Doubt Creeps Into OpenStack
Doubts about OpenStack put the focus on the open source cloud platform's weaknesses.
The OpenStack community recently wrapped up what can only be described as a successful summit in Tokyo, with users, developers and members of the supporting vendors unveiling a wealth of new service offerings and reaffirming their commitment to seeing the format through to widespread deployment.
And yet, amid all of the warm feelings in Tokyo and elsewhere, there remains an undercurrent of apprehension throughout the OpenStack community as more and more voices call attention to the platform’s shortcomings. Is it possible that an open, non-proprietary networking solution is not the best way to support distributed cloud architectures? And has the organization itself and its contribution process become too unwieldy to satisfy constantly shifting enterprise data demands?
OpenStack took a serious shot across the bow this week when no less than British telecom giant BT Group told Light Reading that it was seriously considering switching to a different networking option unless the community addressed some problems in six key areas. These include virtual network functions, service chain modifications, scalability, security, backward compatibility and the issue of “start-up storms” when too many nodes log on at the same time. BT’s chief researcher, Peter Willis, bluntly told the SDN & Openflow World Congress in Germany that if these concerns are not addressed, BT will not use OpenStack as the basis for its forthcoming virtual enterprise services portfolio.
To make matters worse, OpenStack champion Rackspace told investors at its third-quarter conference call this week that demand for OpenStack is ebbing, in part due to more appealing offerings from other clouds. Over the summer, Rackspace extended its Fanatical Support service to the Azure Cloud, and followed that up just last week with Amazon Web Services – two moves that the company says has brought in a wealth of new customers even though Rackspace itself is still powered by OpenStack. Company execs tell Server Watch that the key to a successful cloud service is not to pick the right framework but to offer multiple options to suit a widely diverse user base.
This kind of hand-wringing crops up every six months or so as another OpenStack summit draws near, says Computerworld’s Ben Kepes. This is actually healthy for the organization, because it focuses attention on where the challenges lie. And in the case of the Tokyo summit, perhaps the most significant announcement was the launch of a new certification program for developers. Across the board, the single biggest complaint about OpenStack is that organizations cannot find enough technicians with the expertise to successfully deploy and manage an OpenStack environment. This leads to exactly the kind of launch, integration and operational issues that cause decision-makers to seek the relative stability of proprietary platforms. Still, the fact remains that the number of deployments reaching enterprise production status has nearly doubled since 2013, and the number of large organizations utilizing OpenStack continues to rise.
Indeed, those with the most experience designing and developing open platforms realize that ease-of-deployment is a crucial aspect for any system that hopes to challenge the status quo. Canonical recently launched Ubuntu 15.10, which it has dubbed “Wily Werewolf,” containing a streamlined OpenStack installer that takes much of the guesswork out of the process. The Autopilot system offers the latest OpenStack Liberty edition and works with Canonical’s JuJu and Landscape orchestration and management stacks to ensure that only crucial deployment decisions are left for human administrators. Redundant or non-essential tasks are either automated or reserved for the set-up process. Canonical has also added a hypervisor/container runtime hybrid that provides live migration and other functions and can ultimately be set up as the default option for OpenStack.
So it would seem that the outlook for OpenStack is not completely sunny, but neither is it utterly dismal. Scale-out, distributed and completely virtual networking is a tall order for any platform, open or proprietary, and the enterprise community will have to get used to the idea that it won’t happen in a day.
In fact, it would be more unnerving if there were no complaints about OpenStack at all, as it would be a strong indicator that its supporters are not fully aware of the monumental task they have undertaken.
Arthur Cole covers networking and the data center for IT Business Edge. He has served as editor of numerous publications covering everything from audio/video production and distribution, multimedia and the Internet to video gaming.