Oracle launches GraphPipe for deploying machine learning models

Oracle wants to address machine learning problems with its newly announced open-source project GraphPipe. The project is a “dead simple machine learning model serving” solution.

“There has been rapid progress in machine learning over the past few years. Today, you can grab one of a handful of frameworks, follow some online tutorials, and have a working machine learning model in a matter of hours. Unfortunately, when you are ready to deploy that model into production you still face several unique challenges,” Vish Abrams, architect for cloud development at Oracle, wrote in a post.

According to the company, there are three challenges when it comes to deploying a machine learning model into a project:

  1. Lack of a standard for model serving APIs
  2. Inability to easily build a model server
  3. Not enough focus on performance

“We created GraphPipe to solve these three challenges. It provides a standard, high-performance protocol for transmitting tensor data over the network, along with simple implementations of clients and servers that make deploying and querying machine learning models from any framework a breeze,” Abrams wrote.

GraphPipe supports TensorFlow, PyTorch, mxnet, CNTK and caffe2 models.

Oracle explained that businesses normally train and deploy machine learning models individually with “bespoke” approaches. “This impacts an organizations’ ability to derive value from its machine learning efforts. If marketing wants to use a model produced by the finance group, they will have to write custom clients to interact with the model. If the model becomes popular sales wants to use it as well, the custom deployment may crack under the load,” Abrams explained. The standard aims to provide the tools necessary to derive value from investments.

In addition, the project includes: a set of flatbuffer definitions, guidelines for serving models, examples for serving models, and client libraries for querying models. “In essence, a GraphPipe request behaves like a TensorFlow-serving predict request, but using flatbuffers as the message format. Flatbuffers are similar to google protocol buffers, with the added benefit of avoiding a memory copy during the deserialization step,” wrote Abrams.

The flatbuffer spec is available on GitHub with clients and servers for implementing Python and Go. The team plans on adding a client for Java soon as well as a TensorFlow plugin.

The post Oracle launches GraphPipe for deploying machine learning models appeared first on SD Times.

via Click on the link for the full article

Google updates Firebase for app experiences

Google’s mobile development platform Firebase is getting a number of updates today focused on building better apps, improving app quality and growing businesses. Firebase

“At Firebase, our mission is to help mobile app teams succeed, which means having the capabilities to support companies and teams of all sizes and complexity. In the last couple of years, we’ve matured significantly, from a realtime database to a full mobile app development platform. Firebase is built on top of Google Cloud, so you get all the technical scale, enterprise-grade control and management, and machine learning strength that underpins many of Google’s products,” the company wrote in a post.

For businesses, the company announced Firebase In App Messaging, improved reporting for the cloud messaging console, and change history for Remote Config. The in app messaging solution is a new tool for guiding active app users to targeted and contextual messages. “Now, you’ll be able to communicate with your most valuable users – the ones already interacting with your app – and deepen engagement with them by surfacing relevant information, offers, and tips as they use your app!” the company wrote.

As part of its mission to improve app quality, Google is adding Crashlytics integrations for productivity. Crashlytics integrations are designed to help users keep track of all the different tools they are using within their workflow. The integrations include the ability to export Crashlytics data from Firebase to BigQuery, and integration with JIRA for creating JIRA issues based on crashes. “Combined with the existing integration with Slack, your team can now track the crashes they are working on, with tools they already use,” the team wrote.

Lastly, in order to build better app experiences, the company announced a serverless backend, ability to host and update multiple websites, and an improved console. The serverless backend enables users to easily scale up their databases and takes advantage of the recently released Cloud Functions.

The post Google updates Firebase for app experiences appeared first on SD Times.

via Click on the link for the full article

Report: Python closes in on C++

For the first time in TIOBE Index history, Python is steadily approaching the top three programming languages. The top three TIOBE Index programming languages are Java, C, and C++.

“Programming language Python is getting very close to the top 3 of the TIOBE index. If Python surpasses C++ and becomes number 3, this will be an all time high for the scripting language of Guido van Rossum,” TIOBE wrote in a post.

According to TIOBE, over the last decade interest in Python has been growing. A 2005 study on the most popular programming language taught in US universities found Java to be the winner with 60 percent of the share of introductory programming courses. However, a recent 2014 study found Python was now the most taught language with 70 percent of the market share.

“This Python boost is also visible in the TIOBE index. But industry is adopting Python as well. The Python programming language started as a successor of Perl to write build scripts and all kind of glue software. But gradually it entered also other domains. Nowadays it is quite common to have Python running in large embedded systems. So it is very likely that Python will enter the top 3 and even might become the new number 1 in the long run,” TIOBE wrote.

