Terry Slattery has a few observations from speaking at Interop 2015
Tech Field Day
I have attended a number of the Tech Field Day events at several Interop conferences. This year, we did round-table talks on two network topics, followed by two very interesting presentations by SanDisk.
- Tech Field Day Extra Roundtable – Whitebox Switches (17 minutes)
A discussion on the implications of whitebox switches in Software Defined Networking environments.
- Tech Field Day Extra Roundtable – Ipv4 and IPv6 (14 minutes)
What is happening with the transition to IPv6? What is holding it back? Are we destined to use IPv4 forever?
- Tech Field Day Extra SanDisk Application Acceleration (55 minutes)
SanDisk is doing some really interesting things with flash memory to accelerate application performance. Flash-based disk drives are not new, but what they are doing with them shows that they really understand the difference in flash-based disks and rotating media disks. For example, the disk handling strategies built into the operating systems may not be applicable to flash-based disks, so they have modified the Linux kernel to eliminate the extra code that doesn’t apply. Of course, this speeds up the disk I/O system. If you’ve been tracking flash disks, you may have run into other articles on it: Take time to get to know the flash you think you love and The SSD Endurance Experiment. SanDisk seems to have developed systems that reduce the risk of problems.
- Tech Field Day Extra SanDisk Storage Memory Convergence (16 minutes)
SanDisk talks about using Flash memory to provide non-volatile memory to help accelerate server systems. This memory plugs into a memory slot in the server, just like a regular memory module. It provides an intermediate non-volatile storage between main memory and a high-performance I/O system. Great stuff. I came away with some ideas for using their technology in a network management application that’s disk I/O bound.
Presentations
The main reason for me to attend Interop 2015 was to do two presentations. Fortunately, they were both on the same day. Amazingly, they were in the same room in sequential time slots on Thursday morning. So I walked into the room a bit early, checked out the audio-visual system, and then did both presentations.
The first presentation was “SDN APIs for Communications” (as in Unified Communications). Over forty people attended and I received several good comments afterwards. This topic could be a real sleeper – a topic that would put an insomniac to sleep. I showed the functionality that the APIs allow, which kept it high-level. I showed a limited amount of code, with examples of JSON encoding to set the DSCP value of a call. It was a fast-paced session that kept the audience awake and provided information about what some of the vendors are doing to integrate SDN with UC. Network Computing interviewed me about the topic and posted an article the week of the show: SDN & UC Integration: A Work In Progress.
The next presentation was one that I’ve done before “Building Networks for Real-time Applications: What Works, What Doesn’t?” This was also a well-attended, fast-paced presentation that included a number of stories about network problems around handling real-time traffic (typically voice and video). Since data is sometimes the real-time traffic, there are some examples of handling it. One of the interesting points that I make is the impact of packet loss on TCP throughput. You can see some blog posts on this subject in the NetCraftsmen blog.
To learn more about SDN APIs, join me at our next C-MUG meeting, where I will be presenting with Craig Hill on the current state of SDN, and what’s ahead. You can register here.