach is battle ready.