Other key findings from the August 2018 TIOBE Index included Hack, Groovy and Julia are back in the top 50 and TypeScript interest is waning at position 62 now.

The post Report: Python closes in on C++ appeared first on SD Times.

via Click on the link for the full article

SD Times Open-Source Project of the Week: Fizz

In order to implement the new generation of Transport Layer Security, TLS 1.3, at Facebook, the company built a TLS library in C++ 14 called Fizz. Earlier this week, Facebook announced it was open sourcing that library.

TLS 1.3 added several new features to make Internet traffic more secure, such as encrypting handshake methods, redesigning how secret keys are derived, and a zero round-trip connection setup.

“We are excited to be open-sourcing Fizz to help speed up deployment of TLS 1.3 across the internet and help others make their apps and services faster and more secure,” Facebook wrote in a post.

According to Facebook, Fizz adds several implementation features such as support for asynchronous I/O by default. It also implements scatter/gather I/O which allows users to pass in chunked data, which Fizz will then encrypt in place, eliminating the need for extra copies of data.

In addition, it supports APIs that allow it to send data as soon as a TCP connection is established, reducing the latency of requests.

Fizz was built with security in mind. According to the company, the TLS state machine is complex and poses security challenges. In Fizz, the state machine is explicit, meaning that transitions are defined in a single place based on the messages that are received. Having everything defined in a single place makes it easier to address security issues, the company explained.

Fizz handles millions of TLS handshakes every second, while reducing both the latency and CPU utilization of services that perform trillions of requests per day, the company explained.

“Today, more than 50 percent of our internet traffic is secured with TLS 1.3, and that will continue to grow as browsers and apps add support for TLS 1.3. RFC 8446 will be published very soon, making TLS 1.3 an internet standard. Even before the RFC is published, we’re happy to show that TLS 1.3 has been successfully deployed at scale. With Fizz we’ve built a robust implementation of the next generation of TLS, and we’re excited to share it with the community so that it can be used in mobile apps, services, and load balancers. We hope the community will use Fizz and also contribute to its further evolution,” the company wrote in a post.

The post SD Times Open-Source Project of the Week: Fizz appeared first on SD Times.

via Click on the link for the full article

Prometheus becomes second project to graduate from CNCF incubation

The Cloud Native Computing Foundation (CNCF) has announced the systems and service monitoring framework Prometheus from the foundation’s incubation program has now graduated. This makes it the second project to meet the CNCF’s criteria for graduation after Kubernetes in March of this year.

The news was announced today during the organization’s Prometheus-focused PromCon event in San Francisco. The open-source infrastructure monitoring platform’s new status means that it has met the CNCF’s requirements that the project “demonstrate thriving adoption, a documented, structured governance process, and a strong commitment to community sustainability and inclusivity,” and adheres to the CNCF Code of Conduct, among other accomplishments, according to the foundation.

Prometheus was started in 2012 by developers at SoundCloud to monitor their burgeoning microservice infrastructure and has since become a rapidly growing part of the Kubernetes ecosystem, the CNCF says. It entered incubation at the CNCF in 2016 with its version 1.0 release.

“Since its inception in 2012, Prometheus has become one of the top open-source monitoring tools of choice for enterprises building modern cloud native applications,” said Chris Aniszczyk, COO of the CNCF in the announcement. “Since being accepted as the second project in CNCF, Prometheus has cultivated an active developer and user community, giving the TOC full confidence to graduate the project. As a testament to its maturity, we’re thrilled to see the Prometheus community spin out OpenMetrics, which takes the Prometheus exposition format and works to evolve it as a de facto industry specification.”

With 20 active maintainers, more than 1,000 contributors, popularity among major companies like DigitalOcean and Uber and integration with Kubernetes, the CNCF and CoreOS team, frequent contributors to the project expect Prometheus to continue growing in popularity.

“[GitHub] stars are an imperfect metric, but they do give a good coarse grained measurement for the popularity of an open source project,” software developer Frederic Branczyk at CoreOS wrote in a blog post congratulating the Prometheus team. “Over the years Prometheus has grown in popularity and this metric reflects that popularity. Within the last two years it grew from 4,000 stars to 18,000 stars on GitHub; even though this is a popularity metric, it does show the rising interest in the project.”

The post Prometheus becomes second project to graduate from CNCF incubation appeared first on SD Times.

via Click on the link for the full article

Agile 2018: The Business Agility Report, the Professional Scrum Master II training course, Scrum@Scale, and Zoho Sprints

Despite Agile’s wide adoption, businesses are still struggling to adopt the mindset, structure and practice of the software development approach. Agile Transformation and the Business Agility Institute announced the first Business Agility Report at the Agile 2018 conference in San Diego this week, which found a majority of organizations still rate their business agility maturity as low, and are struggling with culture and processes.  However, despite the challenges, businesses noted they are experiencing tangible benefits from business agility such as speed to market, customer satisfaction, collaboration, and increased employee and customer satisfaction.

