- How Transferable Are Features in Deep Neural Networks? — (answer: “very”). A final surprising result is that initializing a network with transferred features from almost any number of layers can produce a boost to generalization that lingers even after fine-tuning to the target dataset. (via Pete Warden)
- Introducing Atlas: Netflix’s Primary Telemetry Platform — nice solution to the problems that many have, at a scale that few have.
- The Many Facades of DRM (PDF) — Modular software systems are designed to be broken into independent pieces. Each piece has a clear boundary and well-defined interface for ‘hooking’ into other pieces. Progress in most technologies accelerates once systems have achieved this state. But clear boundaries and well-defined interfaces also make a technology easier to attack, break, and reverse-engineer. Well-designed DRMs have very fuzzy boundaries and are designed to have very non-standard interfaces. The examples of the uglified DRM code are inspiring.
- DPDK — a set of libraries and drivers for fast packet processing […] to: receive and send packets within the minimum number of CPU cycles (usually less than 80 cycles); develop fast packet capture algorithms (tcpdump-like); run third-party fast path stacks.
Taking a look at the current issues affecting the Web operations and performance space.
Editor’s note: The European edition of our Velocity conference wrapped up a few weeks ago, and now that the jet lag has passed I’ve had a chance to reflect on the talks and excellent hallway conversations I had throughout. And while I thoroughly enjoyed all the sessions I introduced, one of the downsides to being a chair is that I can’t attend all the other sessions at the same time. As such, I always look around for excellent dissections of the conference from other people; this summary by Peter Arijs from CoScale closely reflects some of the themes I saw, including a few of the standout talks.
November in Barcelona was full of action for web and big data practitioners, with the Velocity and Strata-Hadoop conferences and side events such as WebPerfDays and Papis.io. As a startup in the web application monitoring and analytics space, it was the perfect time to get a pulse on the state of the art, and talk to some of our clients and prospects. Below is a summary of personal take-away points from selected Velocity sessions and personal interactions.
Tools to develop massively distributed applications.
Editor’s Note: At the Velocity Conference in Barcelona we launched “A Field Guide to the Distributed Development Stack.” Early response has been encouraging, with reactions ranging from “If I only had this two years ago” to “I want to give a copy of this to everyone on my team.” Below, Andrew Odewahn explains how the Guide came to be and where it goes from here.
As we developed Atlas, O’Reilly’s next-generation publishing tool, it seemed like every day we were finding interesting new tools in the DevOps space, so I started a “Sticky” for the most interesting-looking tools to explore.
At first, this worked fine. I was content to simply keep a list, where my only ordering criteria was “Huh, that looks cool. Someday when I have time, I’ll take a look at that,” in the same way you might buy an exercise DVD and then only occasionally pull it out and think “Huh, someday I’ll get to that.” But, as anyone who has watched DevOps for any length of time can tell you, it’s a space bursting with interesting and exciting new tools, so my list and guilt quickly got out of hand.