Absolutely right! Of course, when we are talking about Network Services, which are really dynamic composite applications, that gets trickier.
As we know from our work in CloudNFV and Cloud Enterprise in general – not all Functions have the same scaling characteristics, which impacts the whole.
In fact, many vendor Virtual Network Functions being offered were never really designed for Virtualization, they are simply re-packaging of existing capabilities for a new mode of distribution (they may be “on the Cloud”, but they are not “of the Cloud”). At least in the case of VNFs, it becomes an objective measure of reviewing an isolated function – over time, we can expect competition to drive continuous re-factoring of VNFs.
But what about the Orchestrators – the supposedly integrated MANO platforms? If you take a conventional Service Oriented Architecture middleware stack approach circa 2005 (e.g. ESB; BPMN; BRMS; RDBMS; HDFS; Analytics; etc.) – watch out!!!! Not only is that an unwieldly stack with a big footprint requiring systems integration that enterprise IT has struggled to deal with over the last decade – these middleware components don’t offer the same scaling characteristic!!!! Applications (i.e. Network Services) written over the top of them will scale based on the worst performer – that’s a solution stack problem independent of your VNFs. Oh, and by the way, the “integrated” stack approach itself introduces a lot of network chatter (indirection, latency, scheduling, consistency, etc.). You might just be replacing hardware spaghetti w/ software spaghetti and end up doing truck rolls to maintain your middleware instead of your middle-boxes.
This is a huge problem that we’ve called out at various Telecom meetings, but I’m afraid most folks won’t see the wall until they run right into it.
As Tom is suggesting in his post. The industry has to have foresight to understand true complexity of SDN and NFV. I appreciate that “simple” is the best answer from a marketing perspective, but “simple is as simple does” from a technical perspective. The fact is ignoring complexity for convenience is generally a bad thing.
At EnterpriseWeb, we’ve re-imagined the application layer from the ground up to address the challenges of distributed computing. CloudNFV is an application that highlights the benefits of a unified and lightweight middle-tier that enable logic over diverse and distributed endpoints. A policy-based system that enables management to express intent independent of implementation details. An application platform and architecture that not only abstracts MANO, it abstracts challenges of managing state and consistency in distributed and parallel computing (nasty stuff to code per application). A platform with unified scaling characteristics.
Tom blogs truth to power. Ignore the writing on the wall (blog) at your own risk.
http://blog.cimicorp.com/?p=2093
CIOs See a New Cloud Model Emerging blog.cimicorp.com
In some recent chats with enterprise CIOs, I noticed that there were some who were thinking a bit differently about the cloud. Their emerging views were aligned with a Microsoft Azure commercial on sequencing genomes, though it wasn’t possible to…
Originally posted on LinkedIn, click here to access.