The report looked at business agility fluency to understand what organizations are doing and how they are performing, top competencies, lowest competencies, how company size correlates to business agility, and perception of business agility.

Other announcements from the conference included:

Scrum.org newest Scrum training course
Scrum.org announced the Professional Scrum Master II training course at the conference, its newest and more advanced Scrum Master Course. The 2-day course is designed for Scrum Master with at least a year of experience and the understanding how of to use Scrum for value, quality, productivity and user satisfaction.

The course is meant to improve knowledge, skill and ability to support teams with new experiences and capabilities.

“Taking on the role of the Scrum Master is a challenging one,” said Ken Schwaber, Scrum co-creator and Scrum.org founder. “We have spent years training people to become better Scrum Masters, and now it’s time to take what we’ve learned to help them continue improving. We are always inspecting and adapting with Scrum, and Scrum Masters do the same, gaining new techniques, insights and ideas. There is always room to grow as both individuals and Scrum Masters, and it is part of our mission to help them do so successfully.”

CollabNet VersionOne announced 13th annual State of Agile Survey
CollabNet VersionOne is officially opening up withs State of Agile survey, designed to gain insight on agile software development and adoption. The survey will be open until December 1, 2018.

“Technology professionals world round look to the State of Agile Report each year for insights and perspective on software development, methodologies and practices surrounding agile as well as understanding of the challenges faced by organizations,” said Flint Brenton, CEO at CollabNet VersionOne. “This resource adds incredible value to the industry. We invite individuals to share their experiences and by doing so, help contribute to the software development community.”

The Scrum@Scale Guide
Scrum@Scale showcased its guide to transforming every aspect of an organization. Scrum@Scale is a new framework that is meant to scale the core of the originally Scrum framework and extend it to tens of teams, hundreds of teams to even thousands of teams.

“Scrum@Scale is really designed for business agility. It is designed to go across the entire organization. It is the way the venture group I work with builds scaling into their business and into every investment that they work in,” Jeff Sutherland, co-creator of Scrum, in a video.

Zoho highlights Zoho Sprints
Zoho revealed the latest addition to its platform, Zoho Sprints. This addition is meant to be a agile project management tool for Scrum teams to plan work, track progress and build products.

Features include a backlog, board, epics, timesheets, dashboard, reports, meetings and feed. “Create new sprints and backlog items with a fluid drag-and-drop planning center. Break down the nuances of each user story through epics, estimation points, priority listing, and user assignments. Set durations and timely reminders make releasing after every sprint that much easier,” the company wrote on its website.

The post Agile 2018: The Business Agility Report, the Professional Scrum Master II training course, Scrum@Scale, and Zoho Sprints appeared first on SD Times.

via Click on the link for the full article

SD Times news digest: GitLab’s open-source load balancer, Sauce Labs’ Continuous Testing Services, and Eggplant Release Insights

GitLab has announced that its load balancer, GLB Director, is now open source. According to the company, GLB Director is a Layer 4 load balancer that can scale an IP address across a large number of physical machines while minimizing connection disruption during changes in servers.

GitHub hopes that by open-sourcing the solution, others will be able to benefit and “contribute to a common standard software load balancing solution that runs on commodity hardware in physical datacenter environments.”

Sauce Labs launches Continuous Testing Services
Sauce Labs has launched Continuous Testing Services in order to help organizations on their journey to automation and continuous testing. As part of this release, it is also launched a 6-8 week program for teams that are starting to adopt continuous testing called Sauce Start.

According to Sauce Labs, Sauce Start is targeted at companies that have little or no test automation in place; little or no Selenium or Appium experience; have test automation in place, but lack integration with their continuous integration architecture; or are using legacy test automation solutions and are looking to move to modern tools.

Eggplant launches Eggplant Release Insights
Eggplant has launched Eggplant Release Insights to help organizations visualize the impact on user satisfaction and business outcomes if they were to release another version of their product now.

Eggplant Release Insights works together with Eggplant AI to provide an extensive set of comprehensive predictors to determine release quality, such as ones for bug content, development quality, test coverage, and usability quality.

“DevOps teams are drowning in data but, in this world of continuous everything, they lack the ability to easily predict outcomes and the effect on the business,” said Antony Edwards, CTO of Eggplant. “Optimizing products to drive growth is essential for every organization and our “release rating” will provide teams with real time insights to immediately understand the quality and the impact on the user before release. This further cements Eggplant’s leadership in deploying AI, machine learning, and analytics today to optimize the customer experience and accelerate the pace of DevOps.”

