This article is part of an Essential Guide, our editor-selected collection of our best articles, videos and other content on this topic. Explore more in this guide:
4. - 'Hidden' costs, pricing confusion shroud true cost benefits of cloud: Read more in this section
- Learning curve can be costly in the cloud
- Unforeseen cloud computing costs cause sticker shock
- You're charging me for what?! Picking apart your Amazon cloud bill
- Amazon cuts dedicated instance pricing, raises competition
- Amazon cloud cost-cutting tips and tricks
- How to make cloud computing costs models work for you
- Pricing for public cloud nowhere near bottom
- Public cloud ROI takes a hit from app migration snags, subscription fees
- AWS shops save with tools for analyzing cloud costs
- Experts refute 'hidden' cloud computing costs
Explore other sections in this guide:
- 1. - Follow #reInvent on Twitter
- 2. - Where are enterprises in cloud computing adoption?
- 3. - The importance of locking down your cloud
Cloud cost analysis tools can save Amazon Web Services customers thousands by helping them better use its cloud.
Inmar Inc., a Winston-Salem, N.C.-based promotion, logistics and pharmacy-billing firm, used Cloudyn to identify optimal pricing and sizing of Amazon Web Services resources and knocked an estimated $6,000 off its AWS bill over the last two months.
Before using Cloudyn, Inmar's manual bill-verification process showed AWS had overbilled the company by about $1,000 in a month.
Inmar runs 120 to 130 instances on AWS, and three months ago it gave Andres Silva, senior DevOps engineer, the task of analyzing and lowering the amount of money spent on Amazon's cloud.
The cloud cost analysis uncovered that Inmar hadn't been buying Amazon's reserved instances, which are purchased using either a one- or three-year commitment and tend to cost less than its on-demand pay-by-the-hour instances, Silva said. He then determined how many instances the company needed to buy and in what availability zones.
Inmar could have had some code written to do this analysis, but it needed to perform it quickly. Running a Reserved Instance Analyzer report took less than five minutes.
Inmar is also using another cloud cost analysis tool from Cloudyn that calculates which reserved instances to buy based on location and resource consumption, and is selling some of its on-demand instances on the Amazon Elastic Compute Cloud, or EC2, Spot Instances market. Inmar will pay Cloudyn around $300 a month to monitor daily spending thresholds and to prevent overbilling from AWS.
Before using Cloudyn, Inmar's manual bill-verification process showed AWS had overbilled the company by about $1,000 in a month. Cloudyn's monitoring will also check for that kind of error, Silva said.
Cloudyn has a number of competitors, including CloudVertical and NewVem, which Silva also evaluated last year, but found Cloudyn to be the least expensive at that time. But NewVem this week introduced a pricing program for small and medium-sized businesses with pricing that starts at about $300 per month. Officials say the product also offers other services beyond cost-crunching, such as configuration, security and availability analysis.
CloudVertical officials did not respond to requests for comment as of press time. The CloudVertical website lists prices for a professional plan starting at $49 a month, but does not publish pricing details for its enterprise pricing plan.
CloudCheckr cloud cost analysis calculates cuts
Another tool for frugal AWS buyers is CloudCheckr. It caught the eye of Eric Kim, founder and CEO of a Twitter-based marketing startup still in beta called Twylah. Twylah runs about 20 instances on AWS and has used CloudCheckr, also in beta, for about three months.
Through CloudCheckr's analysis, Kim saw that some parts of the Twylah application were very memory-intensive and weren't making good use of CPU resources, while other parts were CPU constrained. Meanwhile, other workloads consumed large amounts of CPU and left memory to spare. "For example, on our Web server, right now we're running on four high-CPU extra larges," he said. "We determined that it would be more optimal to change that to six regular large instances, because we did not need the extra computing power."
Similarly, one large worker server will be split off into its own high-CPU, extra-large instance to cut down on CPU utilization, so that smaller servers of this type can run on smaller instances. Kim estimates the company will save about 20% to 30% on its bill, or between $800 and $1200 a month.
CloudCheckr will come in a free version with limited functionality, along with professional and premium paid tiers based on deployment size. The professional version will start at $179 per month and top out at $899 per month, and the premium version will start at $279 per month and top out at $999 per month, the company said.
For now, Kim gets CloudCheckr free because it's still in beta, and says the jury's still out on whether he will purchase the tool once it becomes generally available. "It's not something you look at every day," he said. "I can see using the service periodically and paying for it periodically."