Cloud pricing can be a frustrating experience. Everything is charged by different metrics. Some of the prices are spelled out, some are hidden behind paywalls or aren’t clear until you get your monthly bill and realize you forgot to turn off an instance that is chewing up your wallet. Some are charged by usage, others by the month. All this means that keeping track of cloud costs can be nearly a full-time job.
Thankfully, there is software that can help predict your bill and figure out if you would be better off moving your VMs to some other provider. Indeed, there is a lot of choice of these cloud costing comparison services. Keep in mind that these tools are used to estimate your monthly cloud bills and don’t really attempt to get much into the actual application performance monitoring itself. For that you will need other tools such as ManageEngine’s Applications Monitor, Scalr (which offers some costing and management features) or DataDog, just to name a few.
Before you use these services, first look at the individual costing pages that each cloud provider offers for free: Here they are from Rackspace, Amazon, Azure, Google, andIBM/Softlayer (the latter has registration annoyingly required). The issue with using these free costing pages is that they can be more confusing than helpful, because they assume you know your way around specifying the exact instance size and parameters. For example, the simple question of how many compute hours are included in your monthly bill isn’t as easy as it seems. Some of these calculators assume a 720-hour month, no matter what the calendar says.
Second, look at some of the free services that are available from third parties that offer comparison reports of the cloud providers. For example one of CloudSpectator’s reports show that Microsoft Azure provides lowest cost block storage, while SoftLayer is the least costly for large Windows instances. Download these reports and study them before you pony up any hard cash for the paid services. If your cloud computing bill is less than $1000 a month you are probably not going to get any further as most of the paid services are for helping you manage larger and more costly cloud installations.
Figuring out the price tag for these costing services is also an exercise in patience. Some of them are completely free, such as Solarwinds and Cloud Harmony. Some have freemium services: you get a small collection of providers for free, more in depth reports and analyses as you pay. For example, Cloudyn has a free service that has limited analysis, a paid service at $229 a month, and will also sell you custom reports that cost and cover more.
Third, figure out whether you should you go wide (with looking at a service that can handle as many players as possible) or deep (diving into just a few cloud providers and examine a lot of their particular service offerings). No one service can do both well, sadly.
Finally, check out some typical questions that you might want to have answered by each costing service or through your own research:
- Who is the cheapest cloud provider for a particular workload size and duration? Most of these services can answer this question adequately. When you add other cloud services beyond CPU, disk, and memory though it can be difficult to compare one provider with another.
- What you can save by using AWS reserved instances (and some of the others) to cut costs if you plan ahead. Both Cloudability and Plan for Cloud have special tools to help with your planning here. CloudHealth also has an item in their reports that show the impact of reserved instances on your overall bill.
- Should you be using burstable CPUs? This is another way to cut costs. Some of the players will let you overcommit your workload for a limited time, and some do so for no additional charge. While none of these costing services can do the math, it is something to keep in mind.
- Should you distribute your workloads? It might make sense to split your workload into smaller instances that are cheaper to run, in some cases significantly cheaper. Cloudorado has the ability to perform this kind of sensitivity analysis. One nice feature about PlanForCloud is that it allows you to assemble deployment scenarios by mixing and matching across their preferred six providers. That is an interesting twist that none of the other costing services offer.
- How much support do you really need? The level of support that you pay for could be costing you money. However, you may not actually know what you need until you have been using a particular platform for a couple of months, so take a closer look down the road and see if you can reduce your support bill. Both PlanForCloud and Cloudorado (to a lesser extent) offer ways you can analyze the effects of reducing your support plan if you don’t need all the hand-holding.
One other thought: the cloud is a dynamic place. Instance characteristics come and go. Each player changes prices almost continuously: AWS has been through dozens of them, mostly drops in price over the years. Any costing analysis that you do perform could be outdated in a few months’ time, so do run it repeatedly to check if anything has changed.