As you move workloads to the cloud, you’ve probably already discovered that keeping track of your monthly cloud computing bills is not an easy task. Certainly, using a cloud provider can be cheaper than purchasing your own hardware, or instrumental in moving a capital expense into an operating one. And there are impressive multi-core hyperscale servers that are now available to anyone for a reasonable monthly fee.
But while it is great that cloud providers base their fees on what resources you actually consume, the various elements of your bill are daunting and complex, to say the least. True, many of the cloud components can be had for pennies per month, but for some resources (such as those heavily loaded multicore CPUs) the meter can run up to real money very quickly.
One of the ways the cloud is evolving is in how these fees are calculated. Some vendors, such as Amazon Web Services, have blue-light specials where if you can plan in advance you can cut your costs significantly. Others, such as IBM’s SoftLayer, offer freebies like inbound and outbound data transfers. (See: Cloud price wars give way to feature battles among Amazon, Microsoft and Google)
And all of the vendors are cutting their prices regularly. According to one vendor, AWS EC2 has had 44 price reductions since its inception in 2006. And that is just for one of their several dozen offerings, each of which has had its own price reductions. Keeping up with these constant changes can be tough.
Certainly, the cloud providers are trying to help with pricing pages on their own websites. But because each provider charges separately for each individual storage unit, CPU core, amount of RAM, disk space, and data transfer, these calculations can be hard to predict with any accuracy, even if you know ahead of time the parameters of your particular instance.
+ ALSO ON NETWORK WORLD Cloud service providers must guarantee zero down time +
AWS has so many different services that they have a link to what they call their economics center where customers can explore how to cut monthly bills and optimize configurations. AWS is probably the most transparent. With some providers, you need to sign up for their services (or at least create an account with a valid email address) before they give you access to their pricing pages.
Added to this is that a virtual CPU isn’t the same across service providers. Cloudorado has written its own vCPU benchmarks, which is a good start, but customers should have more tools like it to enable easier cross-vendor comparison shopping.
And then each provider has its own marketing spin, such as this recent blog post on Google’s site where they compare their cloud offerings to AWS. No surprise, they are cheaper. So, separating fact from fiction isn’t easy.
For this article, we looked at five shopping comparison services, including Cloudorado, CloudHarmony’s CloudSquare, CloudSpectator, Datapipe and RightScale’s PlanForCloud.com. Some of them cover a lot of providers, some only focus on a few.
Each of these services brings something unique and interesting to the table. The services either go wide or deep but generally don’t do both, so you’ll probably want to make use of at least two or more. Most have free reports or cost prediction tools to help in your own analysis. You should definitely download these free reports first and review them carefully before contemplating paying for any of the services, because you can get a lot for free if you spend the time.
However, if you have less than a $1,000 monthly bill from your cloud provider, you probably aren’t going to like any of the paid services since fees are fairly steep and are designed for large enterprises.
Given that there are close to 200 cloud providers it is easy to see that no single comparison service will be able to ferret out details on everyone’s offerings.
- If you are looking to find a smaller or lesser-known cloud provider, then certainly start with those services that compare more of them such as CloudHarmony and then focus in on the three or four players that look the best for your particular circumstances.
- If you are shopping for a particular set of technical traits, such as a provider that makes use of burstable CPUs, then Cloudorado can quickly narrow the field within a few seconds.
- If you want a convenient comparison of a few support plans, then PlanForCloud has a nice page summarizing that.
- If you are just using AWS and you want more insight into your bill and don’t mind paying for the analysis, then take a look at what Datapipe can provide.
- And CloudSpectator will help you determine how much it would cost to move internal operations to the cloud.
