UptimeCloud Eliminates AWS Cloud Bill Surprises

Uptime Software announced Monday that it's offering an online service that can forecast what your cloud bill will be before the end of the month. That's been a blind spot for many early cloud adopters.

Part of the appeal of cloud computing is the ease with which business end users can tap into needed resources. Part of the pain is when end users self-provision on an unpredictable scale and the bill comes as a shock.

Sometimes the shock comes from existing end users, who expand their number of instances, start moving large data sets around, or suddenly decide to download the data they've been amassing, running up export fees. The activity results in additional charges over and above the ones IT expects. In addition to a surprise, they may lead to a budgetary event--a sitdown in the chief financial officer's office.

Even if an IT staff knows how many cloud instances are in use by employees in its organization, it's still hard to guard against surprises at the end of the month. Usually someone is charged with building a spreadsheet, entering user accounts and expected monthly usages. At larger cloud users, that timekeeper needs to spend a day or several days a month trying to keep all the information up to date through an imperfect Q&A polling process with users.

[ Uptime previously made a name for itself with an application that could project usage and provide capacity management in VMware environments. To learn more, see VMware Gets 'Set It And Forget It' Monitoring].

There's even third-party services to assemble your billing information for you--provided you turn over account numbers and passwords to them, something many businesses hesitate to do for what amounts to a clearer presentation of the monthly bill they were going to get anyway.

UptimeCloud, a software as a service from Uptime Software, says it can monitor usage of any given Amazon customer's set of virtual servers as they run through the month. UptimeCloud can calculate existing charges and project them through the end of the month, forecasting what the total bill will be.

The SaaS doesn't require the Amazon EC2 customer's account passwords to do so. Instead of inspecting individual accounts, it plugs into Amazon Web Services' read-only monitoring API based on Java Script Object Notation, the lightweight alternative to the chattier XML. UptimeCloud collects 400 data points on running virtual machines through the API. With the data, it can sum up a company's current bill for all the virtual machines running in the AWS EC2 cloud at any time of the month. Since UptimeCloud is inspecting Amazon's operational metrics of running virtual machines, the calculation is a near real time figure.

It can also calculate subtotals within the company, such as a department, or tally an individual's amount of virtual machine use. Such tracking is difficult for most Amazon customers because Amazon itself uses more general statistics derived through the use of one tag on an account. Suppose an HR department had branches in New York, Chicago, and San Diego. If a virtual machine used by HR was running above its normal monthly budget, IT would want to know which office of the department, or perhaps even who within an office, was responsible for the overage.

Such investigation is difficult, even at the end of the month, with AWS billing because "Amazon's default is using one low-level tag per instance," said Nick Johnson, director of marketing for Uptime Software, in an interview. UptimeCloud, on the other hand, can assign tagging in greater depth, using multiple tags to distinguish the office in San Diego from the one in New York. Some Uptime users implement a hierarchy of as many as 12 tags so that virtual machine use can be pinpointed in the resulting bill, aiding IT in controlling runaway costs.

Budget limits can be built into the system so that going over the budgeted amount triggers alerts. Johnson said Uptime's own quality assurance staff triggered an alert as it commissioned a thousand servers for a particularly intense period of testing. Its own system sounded an alert that the unit was going to exceed its monthly budget, but QA staffers explained the use of so many machines would be for only a three-day period and its monthly bill wouldn't vary widely from the average, said Phil Didaskalou, CEO of Uptime, in an interview.

"Before, we wouldn't have known about it," he observed. If the monthly bill is going to deviate from the budget, he'd rather know about it sooner than later, he added. UptimeCloud can only sound the alarm; it's not authorized or able to shut down virtual machines to keep their operation within budget, since that would interfere with AWS' operation of its infrastructure as a service.

The SaaS bill management application understands AWS's EC2 billing tiers and can analyze a bill and recommend to a customer a money-saving alternative. Amazon has on-demand pricing by which it charges by the hour, such as 8.5 cents for a Linux server or 12 cents for a Windows server. It also offers reserved system pricing for one- or three-year terms that amount to lower rates per hour. It also has spot pricing, where customers bid a certain price to run a job. The job is held until an off-peak period when Amazon may offer a spot price that matches the customer's price. Some workloads are not suitable for the reserved or spot-pricing models and Johnson claimed UptimeCloud could discern whether a workload was a match for the pricing model in its recommendations.

Uptime has used its product to detect when a sales support engineer "ramped up a lot of big machines, then accidentally left them on for two months," Johnson said. An expected $27 charge grew into a $4,000 bill, "and our CFO hates it when that happens," he added. They hadn't expected an alert was needed, so none was set, but at least IT could track down the source of the overage and prevent another occurrence.

The service is priced as a percentage of a customer's total AWS bill. Amazon users with a $10,000 bill per month would pay about $295, Johnson said.

IT's jumping into cloud services with too much custom code and too little planning, our annual State of Cloud Computing Survey finds. The new Leap Of Cloud Faith issue of InformationWeek shows you what to be aware of when using the cloud. Also in this issue: Cloud success stories from Six Flags and Yelp, and how to write a SAN RFI. (Free registration required.)