|
|
Find out more |
Happy Halloween! Infoblox is providing a treat, not a trick, today for the Software Defined Networking (SDN) community: Tapestry, an open-source tool for measuring network complexity is now available for download at FlowForwarding.org.
Tapestry has attracted much enthusiasm and many questions since we first announced the project on September 25th. Here are answers to some of the questions we’ve heard most frequently and a few thoughts on Tapestry’s future.
WHY IS INFOBLOX DOING THIS?
For the past 14 years, Infoblox has been in the business of helping organizations cope with a natural rise in network complexity as more and more of their processes are digitized, networked and automated. We felt that a serious and meaningful quantitative measure of network complexity was lacking, and especially one that would be useful well into the future in an increasingly virtualized, mobile, and cloud-centric infrastructure.
IS THIS AN OPEN-SOURCE VERSION OF AN INFOBLOX PRODUCT?
Absolutely not. Tapestry implements a unique approach to measuring complexity. While the technique has what we believe to be a sound mathematical foundation, we do not know how practical it really is. Tapestry’s approach to quantifying complexity is based on understanding how many network-wide activities are going on simultaneously through analyzing interaction data between endpoints such as sensors, apps, and devices.
This is “far out” head-scratching stuff and not yet something we feel is ready to be productized. Not even close. We are simply helping start the conversation on quantifying network complexity, with what we hope is a reasonable first pass. Better methods and sound criticism are warmly invited.
DOES TAPESTRY WORK?
It’s too early to tell. We are fairly certain the method we have implemented to detect distinct activities on a network does indeed provide useful data since we have done some successful test runs using the excellent work in community detection benchmarks by Andrew Lancichinetti. But we also believe there is room to improve. We encourage the community to put Tapestry through its paces, and drop in other network detection algorithms.
We will know a lot more as Tapestry is widely deployed in real networks that are at the heart of growing and complex modern organizations.
IS THE CODE BASE MATURE?
No. We are essentially releasing a 0.1 version. We plan on improving the code fairly consistently for the next 24 months, with updates as often as once a week. We built Tapestry on a scalable computing and big data platform, but the initial version doesn’t take advantage of all that is available in our toolkit. We are excited about hearing the community’s demand for scaling.
The good news is that Tapestry controls standards-based OpenFlow 1.3 switches to do the “in-line” bits. So buying hard and soft switches from existing vendors such as HP and VMWare should drastically minimize risk to your existing infrastructure. If you don’t mind embracing more risk, you can always deploy the LINC OpenFlow Switchsoftware available at FlowFowarding.org.
WHAT DOES THIS HAVE TO DO WITH SOFTWARE DEFINED NETWORKING?
Nothing and everything. The actual network complexity computation can be run on any interaction data set such as a DNS Log. In that sense, Tapestry has nothing to do with SDN. However, we feel SDN is a perfect way to capture DNS interaction data that would let us get different kinds of interaction data later with minimal effort.
Since we are also interested in future SDN systems, Tapestry has been a useful new end-to-end SDN application that shows what an SDN control plane should do, for example, to outside of the standard “traffic management” use-cases.
WHAT DOES THE “ROADMAP” LOOK LIKE?
Pull up a chair… how much time do you have? Based on initial feedback, this one little app literally opens the door on an entire landscape of features that were not obvious to us when we started. It’s clear we have at least a few years of updates ahead and that is BEFORE the community has really started to pitch in.
Here are a few things we hope to provide in coming months:
- Lots more documentation about how Tapestry works and how to deploy it.
- Even more detail behind the NCI number, such as how many distinct communities have been detected.
- More rigorous testing and benchmarking.
- IPv6 support.
- More detail on the input data, such as the “most chatty and most popular” endpoints and domain names.
- Real-time visualization of the input data (could be an eye-popper).
- GUI tools for deployment including real-time visualization and GUI controls of OpenFlow switches.
- Various “knobs” on the input sources. For example, should Google.com be counted as one endpoint or many (e.g., based on the number of Google servers that Tapestry sees)?
- Scaling improvements; we’d like to be able to handle hundreds of millions of endpoints fairly quickly.
REALLY?
Yes, really. We are committed to supporting Tapestry in the community via FlowForwarding.org until someone else takes it over, or we are convinced no one else cares.
I hope you’ll check it out and tell us what you REALLY think. If you find Tapestry interesting, I hope you’ll join us. I haven’t been this excited about the IT industry since the 80s!
Oh yeah, I almost forgot: We plan to have an iPhone app version of the Tapestry UI out very soon.