Here are the individual reviews:
Cloudorado Cloud Hosting Comparison
Cloudorado looks at 26 cloud providers, which cover a nice mix of both the majors and up-and-coming players, although one striking omission is IBM SoftLayer. Providers pay to be listed on their service so you aren’t charged any fees. If you look through their analysis, you will see that Google, Atlantic.Net and M5 pop up as the least costly providers more often than the others. The site is largely the effort of Marcin Okraszewski based in Gdansk, Poland, who has normalized all the CPU ratings across providers so you can make better and more meaningful comparisons and have some assurance that the workloads you are pricing are equivalent. So even though it is a pay-for-play site, it is worth examining for his insights and side-by-side comparisons.
Another nice feature of this site is that you can calculate costs by distributing a workload across different RAM and CPU configurations. In some cases you can cut your bills significantly if you distribute things in this fashion by renting smaller VM instances. And you can also specify a particular geographic location, along with querying features such as a mobile user interface or having a single consolidated account.
Cloudorado also breaks down each cloud computing provider’s offer into more than 130 distinct features. That is a nice summary of all sorts of technical details in one place, such as whether they support the OpenStack and AWS APIs or can mount particular ISO images. It has a nicely designed user interface that is very straightforward to use. Overall, the site has a nice balance between enough cloud providers and enough details to be a great starting place in your cost comparison journey.
CloudHarmony’s CloudSquare
CloudHarmony is based in Laguna Beach, Calif., and offers a variety of paid consulting services. CloudSquare is a series of free interactive reports that look at 95 providers, which was the largest provider collection of the six sites we examined. You drag and drop the icons from the provider list and can quickly see the specifics of up to six different offerings: compute, storage, CDN, DNS, their overall PaaS and database services. You can filter the overall list by specific geography (for example, 40 out of the 98 providers offer services in both North America and Asia).
There is a second series of comparisons for uptime status that is generated by Panopta for each of the providers. You specify the time period to see how they performed. For example, in the past month India-based Tata was the worst compute-based provider, with four hours of downtime and 40 different outages. Many of the providers had 100% uptime, which is nice to see.
With each provider is a link to their uptime status reports for each of their offerings, which is handy and comprehensive: each data center (AWS has 28, according to their CloudHarmony’s report) is presented with its own uptime statistics. If you click on the provider name, you will see a more comprehensive examination of the various metrics collected. So we could see that the AWS US-west and European sites had 100% uptime over the past year, for example.
Speaking of uptime, we found a lot of 404 errors (meaning their own servers were down) early one Sunday morning when we were investigating their site. Other times the service functioned as expected.
In addition to the interactive comparison charts, CloudHarmony also produces a regular series of in-depth reports on various items, such as managed DNS or compute services. These are done for a particular moment in time but are very comprehensive and are worth examining if you are going to get more involved in a particular provider or if your current provider isn’t delivering what they promised in terms of these services. Here is a link to their most recent report on the “state of the cloud” from July 2014.
CloudSpectator
CloudSpectator, which is based in Boston, is in the business of selling their analysis to enterprises and providers. They will benchmark internal systems to help an IT department figure out the total cost to move into the cloud.
They also have two free offerings. One is a series of reports about cloud costs, performance, and other data, similar to CloudSquare. You have to set up an account to have access to these reports.
Their latest report is dated February 2015 and compares Azure A and D series instances. That is pretty specific, but others are more general and there is a January 2015 report than compares overall costs for 10 providers.
Again, the downside is that this is just a picture in a moment of time in a very fast-changing industry. The upside is that they have lots more detail about pricing for the various cloud providers that they do cover. The January 2015 report runs for 47 pages as an example. They also perform their own sequence of tests to try to make comparisons more rigorous.
One interesting result from this last report is that Microsoft Azure provides lowest cost block storage, while SoftLayer is the least costly for large Windows instances. Like Cloudorado, they try to normalize the variables to make the comparisons more meaningful, so you don’t have to pull your hair out trying to do this.
