What’s the big thing about Open Daylight?


As described in my previous post, What is the state of Software Defined Networking today?, Open Daylight provides an open source, open API, collaborative eco-system, which brings the best technologies from each of the contributors.  But more importantly, Open Daylight shines a light to a part of the Data Center and IT Application development that has been in the dark to Application developers – the network.

OpenDaylight Diagram

Server virtualization abstracted the complexities of the underlying hardware and automated virtual machine scaling (e.g. adding more processor cores or memory to a VM).  More importantly App developers had access to programmatic interfaces that enabled policy based automation.  Using analytics to model current and future demand on a client’s server resources, IBM was able to automate workload placement, quadruple server utilization (see below) and reduce the number of servers by five times, with $4 million in savings estimated the very first year.

CPU Utilization

Similarly, storage virtualization abstracted the complexities of the underlying devices (e.g. what is a cylinder, head and record number?) and automated data placement and scaling.  Again App developers could see programmatic interfaces that enabled automated, intelligent data placement.  For example, by moving more frequently used data to low latency, solid state devices and less frequently used to slower magnetic devices.

Unfortunately, application developers have been in the dark regarding programmatic interfaces for the underlying network.  As a result, today a virtual machine can be brought on-line very quickly, typically in a few minutes.  However, to bring up the entire multi-tier virtual system, which that virtual machine is a part of, takes days.  For example, at OFC 2013, one of the presenters stated that “It takes about 5 days from an end-end point of view to provision something like that (a multi-tier system)” with his goal to “get at least to sub-one day.”

Virtual System Diagram

We are investing in a DOVE Connectivity Service that automates the creation of a multi-tier virtual system, including all the linkages between compute tiers & network appliances.  This service will reduce the multi-day cycle that exists today, down to minutes.  It is a key element of IBM’s SDN controller for Virtual Environments (SDN-VE).  The SDN-VE code will be contributed to the Open Daylight project.  We are focused on enabling, through Quantum APIs, this network connectivity service for Open Daylight.

Another example of how programmatic network interfaces changes things is the use of feedback by placement engines. Today, applications are completely oblivious to the network and vice versa.  That is, today, the network cannot provide feedback to applications on things like network latency, jitter, operations/second (for small sized traffic), bandwidth (for large sized traffic), etc. More importantly, applications cannot take that feedback into account for things like workload placement.  Unlike server virtualization, which provides a rich amount of feedback (e.g. CPU utilization, IO utilization, memory footprint, etc.) that can be used for workload placement.  We are focused on ensuring Open Daylight provides APIs that pass network feedback to upper level software, such as IBM CloudStack, which is OpenStack based.

[1] Source: 2012 IBM Data Center Study:   http://www.ibm.com/data-center/study

Renato Recio is an IBM Fellow & CTO of IBM System Networking, specializing in System I/O and Network Architecture, Strategy, and Standards. For the past 15 years Renato has played a leadership role in the strategy, architecture and design of future IBM system I/O and Networks. He is currently responsible for cross-IBM Data Center and Virtual Switch Networking product strategy and roadmap.

Smarter Computing Analyst Paper - HurwitzTo effectively compete in today’s changing world, it is essential that companies leverage innovative technology to differentiate from competitors. Learn how you can do that and more in the Smarter Computing Analyst Paper from Hurwitz and Associates.

Subscribe to the Smarter Computing Blog

Recent Posts

SPARKing a data revolution on z Systems mainframe

Paul DiMarzio

What does Spark have to do with the mainframe? A lot! Linux and the mainframe have been a perfect match, and so it will be with Spark – for both z/OS and Linux on z Systems.

Continue reading

A window to open innovation

Andy Dominguez

But what if you could respond to clients with lightning fast response times—and in a fashion that is tailored to the individual? What if you could give them waitless response times, coupled with waitless answers, solutions and products?

Continue reading

Leave a Reply

Your email address will not be published. Required fields are marked *

* Copy This Password *

* Type Or Paste Password Here *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>