What's it like using serverless in production? We answer that question in this Day Two Cloud podcast with guest Robb Schiefer, who's running a live application on AWS Lambda for a healthcare company. We talk about why the organization adopted serverless, how they chose AWS, and how DevOps practices are key for ongoing operations.
The post Day Two Cloud 011: Going All In On Serverless And DevOps appeared first on Packet Pushers.
There are two trends converging in AI inference and so far, only a small number of companies are enmeshed. …
Will Analog AI Make Mythic a Unicorn? was written by Nicole Hemsoth at .
Cooperation has been key to expanding Internet access around the globe. Ten years ago, the African region created AfPIF, a space focused on collaboration about among regional actors on topics related to peering and interconnection. Inspired by that project, in 2014 I approached Bevil Wooding to create a similar space for the Caribbean.
In recent years, the Caribbean has been losing its traditional industries, such as sugar and banana production. In this context, the Internet can be seen as a good opportunity to leverage the local economy. Fortunately, the idea gained the support of the Caribbean Telecommunications Union (CTU) and the Caribbean Network Operators Group (CaribNOG). That’s how the Caribbean Peering and Interconnection Forum (CarPIF) was born.
From its inaugural meeting in 2015, CarPIF has sought to bring together key infrastructure, service, and content providers to improve network interconnection, lower the cost of connectivity, and increase the number of Internet users and services in the Caribbean. This year, the meeting will be held from 12 to 13 June in Grenada, with the aim of highlight the active role played by the Organization of Eastern Caribbean States (OECS) in the successful deployment of Internet Continue reading
The Swedish vendor upgraded its forecasts by an additional 400 million subscribers or 27% as 5G...
Wave Computing was one of the earliest AI chip startups that held significant promise, particularly with its initial message of a single architecture to handle both training and inference. …
AI Chip Startup Makes Training to Edge Inference Transition was written by Nicole Hemsoth at .
Given our focus on the systems-level of AI machine building, storage was a big topic of discussion at the sold-out Next AI Platform event we hosted in May. …
The Challenge Machine Learning Brings To Storage was written by Timothy Prickett Morgan at .
Today is the 5th anniversary of Cloudflare's Project Galileo. Through the Project, Cloudflare protects—at no cost—nearly 600 organizations around the world engaged in some of the most politically and artistically important work online. Because of their work, these organizations are attacked frequently, often with some of the fiercest cyber attacks we’ve seen.
Since it launched in 2014, we haven't talked about Galileo much externally because we worry that drawing more attention to these organizations may put them at increased risk. Internally, however, it's a source of pride for our whole team and is something we dedicate significant resources to. And, for me personally, many of the moments that mark my most meaningful accomplishments were born from our work protecting Project Galileo recipients.
The promise of Project Galileo is simple: Cloudflare will provide our full set of security services to any politically or artistically important organizations at no cost so long as they are either non-profits or small commercial entities. I'm still on the distribution list that receives an email whenever someone applies to be a Project Galileo participant, and those emails remain the first I open every morning.
Five years ago, Project Galileo was born Continue reading
Lack of transparency, false incentives and lack of trus
The post Don’t use Opera browser Because Privacy Threat appeared first on EtherealMind.
In my quest to understand how much buffer space we really need in high-speed switches I encountered an interesting phenomenon: we no longer have the gut feeling of what makes sense, sometimes going as far as assuming that 16 MB (or 32MB) of buffer space per 10GE/25GE data center ToR switch is another $vendor shenanigan focused on cutting cost. Time for another set of Fermi estimates.
Let’s take a recent data center switch using Trident II+ chipset and having 16 MB of buffer space (source: awesome packet buffers page by Jim Warner). Most of switches using this chipset have 48 10GE ports and 4-6 uplinks (40GE or 100GE).
Read more ...Hello my friend,
This article is kind of a special one for me. It doesn’t mean that everything I have written before has a little sense. Everything what I have written about the Data Centre Fabric project was steps towards fully automated data centre operation, and today we make a final step towards the closed-loop automation based using the real-time data analytics by InfluxData Kapacitor.
2
3
4
5
retrieval system, or transmitted in any form or by any
means, electronic, mechanical or photocopying, recording,
or otherwise, for commercial purposes without the
prior permission of the author.
According to the official website, InfluxData Kapacitor is alerting system following publish-subscribe design pattern, which supports both steam and batch data processing. If we translate it from the geeks’ language, it means that Kapacitor can subscribe to a certain to topics in the data producer (e.g., time series database – InfluxDB or collector – Telegraf) and start getting information out of it:
Beyond data and model parallelism for deep neural networks Jia et al., SysML’2019
I’m guessing the authors of this paper were spared some of the XML excesses of the late nineties and early noughties, since they have no qualms putting SOAP at the core of their work! To me that means the “simple” object access protocol, but not here:
We introduce SOAP, a more comprehensive search space of parallelization strategies for DNNs that includes strategies to parallelize a DNN in the Sample, Operator, Attribute, and Parameter dimensions.
The goal here is to reduce the training times of DNNs by finding efficient parallel execution strategies, and even including its search time, FlexFlow is able to increase training throughput by up to 3.3x compared to state-of-the-art approaches.
There are two key ideas behind FlexFlow. The first is to expand the set of possible solutions (and hence also the search space!) in the hope of covering more interesting potential solutions. The second is an efficient execution simulator that makes searching that space possible by giving a quick evaluation of the potential performance of a given parallelisation strategy. Combine those with an off-the-shelf Metropolis-Hastings MCMC search strategy and Bob’s your uncle.