When disruptive cloud architectures began to shake corporate IT foundations, it required a ‘cloud stack’ to create, manage, and deploy infrastructure cloud services. Today, such stacks are available from a wide range of providers, some use proprietary code, but increasingly corporate users are adopting open source products. However, like other disruptive paradigms, most cloud deployments emerge in a legacy infrastructure where entrenched workflows and incumbent vendors like Microsoft, Oracle and VMware are evolving their own software platforms into enterprise cloud stacks, and Amazon AWS reigns supreme in the public cloud market. So don’t expect to see enterprises in droves switch out their recently deployed virtualization platforms like vSphere and Hyper-V for a multi-hypervisor cloud stack. The IT infrastructure is closely tied to critical business processes, so radical change does not happen overnight – but change will occur.
The Open Source Stacks
Open source stacks include SmartOS, Open Nebula and Eucalyptus, but the leaders are CloudStack and OpenStack. OpenStack has a modular architecture and is primarily deployed as an infrastructure-as-a-Service (IaaS) offering, with a series of interrelated projects that control processing, storage, and networking resources managed through a web-based dashboard. The original codebase comes from hosting company Rackspace [and NASA]. In 2013, the code was transferred to the OpenStack Foundation, a non-profit corporate entity set up to promote OpenStack software and its community. More than 200 companies have since joined the project, including Arista Networks, AT&T, AMD, Avaya, Cisco, Dell, EMC, Ericsson, Hewlett-Packard, Huawei, IBM, Intel, Mellanox, Mirantis, NEC, NetApp, Nexenta, Oracle, Red Hat, SolidFire, SUSE Linux, VMware and Yahoo!. This year’s OpenStack Summit, held in Atlanta, drew 4,500 attendees.
The CloudStack code base was donated to the Apache Software Foundation by Citrix in 2011. It is a more integrated package that internally manages a pool of virtual appliances such as firewalling, routing, DHCP, VPN access, console proxy, storage access, and storage replication to support the cloud itself. Both stacks work with a variety of hypervisors, and a single cloud deployment can contain multiple hypervisor implementations. The current release of OpenStack and CloudStack support pre-packaged enterprise hypervisors like Citrix XenServer, Hyper-V and VMware vSphere, as well as KVM or Xen running on Ubuntu or CentOS.
The 2014 CloudStack Collaboration Conference took place in November in Budapest and brought together some 400 software developers and 20 exhibitors for two days of intensive discussions centred on how to get the CloudStack message out to the global cloud provider community. While the CloudStack community is much smaller than its OpenStack counterpart, it claims to have invested much more than OpenStack in developing and integrating its code base.
The conference was intimate – almost ‘bare metal’ – clean and direct, not much hype and a strong focus on promoting extensible, scalable and proven deployments, spearheaded by Interoute’s Virtual Data Centre service and BT’s Cloud Compute service. But there was also recognition that, while the Apache organization is great for driving technical developments, it is not a great marketing engine. Certainly, CloudStack needs a marketing and trade alliance and a bigger ecosystem to ensure regular update cycles, product enhancements and a better governance structure.
What The Customer Wants
CloudStack vendors recognize the attitude shifts in enterprise purchasing behavior, both in hardware and software, notably:
- The focus is on services and on OPEX, avoiding CAPEX as much as possible.
- Customers want QoS and support
- They want well-defined update cycles
- They want standardized hardware and software to maximize flexible and scalable deployments
Meeting these requirements will translate into broader acceptance of open standards-based products and services. However, if competing open source stacks creates confusion in the fledgling enterprise user community it will have been a detrimental effect. So, is further integration possible?
There are several projects in OpenStack, such as the Swift object storage and Neutron networking that may enhance CloudStack performance. However, adding APIs to make these projects available in CloudStack threatens the stack’s integration levels. OpenStack needs to provide better core component integration and simpler management capabilities, and demonstrate that it is not just a staging ground for vendors to roll-their-own semi-proprietary services.
Ultimately, all the open standard stack vendors need to stop focusing so much on product enhancements and mirror what has happened in the physical goods transport sector. They must evolve a standard cloud services container accessible with standard APIs and providing the minimum viable cloud services for quickest time to value and greatest flexibility.
This could be where the next battleground emerges. The rise of Docker as a container for applications and functions could lead to less of a need for a full-function cloud stack. But, that will be a different story.