Running a hybrid cloud environment and having the option to burst from your own private cloud infrastructure to a public cloud can be very convenient, but here's the problem: how do you know which of your virtual machines would be the best ones to move to the public cloud?
You obviously want to make the maximum possible use of your own resources, so there is no point leaving spare capacity behind when you move VMs to a public cloud. But there are constraints: regulatory or other reasons may mean that some VMs have to stay where they are, even when it would be easier to optimize the usage of your own capacity by shifting them.
In short, it's not always as simple as it first looks.
One company that's seeking to make the decision easier is VMTurbo. The company has just released version 4.0 of its Operations Manager software, which has out-of-the-box support for Amazon's AWS as well as Microsoft's Azure cloud.
A quick recap on what VMTurbo's software does: put very simply, it creates a market for resources and allows you to give applications a "budget" to spend on them.
Mission-critical apps get more resources than non-mission critical ones, and the system uses its Economic Scheduling Engine to match supply and demand and allocate the resources efficiently. The most important applications are the "richest," so they are the ones that can choose the resources they need, while the "poorer" applications can "buy" whatever is left.
The idea is that rather than operating a Soviet-style planned system where applications are allocated resources based on a combination of prediction and guesswork, it's capitalism and Adam Smith's "invisible hand" of market forces that handle the allocation work for you, quickly and efficiently. And we all know that capitalism beats communism hands down when it comes to the efficient allocation of resources.
VMTurbo's newly released software automatically "discovers" AWS or Azure clouds and feeds its resources into the marketplace model, so that these resources can be taken into consideration when it comes to moving applications around.
"We understand the characteristics of AWS (and Azure) and we know the profiles of workloads — which ones have HIPAA considerations and so on — so we can recommend which ones to move to the cloud and when, and also when to bring them back in-house," says Lauren Whitehouse, the company's product marketing director. "The objective is to make the most efficient use of the resources that are on hand so that you don't have spare resources of your own and end up paying for the cloud as well."
Curiously, for a system that uses the concept of pricing to allocate resources, VMTurbo's software doesn't take into account the actual dollar cost of using AWS or Azure. It's more a case of using the public cloud when there's no other option, not when it is cheaper to do so than any alternative.
VMTurbo also wants to help when it comes to allocating storage resources for VMs. "Storage can be a challenge for companies that virtualize their infrastructures because you lose visibility into workloads and the storage they are running on," explains Whitehouse.
VMTurbo aims to understand each piece of storage — its I/O capabilities, whether it is SSD or HDD, snapshotting habits and so on — and that is all taken in to account when the storage is allocated to VMs. "We are not just doing that one time — we tune the storage so that it is always optimized," she adds.
If it works as it should, VMTurbo's Operations Manager can offer capex cuts of 20-30% due to increased allocation efficiency, and opex cost cuts of about 25%, Whitehouse claims.
Does it really work? Plenty of customers seem to think so.
When we last looked at VMTurbo some eighteen months ago the company had 300 customers. That number has since grown by 30% to almost 400. Some of these are mid-market companies with 100 - 1000 employees, but most are 1000-plus employee enterprises, and there are now a couple dozen large-scale service providers as well.
"The bigger the company, the more complex and bigger the interdependencies. That's not something that a human can manage, and we believe that you have to give control to software," Whitehouse concludes.
Capitalism's victory over communism was inevitable because an economy is too complicated for a human to manage. Now that public clouds are getting that way, software-based resource allocation — such as that offered by VMTurbo — is likely inevitable as well.
Paul Rubens is a technology journalist and contributor to ServerWatch, EnterpriseNetworkingPlanet and EnterpriseMobileToday. He has also covered technology for international newspapers and magazines including The Economist and The Financial Times since 1991.