投稿時間:2021-12-02 05:36:06 RSSフィード2021-12-02 05:00 分まとめ(44件)

カテゴリー等	サイト名等	記事タイトル・トレンドワード等	リンクURL	頻出ワード・要約等/検索ボリューム	登録日
AWS	AWS Architecture Blog	Top 5: Featured Architecture Content for November	https://aws.amazon.com/blogs/architecture/top-5-featured-architecture-content-for-november/	Top Featured Architecture Content for NovemberThe AWS Architecture Center provides new and notable reference architecture diagrams vetted architecture solutions AWS Well Architected best practices whitepapers and more This blog post features some of our best picks from November s new and updated content Game Industry Lens for the AWS Well Architected Framework This new Lens specifies best practices for the unique characteristics …	2021-12-01 19:21:49
AWS	AWS Database Blog	Configure an AWS SCT multi-server project	https://aws.amazon.com/blogs/database/configure-an-aws-sct-multi-server-project/	Configure an AWS SCT multi server projectThe AWS Schema Conversion Tool AWS SCT accelerates migration for commercial database and data warehouse schemas along with code objects to open source engines or AWS native services such as Amazon Aurora and Amazon Redshift For your database migration to the cloud you can choose a rehost lift and shift replatform lift and reshape or refactor approach …	2021-12-01 19:25:19
AWS	AWS Machine Learning Blog	Evolution of Cresta’s machine learning architecture: Migration to AWS and PyTorch	https://aws.amazon.com/blogs/machine-learning/evolution-of-crestas-machine-learning-architecture-migration-to-aws-and-pytorch/	Evolution of Cresta s machine learning architecture Migration to AWS and PyTorchCresta Intelligence a California based AI startup makes businesses radically more productive by using Expertise AI to help sales and service teams unlock their full potential Cresta is bringing together world renowned AI thought leaders engineers and investors to create a real time coaching and management solution that transforms sales and increases service productivity weeks after application deployment Cresta …	2021-12-01 19:26:58
AWS	AWS Startups Blog	Koo App Connects Millions of Voices in Their Preferred Language with AWS	https://aws.amazon.com/blogs/startups/koo-app-connects-millions-of-voices-in-their-preferred-language-with-aws/	Koo App Connects Millions of Voices in Their Preferred Language with AWSWhen the social media revolution began e commerce sites mostly catered to English speakers which left out a huge population of would be participants Koo a microblogging platform based in India noted the lack of inclusivity and made it their mission to create an app that is accessible to the entire spectrum of languages spoken in India Koo started with just three languagesーEnglish Hindi and Kannadaーand expanded from there Learn how this ambitious startup used AWS to scale and give a voice to millions of users	2021-12-01 19:59:25
Program	[全てのタグ]の新着質問一覧｜teratail（テラテイル）	メール認証による会員登録機能で、メール登録後の画面で会員登録用URLを表示する理由がわからない	https://teratail.com/questions/371895?rss=all	会員登録	2021-12-02 04:40:24
技術ブログ	Developers.IO	[速報] 機械学習を身近に Amazon SageMaker Studio Lab が発表されました	https://dev.classmethod.jp/articles/new-amazon-sagemaker-studio-lab/	amazonsagemakerstudiolab	2021-12-01 19:38:48
技術ブログ	Developers.IO	[速報] Amazon Lex Automated Chatbot Designer が発表されました！(プレビュー) #reinvent	https://dev.classmethod.jp/articles/new-amazon-lex-automated-chatbot-designer/	amazon	2021-12-01 19:26:12
技術ブログ	Developers.IO	[速報] トレーニングデータセット作成をもっと簡単に Amazon SageMaker Ground Truth Plus が発表されました	https://dev.classmethod.jp/articles/new-amazon-sagemaker-ground-truth-plus/	amazo	2021-12-01 19:24:47
技術ブログ	Developers.IO	[速報] 人工知能・機械学習を学ぶ学生のための支援プログラム AWS AI＆ML Scholarship が発表されました #reinvent	https://dev.classmethod.jp/articles/aws-aiml-scholarship/	awsaimlscholarship	2021-12-01 19:18:52
技術ブログ	Developers.IO	[速報] ノーコードで完全に機能するカスタマイズ可能な検索エクスペリエンス Amazon Kendra Experience Builder が発表されました	https://dev.classmethod.jp/articles/new-amazon-kendra-experience-builder/	amazon	2021-12-01 19:07:14
海外TECH	Ars Technica	Omicron found in US—plus 23 other countries in 5 of 6 global regions	https://arstechnica.com/?p=1817319	defeat	2021-12-01 19:43:49
海外TECH	Ars Technica	I like Microsoft Edge. But if it doesn’t get less annoying, I’ll switch again	https://arstechnica.com/?p=1815164	microsoft	2021-12-01 19:16:57
海外TECH	MakeUseOf	What Is Discovery+? Everything You Need to Know	https://www.makeuseof.com/what-is-discovery-plus/	discovery	2021-12-01 19:30:12
海外TECH	MakeUseOf	Facebook May Be Forced to Sell Giphy: Here's Why	https://www.makeuseof.com/facebook-may-force-sell-giphy-why/	giphy	2021-12-01 19:26:53
