Life changed for programmers and operations teams when the cloud arrived. Instead of waiting weeks, months, and sometimes more than a year for new hardware to be purchased and provisioned, the cloud of servers makes it possible to get a new idea up and running in seconds with just a click or three.
Alas, every great leap forward usually comes with some backsliding. Now that anyone can start up a server in seconds, everyone is doing just that. The number of machines is proliferating and our jobs are now dominated by the task of herding the dozens, hundreds, or even thousands of machines. Where we once could keep everything straight with a clipboard or a checklist in a Word or Excel document, now we need serious tools to juggle everything.
Programmers have solved this challenge as they solve everything: by creating another layer of code that stores bits in another set of tables in another database. In this case, the code is a hydra-headed tool sometimes called hybrid cloud manager that can connect to all of the clouds out there, create and destroy machines on its own, and at the same time fill its own database tables with lists of the machines and where they are.
These so-called multicloud or hybrid cloud management tools keep track of our machines and make it possible to orchestrate hundreds or thousands of instances so the work gets done by the cheapest, fastest, and best-configured option. They also make it possible to blend multiple public clouds with in-house servers to lower overall costs, improve security, enhance redundancy, and support world-wide operations.
Do you worry about “cloud lock-in?” That your code may run only on one cloud, leaving you with no alternatives? These hybrid cloud management tools are designed to prevent that from happening. Your developers should be able to ask for a cloud machine for a particular purpose, and the multicloud manager will deliver the best fitting option, whether it’s in your local hardware or in any of the supported public clouds.
But it’s not just about playing the public clouds off of one another. The tools can also enforce uniform security and configuration rules, simplifying the job of keeping all of the so-called commodity machines actually behaving in the same, correct way. They also will track usage and help you make sure that your different development teams and projects are sticking to your desired budget for computational power and monthly charges. In other words, they can do a better job keeping the cloud machines in order than you used to do with that clipboard or spreadsheet.
Here we’ve rounded up some of the most prominent options for multicloud deployment and management that caught our eye.
Scalr
Scalr tries to balance tight, hierarchical control of all machines with a spirit of self-service that promotes agility by letting all of the users create machines as needed—but only within the parameters dictated by the cloud-based management system.
The standard definitions and templates for the resources are arranged into an object-oriented like hierarchy where they can inherit default settings from other templates. You could set up a default instance for database machines that insists upon safer firewall configuration and encrypted file systems. The inheritance makes it simpler to make sweeping, company-wide changes in the defaults without going into individual templates.
The users are given guardrails that limit choices and enforce rules depending on their identity. A QA team might only get the ability to start up clones of the production machines to ensure accurate testing while the developers get access to the full API of each cloud provider. Or, if you’re worried about the developers overprovisioning machines, you can tighten up some of the limits for them to encourage careful programming while giving the devops team more latitude to spend in emergencies.
The costs for the various machines are assigned to developers, teams, and applications, making it possible to find the parts of your stack that are costing the most money. Scalr works with Amazon, Azure, Google, VMware, and OpenStack clouds.
RightScale
RightScale splits its cloud-based cloud management solution into two major parts: CMP and Optima. CMP (Cloud Management Platform) aims to be a single dashboard for all of your clouds, external and internal. You can use CMP to request a new machine and it will help you make the right decisions. The options are spelled out using predefined templates that can adjust to requests. You might, for instance, steer your developers building LAMP stacks to particular versions of Linux in particular clouds. The developers are also prompted to set expiration dates and to create policies that might shut down instances after hours when development machines won’t get used. Shutting down development instances at night and on weekends is a great idea.
Complementing CMP is Optima, which provides tools for more sophisticated accounting of cloud costs. Optima adds an extra layer of accounting detail by splitting up the bill into parts, so each team can get “visibility” into which charges they’re responsible for adding to the monthly total. You’ll need to allocate the instances and then Optima will automatically track costs and send out regular emails to give your project managers one more thing to do.
If you’re worried that the project managers won’t pay attention, Optima includes several automated features that will scan through your collection of machines and flag instances that seem to be overprovisioned. When it can, it will make suggestions to shrink some machines and cut the bill.
There is also an option to create policies that will constrain how your team selects instances. Optima can flag unattached volumes that may be holding unwanted data or watch for instances that live longer than expected.
RightScale integrates with Amazon, Azure, Google, IBM, Rackspace, OpenStack, CloudStack, and VMware clouds.
Red Hat OpenShift
Red Hat offers a simple model for multicloud deployment with the OpenShift platform. If you’ve got a Docker container and you want to run it under Kubernetes, OpenShift will help you move it around between the major clouds and your own private collection of machines. You hand your code to the continuous integration process and then OpenShift handles the configuration and deployment.
Red Hat’s set of tools is long on integration with IDEs as well as the bits of infrastructure and standard containers that you might want to put together with your own code. There’s even a neat tool that lets you drill down into the JVM console on a particular instance. It’s not so focused, though, on managing the costs and proliferation of containers. As with Pivotal Cloud Foundry (below), you’ll have to use other tools once your application is up and running.
Pivotal Cloud Foundry
Pivotal doesn’t sell Pivotal Cloud Foundry as a multicloud management tool, but the platform-as-a-service software offers some of that functionality along with its tools for deploying apps, containers, and functions. Once your code fits into the Cloud Foundry definition of an app or a function, you can deploy it to all of the major clouds or your own internal stack of boxes. Pivotal’s platform does most of the work.
