A few weeks ago we ran into a client who stated that he needed to achieve the holy grail of system uptime, commonly known as the “five 9s.” This descriptive phrase is derived from the figure of 99.999 percent. If your goal is five 9s, this means that your system stays up and running 99.999 percent of the time. Ultimately, though, this client had to settle for something less as five 9s was outside of his budget.
In this particular instance, we were dealing with a relatively small database, running on Microsoft’s SQL Server database software. To get to five 9s on SQL Server requires implementing a solution called AlwaysOn Availability Groups. AlwaysOn requires multiple Windows servers, configured as clusters. Windows servers, of course, require licenses. Furthermore, the hardware upon which the servers are built need to be fairly robust, even if virtualized with something like VMware or Hyper-V.
But what really drives up the cost is that AlwaysOn requires an Enterprise license of SQL Server for each database server. When you factor in the need for at least two database servers, the cost increases rather dramatically.
While we focus on SQL Server in this narrative, these principles hold true for all vendors’ products, including Oracle, IBM and even open-source offerings. We’re not just picking on Microsoft here, as popular as that might be. Every vendor requires multiple servers, good hardware and top-of-the line database licenses to implement such a solution.
It’s not just the software and hardware costs that need to be considered. Such solutions, regardless of vendor, are complex, which means a lot more moving parts, which means a lot more things to go wrong. Even with highly skilled and trained personnel, the chance of a misconfiguration or overlooked setting is high.
Not only that, but regular maintenance of such solutions is also a requirement. Logs need to be reviewed periodically, and alerts need to be set up to send emails or texts when anomalies crop up, and they always do.
While five 9s sounds cool, a more pragmatic approach is to first determine how much downtime your organization can stand. After all, in a month, five 9s allows for only 26 seconds of downtime. Is that really what you need?
In theory, the cost to maintain uptime should be less than the cost of an outage. Don’t spend $10,000 to prevent an outage if the downtime costs you only $1,000. Of course, this is an oversimplification of the equation, but don’t let the basics of the equation get away.
John Agsalud is an IT expert with more than 25 years of infor- mation technology experience in Hawaii and around the world. He can be reached at johnagsalud@yahoo.com.