They also have a second offering that provides interactive reports on seven providers: Digital Ocean, AWS, Rackspace, Google Cloud, Century Link, Profhost, and Microsoft Azure. Not all of these providers are current clients of CloudSpectator. When you access these interactive reports, you select which of these providers you want to see the benchmarks run, and then the site produces a series of reports that show response time, input/output operations, and other statistics.
All this information is presented graphically, which makes for pretty charts but not much in the analysis department.
They will have a third offering, called Selector, that combines analysis and performance data, which is similar to some of their competitors. This is scheduled for the summer and will cover at least 15 providers.
Datapipe Cloud Analytics for AWS
Datapipe has a number of their own service offerings in the managed hosting space and they include a cloud costing service as part of their overall management tool. Basically, you allow them to manage your AWS accounts and they will provide a raft of analytics that you can use to see what your instances are doing. This is designed for larger environments with fees starting at $3,500 a month. Datapipe takes a small percentage of your overall AWS bill as part of its fee.
Datapipe is using technology it bought from Newvem several years ago; earlier this year they acquired cloud provider GoGrid. They are based in Jersey City, N.J., with data centers all over the world.
While their Cloud Analytics for AWS is very specialized: as you might have guessed, it is just for AWS. They also have a similar offering for Azure, and since the GoGrid acquisition are thinking about how to leverage those customers too. One product that we would like to see is analytics that show you how much you save or spend if you switch instances from AWS to Go Grid or Azure.
Once you connect the Datapipe tool to your AWS account, it starts reviewing your configuration for the various AWS offerings. You can visualize historical trends, and there’s an interesting heat map to help with capacity planning decisions.
+ ALSO: Cloud price cuts are nice but enterprises still want more +
It shows you changes in your running instances over time, which is unique among the various cloud costing services.
There is also a unique security review of your account. Datapipe will also help you figure out the reserved instances that you should purchase, based on your past account history. You can consolidate multiple AWS accounts into one dashboard, provided that Datapipe manages all of them.
RightScale’s PlanForCloud.com
RightScale has its own cloud management and analytics business that it sells under the RightScale Cloud Analytics bundle of tools. IT has a free 60-day trial, but otherwise this is geared towards larger enterprises, with fees starting at $6,000 per month. Their PlanForCloud service, which is based in Scotland, is a free offshoot and leverages some of the same information that the company collects for its analytics platform.
You start with assembling various deployment scenarios so you can do some sensitivity analysis about your assumptions. They support just six cloud providers: AWS, Rackspace, Azure, Google Cloud, SoftLayer and HP’s cloud. You can login as guest and poke around with a sample three-tier Web app scenario.
The deployment scenario starts with a model where you purchase a particular virtual server, attach resources such as RAM and CPUs, and then specify a growth pattern. There are five other tabs as part of your scenario: you add storage repositories, databases, data transfers in and out to your cloud, a particular support plan, and various miscellaneous costs.
The support plan tab is the most noteworthy, again a nice summary of what you are going to be dealing with if you have to purchase extra support. This only shows AWS, Google, HP and Rackspace.
If you want more, you can create a free account, which enables you to save your scenarios so you can access them later, and also specify non-US dollar currencies to display your cost analysis.
Once you add your server from their list of what instances are available, it can produce a three-year cost report, showing you what metrics contribute to the costs and more importantly, how these costs vary from month to month. This is one thing that none of the other costing services has.
Finally, you can change the plans that you pick from each provider to a reserved instance, when it is available, and see how your costs change.
Like with Datapipe, you login to your AWS (or RightScale) account and can import your configuration directly into their tool.
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 can offer.
While their comparison tables are a bit simplistic, you can find out quickly who has the least costly instance for particular configurations, such as HP’s “double extra large” offering with its 32 virtual CPUs at 90 cents an hour, lower than any AWS instance.
Strom is the founding editor-in-chief of Network Computing magazine and has written thousands of magazine articles and two books on various IT and networking topics. His blog can be found at strominator.com and you can follow him on Twitter @dstrom. He lives in St. Louis.
