PyTorch-BigGraph: a large-scale graph embedding system

PyTorch-BigGraph: a large-scale graph embedding system Lerer et al., SysML’19

We looked at graph neural networks earlier this year, which operate directly over a graph structure. Via graph autoencoders or other means, another approach is to learn embeddings for the nodes in the graph, and then use these embeddings as inputs into a (regular) neural network:

Working with graph data directly is difficult, so a common technique is to use graph embedding methods to create vector representations for each node so that distances between these vectors predict the occurrence of edges in the graph.

When you’re Facebook, the challenge in learning embeddings is that the graph is big: over two billion user nodes, and over a trillion edges. Alibaba’s graph has more than one billion users and two billion items; Pinterest’s graph has more than 2 billion entities and over 17 billion edges. At this scale we have to find a way to divide-and-conquer. We’ll need to find some parallelism to embed graphs with trillions of edges in reasonable time, and a way of partitioning the problem so that we don’t need all of the embeddings in memory at each node (‘many standard methods exceed the memory Continue reading

BrandPost: Digital Transformation: IT Leaders Discuss SD-WAN’s Emerging Role

The digital economy is here, and the ability to innovatively meet ever evolving customer expectations has become the new reality.  For most organizations accomplishing this goal starts with digital transformation.In order to realize digital transformation benefits, companies need to move beyond a legacy network and invest in next generation components, according to participants in a recent IDG TechTalk Twitter chat.Unfortunately, past investments in legacy technology are still serving as a significant road block. A1) Issues such as technical debt, legacy security problems, poor/inconsistent Ops & maintenance, and a long list of infrastructure/bandwidth inadequacies can hold back #DigitalTransformation. #IdgTechTalkTo read this article in full, please click here

Automation Workflow Patterns

Workflows vary from seriously simple to notoriously complex and as humans, we might not even consciously observe the subtleties of what a workflow comprises of. Workflows are the source of control semantics and comprise of many elements, some obvious some not so. This post is a primer to help you think about the kind of workflows you encounter drawn from my experiences. This post offers a view with conviction backed by experience.

To set the tone, workflows have logical flow, temporal behaviour, consume and transmit data, for processing triggers, acting on decision points and returning states. Since the 1970s, I believe we haven’t actually come that far from a workflow orchestration standpoint. Atomic units of code exist that do one thing well, a real win for the 1970s and good automation systems understand how to instantiate, feed these atomic blobs of logic data and grab their exit state and content. On a *nix system, it’s possible to use bash to create a single chain of tasks using the | operator. One blob of logic effectively feeds it’s output to the next blob of logic. Who needs an orchestrator? It’s sensible to include detection logic within each blob of code to Continue reading

At A Glance: The Mid-Atlantic + Government Docker Summit

Last week, Docker hosted our 4th annual Mid-Atlantic and Government Docker Summit, a one-day technology conference held on Wednesday, May 29 near Washington, DC. Over 425 attendees in the public and private sector came together to share and learn about the trends driving change in IT from containers, cloud and DeVops. Specifically, the presenters shared content on topics including Docker Enterprise, our industry-leading container platform, Docker’s Kubernetes Service, Container Security and more.

Attendees were a mix of technology users and IT decision makers: everyone from developers, systems admins and architects to Sr. leaders and CTOs.

Summit Recap by the Numbers:
  • 428 Registrations
  • 16 sessions
  • 7 sponsors
  • 3 Tracks (Tech, Business and Workshops)
Keynotes

Highlights include a keynote by Docker’s EVP of Customer Success, Iain Gray, and a fireside chat by the former US CTO and Insight Ventures Partner, Nick Sinai, and current Federal US CIO, Suzette Kent.

The fireside highlighted top of mind issues for Kent and how that aligns with the White House IT Modernization Report; specifically modernization of current federal IT infrastructure and preparing and scaling the workforce. Kent mentioned, “The magic of IT modernization is marrying the technology with the people and the Continue reading

Heavy Networking 454: Analyzing Encrypted Traffic In The TLS 1.3 Era With ExtraHop (Sponsored)

Today's Heavy Networking examines packet analysis with sponsor ExtraHop. We drill into the company's marketing claims about deep analysis at line rate with Mike Ernst, VP of Sales Engineering. We also tackle how ExtraHop handles encrypted traffic, incuding TLS 1.3 and Perfect Forward Secrecy. Mike promises to keep his inner salesperson on mute for this conversation.

The post Heavy Networking 454: Analyzing Encrypted Traffic In The TLS 1.3 Era With ExtraHop (Sponsored) appeared first on Packet Pushers.

