Last week I attended the GigaOM Structure Conference in San Francisco, which is a key event in the calendar for companies in all areas of internet infrastructure and cloud computing, bringing together thought leaders from America and beyond to discuss the future of internet infrastructure
Some fascinating highlights and trends have come out of the conference:
An entertaining panel involved Chris Kemp from Nebula (for Openstack), Sameer Dholakia (for Citrix) and Marten Mickos (for Eucalyptus) debating API standardisation vs innovation etc.
To summarise, Citrix and Eucalyptus both came over open-minded about supporting multiple APIs, either now or in the future (Eucalyptus is based on the AWS API, Citrix supports both their own and AWS), whereas Chris Kemp from Nebula came over as rather narrow minded, in that Nebula will only ever support their own API and that has enough groundswell to become the new standard. Lew Moorman for Rackspace (supporting Openstack) also presented previously at the conference presenting a similar stance.
At Flexiant, we believe that compatability with other API’s is good for users and the ecosystem, as it makes it easier for new customers and partners to integrate easier, using existing technologies. However, using someone else’s API (e.g AWS) as your standard API to build your business logic against could prevent you innovating, or at least innovation would force you to diverge from the API, thus negating the benefit.
Multi Hypervisor Solutions
While not deliberated on a panel, numerous discussions with other attendees suggests that customers interested in Hyper-V are rapidly increasing, and as such are also starting to look at multi hypervisor solutions, rather than single solutions (to support Hyper-V and their existing hypervisor, usually VMware). This is something we have already seen within service providers for some time, but it’s interesting to see it now developing in the Enterprise market as well. Flexiant has always maintained a hypervisor agnostic technology platform, providing a common standard of capabilities and interface across all hypervisors, so this is welcome news.
Software Defined Networking (or SDN)
Software Defined Networking became the somewhat surprise topic of frequent discussion within panels and conversations. Basically the idea is that switches and routers are becoming dumber and the intelligence is all moving to the software stack. There wasn’t any specific product announcements in this area though.
At Flexiant, we think SDN is really useful, as it not only moves the logic away from the hardware into the software, but more importantly as virtualisation has done for virtual machines, it abstracts it as well. This allows you to build more flexible networking capabilities that wouldn’t previously be possible.
That said, a number of companies, Flexiant included, have solutions that involve SDN and have had for some years, so it’s hardly new technology, but is clearly becoming of increasing interest this year, so we look forward to seeing how this develops.
What are your thoughts on API’s, multi-hypervisor solutions and software defined networking and how do you see each area developing? Please leave your thoughts in the comments section.