Revulytics adds new feature tracking and reporting capabilities
Revulytics has launched new feature tracking and reporting capabilities in its analytics offering Lifetime Feature Usage. The update enables every feature tracked in Usage Intelligence to be filtered based on environmental properties as well as usage by particular audiences.

Other improvements to Usage Intelligence include Sunburst visualizations, ReachOut delivery caps, regular expressions, a date installed filter, and a date last seen filter.

“Our mission is to increase the power of product usage analytics to help our customers build better products by answering more questions about how users interact with their application. The introduction of Lifetime Feature Usage analysis extends our existing set of feature usage reports, giving our customers a deeper understanding of how specific user groups behave throughout their lifetime and reasons for churn,” said Keith Fenech, VP of Software Analytics. “Additionally, customers can better target ReachOut in-application messaging campaigns to influence that behavior and increase feature adoption and retention.”

The post SD Times news digest: GitLab’s open-source load balancer, Sauce Labs’ Continuous Testing Services, and Eggplant Release Insights appeared first on SD Times.

via Click on the link for the full article

Agile 2018: Data is a people problem

Businesses are looking for new ways to collect, manage and analyze data to make better and more informed decisions, but they are missing the mark. Data is not just about the tools. The number one problem to using data to your advantage is people, according to Troy Magennis, an Agile consultant and founder of Focused Objectives who made a keynote presentation at the 2018 Agile conference in San Diego this week.

According to Magennis, you can collect all the data in the world and make charts and dashboards, but if people don’t understand what you are trying to tell them — it is not going to make a difference.

Magennis opened the keynote with an Agile community call to action: “Having the data isn’t enough. You have to tell the story that leads to the right action.”

That means the way in which we present data in order to get understanding conveyed is very important. “Stop using boring plots,” Magennis stated. “That is not going to convince anything to anyone.” Some ways to do that is to present data in a visually compelling way, and compare it to other scenarios that might make it easier to understand.

Other call to actions Magennis expressed throughout the keynote included:

  • Businesses needs to find safer ways to capture and utilize data from teams and organizations
    • Safe = better data and happier people
  • Businesses needs to find better ways to measure outcome rather than output
  • Businesses need to balance our predictability expectations with creativity and novelty
  • Businesses need to incorporate new information faster
    • Deployment ends speculation

The point Magennis kept stressing is that there needs to be a conversation around data, and so far the industry has been bad at having that conversation. Magennis himself has been subject to bad data practices such as predictability, trying to plan better, and sticking to a narrow range for expectations. That isn’t a good business process because the real value lies in the unpredictability, he explained. “We need to be able to accept in a creative world of doing products and innovations that we are entering uncertain territory,” he said. “We are going to have to accept there is a lot we don’t know and plan with a high degree of uncertainty.”

Organizations are still just in the beginning of this journey of understanding how to do data and how to do it well, and businesses need to move to measuring value and outcomes rather than output, according to Magennis. In order to learn and learn fast, businesses have to be able to accept bad information too, meaning uncertain information that may set you back, but may provide more value than “good information.” Uncertainty is a strategic weapon and being able to embrace unpredictability is a bold movement, he explained.

“You don’t learn as much from an expected outcome as [you do] with an unexpected outcome.”

Forecasting and estimates should not tell you if you are right, they should detect if you are wrong and discover what you missed.

Some best practices include experimenting, using the most recent 10 samples when forecasting with historical velocity or throughput data and starting early. “Forecasting is about knowing when to start,” he explained.  

Other points he made included:

  • “Give multiple options of investment and implementation
  • Give a duration and define what started and complete means
  • If the original data is in doubt, know earlier and react faster”

The post Agile 2018: Data is a people problem appeared first on SD Times.

via Click on the link for the full article

The Dart programming language reaches version 2

Google is trying to revive the Dart programming language as a mainstream language for mobile and web development with its latest announcement. Dart 2 features a complete rewrite of the web platform with a strong focus on productivity, performance and scalability.  

“Dart 2 marks the rebirth of Dart as a mainstream programming language focused on enabling a fast development and great user experiences for mobile and web applications,” Kevin Moore, product manager for Google, wrote in a post. “We want to enable developers building client applications to be productive, with a language, framework and components that reduce boilerplate and let them concentrate on the business logic, along with tooling that identifies errors early, enables powerful debugging and delivers small, fast runtime code.”

Google created Dart with a focus on client-side development, and the ability to create high-quality, mission critical apps for mobile and the web.