Weekend Reads 060619

When most people speak of exascale supercomputers, they tend to focus on the computational aspect of these systems. That’s certainly understandable inasmuch as exascale has been generally equated with performing calculations at the level of exaflops. —Michael Feldman

Although it was originally developed by Google, Kohn said Kubernetes has been built on a decade of learning and incorporates a host of technologies from other companies. “There is no vendor lock-in,” he said. “Google has encouraged developers from other companies to take part.” —Cliff Saran

The exposed areas needing improvement by ICANN — both the organization as well as its global community of stakeholders — provide a target-rich environment. —Greg Thomas

The market-dominating behavior of four of the biggest tech companies now appear to be in the crosshairs of the U.S. government, but the big question for consumers and investors is, will anything actually change this time around? —Therese Poletti

On the one hand, testing is all too often the roadblock that stands between highly accelerated Dev processes and highly automated ops-driven delivery processes. But on the other hand, testing is essential for ensuring that the release doesn’t place the business at risk — undermining the very “customer experience” Continue reading

New switches, Wi-Fi gear to advance Arista’s campus architecture

Arista is rolling out more products and services in its continued assault on both the campus network and enterprise hybrid-cloud environments.In particular, the company is readying a new family of what it describes as its first purpose-built campus leaf switches as well as a Wi-Fi 6 (802.11ax) access point that fit into its overall grand plan called Cognitive Campus, with which the company says customers can more easily automate deployment, configuration, troubleshooting and deploying security. [ Read also: How to plan a software-defined data-center network ] Arista is also fashioning an alliance with Microsoft to better support enterprise use of hybrid cloud.To read this article in full, please click here

New switches, Wi-Fi gear to advance Arista’s campus architecture

Arista is rolling out more products and services in its continued assault on both the campus network and enterprise hybrid-cloud environments.In particular, the company is readying a new family of what it describes as its first purpose-built campus leaf switches as well as a Wi-Fi 6 (802.11ax) access point that fit into its overall grand plan called Cognitive Campus, with which the company says customers can more easily automate deployment, configuration, troubleshooting and deploying security. [ Read also: How to plan a software-defined data-center network ] Arista is also fashioning an alliance with Microsoft to better support enterprise use of hybrid cloud.To read this article in full, please click here

Home on the Palo Alto Networks Cyber Range

You’ve probably heard many horror stories by now about the crazy interviews that companies in Silicon Valley put you though. Sure, some of the questions are downright silly. How would I know how to weigh the moon? But the most insidious are the ones designed to look like skills tests. You may have to spend an hour optimizing a bubble sort or writing some crazy code that honestly won’t have much impact on the outcome of what you’ll be doing for the company.

Practical skills tests have always been the joy and the bane of people the world over. Many disciplines require you to have a practical examination before you can be certified. Doctors are one. The Cisco CCIE is probably the most well-known in IT. But what is the test really quizzing you on? Most people will admit that the CCIE is an imperfect representation of a network at best. It’s a test designed to get people to think about networks in different ways. But what about other disciplines? What about the ones where time is even more of the essence than it was in CCIE lab?

Red Team Go!

I was at Palo Alto Networks Ignite19 this past Continue reading

Hong Kong Chapter: Why Aren’t There More Women in Tech?

The tech industry in Hong Kong and across the world remains male dominated. Why aren’t there more women and what can be done to fix this?

To mark International Girls in ICT Day, which aims to encourage girls and young women to work in information and communications technology, the Internet Society Hong Kong Chapter organized an event to tackle these questions. Ladies X Tech X Gents: How Are the Three Compatible? brought together four successful developers to lead the dialogue:

  • Ivy Luk, Sales Engineer at Clare.AI (an Artificial Intelligence digital assistant solutions provider)
  • Emma Wong, Organiser of Google Developer Group and Women Techmakers Hong Kong
  • May Yeung, Director of Internet Society Hong Kong Chapter
  • Rick Mak, Co-Founder of Oursky (a web and mobile application development company)

Why are there so few women in the tech industry?

A common observation among the speakers was the high dropout rate of women developers in the tech industry – amid the already low women to men ratio. The speakers noted that it drops from roughly 3:7 at school to 1:10 at work.

One of the main reasons women leave the tech industry is the gender stereotype that it is a masculine profession. Continue reading

As Expected: Where Have All the SDN Controllers Gone?

Roy Chua (SDx Central) published a blog post titled “Where Have All the SDN Controllers Gone” a while ago describing the gradual disappearance of SDN controller hype.

No surprise there - some of us were pointing out the gap between marketing and reality years ago.

It was evident to anyone familiar with how networking actually works that in a generic environment the drawbacks of orthodox centralized control plane SDN approach far outweigh its benefits. There are special use cases like intelligent patch panels where a centralized control plane makes sense.

Read more ...

Towards federated learning at scale: system design

Towards federated learning at scale: system design Bonawitz et al., SysML 2019

This is a high level paper describing Google’s production system for federated learning. One of the most interesting things to me here is simply to know that Google are working on this, have a first version in production working with tens of millions of devices, and see significant future expansion for the technology (‘we anticipate uses where the number of devices reaches billions’).

So what exactly is federated learning?

Federated Learning (FL) is a distributed machine learning approach which enables training on a large corpus of decentralized data residing on devices like mobile phones. FL is one instance of the more general approach of “bringing the code to the data, instead of the data to the code” and addresses the fundamental problems of privacy, ownership, and locality of data.

Note that this beyond using an on-device model to make predictions based on local data. Here we’re actually training the model in a distributed fashion, using data collected on the devices, without the data ever leaving those devices. The FL system contains a number of privacy-enhancing building blocks, but the privacy guarantees of any end-to-end system Continue reading

1 … 1,207 1,208 1,209 1,210 1,211 … 3,835