What’s the cost to your organisation if your systems aren’t running?

What’s the cost to your organization if your systems aren’t running? Today, we’re going to delve into high availability and disaster recovery, crucial topics surrounding this question. Depending on the value of a system and the cost implications of downtime, these factors justify the investment in a robust disaster recovery strategy. Let’s explore this issue further. Ultimately, in today’s cloud-based infrastructure, the cloud is essentially someone else’s computers. Your software or application runs on one or more of these computers, located in data centers across various regions. When we look at this as a whole, the first thing we want to avoid is having any single points of failure, because if there’s just one component your application depends on and it fails, you’ve lost it. Therefore, we always ensure redundancy. For example, in our applications, we create… That’s why Web Farms have pools of web servers. If one fails, the others can take over.

Similarly, in an Azure region, a region may have multiple data centers. You’ve got redundancy between data centers. Additionally, some of the services we use, such as storage services, can be zone redundant. This means that when we instruct the system to store something, it doesn’t just save it on a computer but across multiple computers in several physically separate data centers. This way, if one data center loses power or connectivity, it doesn’t matter; your data is still safe elsewhere. But what if an entire region fails? Then, you might decide you want your data in more than one region, which complicates matters. This approach is reserved for your most critical applications. For instance, if we’ve got UK South, that has three data centers in it. But if you say, “Okay, I’ll fail over to UK West,” that’s when they get warmed up. Not everything works the same way across different regions.

So, we need to account for that. But also, if we consider the traffic coming in, somehow you have to balance that traffic so that it can go to this region or that region, or both. Depending on what you need to do, there are different strategies. Do you need a cold standby? In other words, if this goes down, can you bring this up from cold and then start working with it? Is that alright? Or do you need a warm standby? You need to have this ready to go, but it’s just not being used right now. If you pull the switch, it’s there. Or is it a hot standby? You’re using both of them all the time because then you know that no matter what happens, everything is there immediately. That’s your really critical processes. That’s the stuff that’s really customer-impacting or cost-impacting. That’s often expressed in something called RTO, recovery time objective. How long does it take me to recover? Do I need to recover in a millisecond, or is sometime tomorrow afternoon okay? There’s a big impact on cost on these, and it depends on how valuable that is.

Similarly, there’s something we call RPO, or recovery point objective. In other words, if this goes down, how much data do I lose? Do I lose the last millisecond of data, or do I lose the last week? It’s like if you do weekly backups and they’re sitting on the same server, and once a week, you take the tapes out, you could lose a whole week. That’s not good. The same principle applies to the cloud. How much data are you willing to lose in your recovery point objective? And if that is zero, then you’ve got to pay close attention to how you ensure that data is replicating in real-time between different locations. And that has performance and cost implications. So, as a general overview of high availability in disaster recovery, big things to bear in mind are, have multiple things in multiple data centers or multiple regions. The hotter your standby, the less time it takes to recover, and the less data you lose, the costs are likely to go up. So make sure you invest in the things appropriate for what that value is to your business.