According to Moore, Dart has seen great growth over the last year with a “ten-fold growth in external usage,” and an increase in pull requests on GitHub.

A lot of the work in Dart 2 went into cleaning up the language, adding more support for web and mobile frameworks, and providing more tools and components to support Dart outside of Google.

One of the features ensures developers “can’t get into a state where an expression evaluates to a value that doesn’t match the expression’s static type,” Moore wrote. This will enable developers to scale their codebase, work on large refactoring projects and better deploy code, Moore went on to explain.

For the language’s framework, Moore said Dart 2 provides faster incremental builds and smaller compiled JavaScript with the help of the language’s sound type system and new build system. In addition, the team worked on reducing the “time to interactivity” on web pages by improving the amount of code being processed during page load.

Version 2 also embraces a JavaScript development model with JavaScript compilers dartdevc and dart2js. Dartdevc is designed for faster incremental compilation while dart2js focuses on optimizing JavaScript for progressive web apps.

Other key features include an improved package manager, a static analyzer, and a linter that can be consumed from the command line or as a package.  

“Dart has become a general purpose language that is optimized for web and mobile development. We’ve been working on Dart 2 for several years, during which time it has touched every aspect of our ecosystem and required migrating millions of lines of code and hundreds of packages,” Moore wrote.

The post The Dart programming language reaches version 2 appeared first on SD Times.

via Click on the link for the full article

Analyst View: Scope, silos stifle software innovation

Software innovation can be constrained by scope limitations, and the barriers created by artificial silos in existing systems and organizations. But help is at hand. This article looks at two very different approaches — the application of systems engineering, and customer-centric thinking.

Legacy sets limits
In the early days of a software development project, there are visionary ideas, novel approaches, and exciting goals. But the cold light of reality can set in at any moment — available resources and required delivery dates can be defined at a very early stage. The project adapts its thinking to meet these constraints.

Both traditional and agile methods allow the project team to keep faith with their vision, and get on with the ‘core’ (to match the budget and schedule), while recording everything else on the WIBL (Wouldn’t It Be Lovely…) list, which hopefully will feed into some later phase of the project. So the big idea which got things started isn’t lost, it just gets a bit fragmented.

Sometimes there are legacy issue which can be even more of a constraint — database schemas which no one wants to touch, or perceived turf wars, or tough technical mountains.

These issues create even harder constraints than resources and timescale. Project teams adapt their thinking. It’s not easy to think of what should go on the WIBL list. The easiest response is to redefine the project to stay entirely within the scope of control of the project team. The big-idea can suffer.

Systems engineering
Systems engineering got bad press because in the early days it was a documentation-heavy methodology, which fitted a big-budget project and a waterfall approach to development. The systems engineers had to spend up-front time on extensive requirements and systems architecture documents. More “ready-aim-aim-aim-aim-…” than the agile “ready-aim-fire-what have we got-let’s go again” approach.

The early days are over, and systems engineering is back center stage. “Model based systems engineering” (MBSE) replaces documentation with models. Tools can guarantee model consistency. Some models allow component performance estimates to be attached, enabling ‘execution’ of the model to simulate system performance and visualize some aspects of function. This approach fits agile projects.

But in relation to innovation, the key role of systems engineers is in setting scope. As always, it remains the duty of the systems engineer to consider all the relevant domains which impact the system being worked on by the team. Systems engineers must optimize and prioritize requirements from all these domains. Of course, the central domain is the function and performance of this system itself. But, for a systems engineer, the domain of software development processes is also in scope, as are the domains of delivery, provisioning, deployment, operations and maintenance environments for this system.

So the system engineer has a chance to see and articulate the legacy and silo issues which constrain the innovation potential of the project. Of course this does not solve the problem, but at least it brings the issues to the surface for reasoned consideration.

Customer-centric thinking
Let’s assume that the customer for the results of the development project will be a user who pays for the right to use the software. In this case, true customer-centric thinking means that the project team should consider the whole customer experience.

This means checking for gaps in requirements by looking at the whole lifecycle experience for the customer — what will the customer see and do at every stage — discover, investigate, buy, deploy, use, maintain and upgrade the new software? This is quite a long list, especially when we all know that handling just ‘use’ can seem challenging.

So, just like systems engineering, this doesn’t solve the problem. But handling the other steps will probably mean talking to other groups in your own organization. These conversations will help identify the constraints that threaten the big idea. And who knows, your colleagues may find time, budget, resources and perhaps other ways to support the project’s full innovation opportunity.

The post Analyst View: Scope, silos stifle software innovation appeared first on SD Times.

via Click on the link for the full article