The Elastic Cloud: Leading Cloud Stacks Shape API Conversations

APIs (Application Programming Interfaces) and stack-based delivery models are a vital part of the cloud process.

More organizations are finding great ways to use cloud computing to create a more elastic infrastructure. But as they boost their reliance on cloud computing, some organizations worry that their environment won’t be compatible with other platforms. So APIs (Application Programming Interfaces) and stack-based delivery models were introduced into the cloud computing technology matrix.

These APIs are actually a vital part of the cloud process. Why? They help create direct cloud computing connections. Basically, they are pushing towards a more agnostic cloud platform. With more cloud deployments, data center administrators will need to find ways to extend, connect, or integrate their cloud environment with other services.

Today, there are three core areas where cloud computing can help: Infrastructure, services and applications.

Within those categories, you’re able to place services like SaaS, PaaS, IaaS and so on. But what happens when we start delivering “Everything-as-a-Service?” What happens when the cloud model continues to grow and evolve, and new types of connections are required? Let’s take a look at three stack models and how they’re helping shape the cloud connectivity and API conversation.

Apache CloudStack

CloudStack has been growing in popularity with many different organizations. Originally developed by Cloud.com, CloudStack was purchased by Citrix and then released into the Apache Incubator program. From there, the first stable version was released in 2013. The platform is already compatible with hyervisors like KVM, vSphere, and XenServer. Apache CloudStack is an open-source cloud management platform designed for creating, controlling, and deploying various cloud services. Similar to the other stack-based models, CloudStack supports the Amazon AWS API model and many other APIs.

The good: CloudStack 4.0.2 is the first stable release – but it’s very new, just five months old. Still, the latest version includes great features, like scaling storage independent of compute or having VMs maintain their machine state without having to experience compute changes. New security features now allow the administrator to create security zones across various regions. The overall deployment of CloudStack is smooth. In a typical setup, you would have one VM running the CloudStack Management Server and the other VM acting as the actual cloud infrastructure. From a testing and deployment perspective, you could deploy the whole platform on one physical host.

What to look for: Remember, the latest release of CloudStack is very new. The other challenge is that we have yet to see any major cloud provider adopt the platform.  Finally, from an engineering perspective, some have pointed out challenges around the monolithic architecture and installation process. In some cases, although simplified, the installation process will require a bit of knowledge. Still, the platform is being adopted by a few other big players. During the summer of 2012, Datapipe announced that its global infrastructure will run on CloudStack. Other organizations like SunGard, Citrix and WebMD have already adopted the CloudStack model.

OpenStack

With more than 200 companies adopting this platform, OpenStack is certainly one of the more popular cloud models out there. Currently managed by the OpenStack Foundation, OpenStack consists of multiple, interrelated stack-based parts. These components all tie together to create the OpenStack delivery model. Much like CloudStack, there is further agnosticism when it comes to the underlying hypervisor and infrastructure on which OpenStack may run. It’ll support platforms which include VMware, Citrix and KVM.

The good: Let’s face facts - OpenStack is, arguably, the most mature stack-based cloud control model out there. Furthermore, OpenStack's adoption momentum has been very strong. The latest release of Havana shows some pretty big improvements around all major components in the stack. The networking component (Neutron) allows administrators to do some pretty amazing things with their cloud model. Now, with direct integration with OpenFlow, Neutron allows for greater levels of multi-tenancy and cloud scaling by adopting various software-defined networking technologies into the stack. Furthermore, the networking framework around OpenStack has new services like intrusion detection services (IDS), various load-balancing features, firewall technologies and even a VPN that you can deploy and manage. Traffic and IP redirection is made easier – thus creating a stack platform capable of even greater resiliency and failover.

Next: The Path Ahead for Open Stack, A Look at Eucalyptus

OpenStack (Continued)

What to look for: Adoption around OpenStack continues to be strong. CERN, Yahoo!, Rackspace Cloud, HP Cloud, AT&T and several other large organizations are actively working on and developing around the OpenStack platform. Still, even with so much adoption, OpenStack can still be challenging to assemble and deploy. Administrators have said that many key components need to be managed from various command-line consoles. The platform has eight different modular components - Compute, Open Storage, Block Storage, Networking, Dashboard, Identity Service, Image server, and Amazon Web Services compatibility. To some, this comprises a somewhat fragmented architecture. Still, with amazing developments around the platform and strong adoption, OpenStack continues to lead the way for cloud delivery and control.

Eucalyptus

Formed in 2009, Eucalyptus has been working hard to create an AWS-compatible private/hybrid cloud infrastructure. Through the years, they’ve incorporated the pooling of compute, storage, and network resources for a respective cloud platform. The Eucalyptus platform enables users to migrate seamlessly between private cloud resources and those located within Amazon.

The good: On June, 2013, Eucalyptus 3.3 was released, with some pretty great new features. Following the demands of the market, Eucalyptus now includes three core features which further help enhance the entire cloud computing process: Auto-scaling, ELastic Load Balancing and CloudWatch.

Furthermore, Eucalyptus 3.3 is also the first private cloud platform to support Netflix's open source tools - including Chaos Monkey, Asgard, and Edda. Furthermore, the Euca-community has been known to be strong when it comes to open-source support. Finally, the latest version of 3.3 boasts simplified deployment process where almost any organization can now easily run and deploy their Euca-cloud on commodity hardware.

What to look for: The Eucalyptus platform continues to grow and expand. Their latest release added some core features when it comes to enhanced workload automation and orchestration. Still, some users are a bit nervous about the long release periods for the product. The year and a half wait between Eucalyptus 2.0 and 3.0 create a lot of questions around the development cycle of this platform. Still – the team has been going strong with many more consistent releases as of late. Furthermore, there are still wide adoptions of the Euca-cloud. Adopters include organizations like Sony, Puma, NASA, Trend Micro, Nokia and several others.  This platform is continuing to generate a lot of excitement because of the ease of deployment and manageability. Plus, adoption trends have been increasing for the Euca platform.

Although we mentioned three big cloud players here – there are more emerging platforms that are making waves. The stack-based computing model and the cloud API structure is heating up because there are more cloud workloads out there and more organizations actively moving towards the cloud. Development to help big (and small) companies tie their resources together within an agnostic cloud platform create an environment which can scale dynamically and increase resiliency. As cloud adoption continue to grow – more of these platforms will be tasked with handling yet additional features, services, and deployment models.