Is this an advantage? Pivotal likes to claim “Never patch an OS again”—and that’s great until you want to use a different operating system. Pivotal controls that part of the stack and you surrender all interest in it when you sign on board. That’s why they also suggest that companies can enjoy a ratio of developers to operators of 200 to 1. Pivotal’s standardization cuts through the operational complexity.
The platform can be seductive for programmers who want to use fairly standard models like Java Spring and .Net. Pivotal has added deployment mechanisms to the continuous integration portal so your code pretty much flows from commit to deploy automagically. If you want to deploy your app elsewhere, you’ll need to duplicate the work done by Pivotal, but the application frameworks of Cloud Foundry are well-supported and generally open source.
Pivotal is also moving down the road toward Lambda-like functions with their own model that connects events with HTTP requests that bounce around a collection of microservices. This isn’t ready for general use, but it too is multi-cloud, unlike the popular options from major cloud providers.
The ease of use, however, has limits. While the Pivotal Cloud Foundry platform can take its applications to any cloud, Pivotal offers few of the sophisticated management tools for tracking deployments and controlling costs like some of the other tools here.
Morpheus
The team behind Morpheus likes to say their product was built by developers who needed it themselves—to optimize cloud usage across all of the companies in a private equity portfolio. When they were happy with using it, they decided the rest of the world might want it too.
The Morpheus solution emphasizes automation with deep hooks for connecting to Jenkins, Chef, Puppet, and other tools. Your code goes in one end and then Morpheus takes it all of the way to deployment, enforcing site-wide policies using local templates and restrictions. Details about the network routing and load balancing are also available to smooth deployment and limit mistakes (or at least ensure that the mistakes are made everywhere in exactly the same way).
Morpheus comes with deep integration with many common open source tools so there’s a good chance you’ll be able to migrate your current build process. They joke that you won’t need your own army of scripting monkeys because they have already brought them under their roof. Morpheus’s list of integrated tools is very long and quite complete.
The creators have also been paranoid about losing data, building in several different opportunities to back up or clone your files or databases. This can also help smooth migration and make it simpler to deploy across multiple clouds.
Morpheus supports Amazon, Azure, Google, IBM, Oracle, VMware, and OpenStack clouds, along with a few other public clouds and many other private cloud platforms.
Embotics
Embotics vCommander is full-service management software that takes your virtual machines and containers and controls how they are deployed across public and private clouds. Your containers are placed where they will be most effective and economical. All of the accounting details are available for you to watch for anomalies and overspending.
Embotics vCommander is built to be relatively open and flexible, providing many APIs and tunnels for passing API calls through it. You can send calls directly to the various providers, through what Embotics calls “Cloud Direct,” or you can rely on vCommander to do this work with a “Cloud Brokered” connection. In both cases, vCommander will track usage.
The Cloud Brokered option helps when you’re layering vCommander on top of your current infrastructure. There’s no need to rewrite all of the direct calls to any particular cloud immediately. You can migrate them to Cloud Direct later.
There are also deeper options that offer more architectural structure. The Microservices option, for instance, provides much of the automation that you need to turn your Git repository and the Docker container it produces into a microservice.
The Virtual Lab option is designed to make testing a bit easier both for QA teams and the developers who must fix the problems. The tool juggles the virtual machines to allow testing and replication to occur in a fenced and secure corner of the cloud.
The Intelligent Placement option lets vCommander make the decision about where to run a container given the policies that you’ve pre-determined. The biggest knob you can turn is the projected cost, but you can also add limits or quotas on particular data centers or require particular local features. The catalog of different cloud options includes details on compatibility with different deployment tools such as Puppet, Chef, and Ansible.
Embotics vCommander works with AWS, Azure, VMware vSphere, Microsoft Hyper-V, and Kubernetes infrastructures.
DIY or native cloud management
For all of the advantages of using a multicloud manager, it’s worth noting that there may be times when it’s not the perfect solution. All of the major clouds offer substantial price discounts for bulk use, so it can be cheaper to concentrate all of your machines in one place. If you’ve got the ability and the desire—and a server room and infrastructure already in place—then adding a few machines to act as your in-house cloud can be even cheaper. It all depends on how you account for the costs of heat, storage, and upkeep.
But living without the public cloud does mean forgoing all of the power that’s built into the cloud platforms. You might ask yourself why you’re adding an extra layer to create a new instance when you could just go to the website for the cloud provider and create one right there. Choosing “none” and doing everything through the standard public interface is a real option. The built-in tools of AWS, Azure, and Google Cloud are already pretty good for tracking your machines and accounting for their usage.
Running without the extra help of a hybrid cloud manager isn’t a bad solution, but you would miss the extra instrumentation and features built into the third-party tool. The multicloud managers often do a better job of managing the cloud than the standard portal. They’re selling their ability to help you manage costs, and you won’t get much of that flying alone and putting all of your eggs in one basket. None of the cloud companies ever suggests using a machine with less CPU power or fewer gigabytes. The multicloud tools have bots that will scan for underutilized or over-provisioned instances.
The multicloud management tools will also prevent or at least minimize cloud lock-in, giving you the freedom to move to another cloud more easily than if you didn’t use one. These cloud managers also tend to do a better job of enforcing security rules everywhere. They do more than just make it possible to play the different cloud companies off of each other and get the lowest rates. The control features are so desirable that you might want to use a multicloud manager even if you’re going to put all of the machines in one cloud.