海外TECH	MakeUseOf	How to Add Software Shortcuts to Windows 11’s Desktop Context Menu	https://www.makeuseof.com/windows-11-add-shortcut-desktop-context-menu/	How to Add Software Shortcuts to Windows s Desktop Context MenuWith just a few registry tweaks you can add shortcuts to Windows s desktop context menu and make it a lot easier to boot up your favorite apps	2021-12-01 19:15:22
海外TECH	MakeUseOf	Respira Labs Announces Sylvee, a Wearable Lung Monitor: What We Know So Far	https://www.makeuseof.com/respira-labs-announces-sylvee-what-we-know-so-far/	Respira Labs Announces Sylvee a Wearable Lung Monitor What We Know So FarA new wearable lung monitor from Respira Labs could bring breakthroughs on the respiratory disease front Here s what we know now	2021-12-01 19:00:37
海外TECH	DEV Community	Chapter: A guide to Docker	https://dev.to/atharvaa/chapter-a-guide-to-docker-36mj	Chapter A guide to Docker What is Docker Docker is a platform that allows developers to containerise build test deploy ship applications much faster within multiple workspaces and delivering production ready applications Docker wraps your application inside an abstraction called container which makes your development workflow quick and easier Why Docker If you need streamlined fast lightweight efficient and orchestrated way to run scale deploy or test your application software on one or more virtualised physical systems Docker is your solution Docker being lightweight and portable makes it a cost effective alternative to hypervisor virtual machines With its simple CLI commands we can easily build delete deploy and manage containerised images on our machines Moreover these containers are highly portable which enables them to be able to run on local systems virtual machines cloud providers or even in hybrid workspaces without worrying about setting up libraries and all other configuration files Docker ArchitectureDocker implements Client Server architecture So what is Client Server architecture Consider online shopping You the client orders a product from app store and the service provider server places the order takes necessary actions and ships it to your address Similarly a client server architecture is an architecture where the client requests a service in this case through Docker CLI and the service provider in this case Docker daemon works on the request and provides desired result Referring the above diagram if user wants to create a container from one of the installed images and therefore enters docker run command docker daemon responds to the request and gives user a configured virtualised container along with its metadata What is Docker Daemon Daemon is a program that runs consistently in the background which responds to particular request or action and is not under any direct control of user that means user cannot change the nature of the program Usually daemon files processes names are suffixed by d for example sshd mysqld dockerd Docker Daemon listens to Docker Client through API requests and manages docker objects like images containers volumes network Docker ClientDocker client is the primary way to interact with Docker When we run a Docker command through terminal the client sends the request to Docker Daemon which is responsible for managing commands Docker ImageDocker Image is built of read only stacked layers generated from instructions inside the image s dockerfile Each layer is a representation of an instruction from the dockerfile Images that are pre built by developers and are available inside Container Registries a place to store and download images One such popular public registry is DockerHub Few companies also have private registries to store their images to install an image in your systemdocker pull lt image name version gt example docker pull ubuntu to see if the image is installeddocker image ls ordocker images to remove an image docker image rm lt image name or image id gt if you get an error message similar to daemon conflict unable to delete then you must delete the container firstread more about containers below to check layers of an imagedocker image inspect lt image name or image id gt DockerfileTo build your customised docker image create a Dockerfile with the extension dockerfile A new image layer is stacked for an instruction you define inside Dockerfile but not every instruction is responsible for creating a layer Following is an example of dockerfile this is a commentFROM ubuntu LABEL org opencontainers image authors org example com COPY appRUN make appRUN rm r HOME cacheCMD python app app py dockignoreIf you have worked with GitHub before you may have used gitignore file to exclude env variables dependency directories temporary files etc Similarly for Docker we have dockignore file to exclude files that are not relevant to the build Before a request from docker clients CLI reaches docker daemon it goes through dockignore file and checks whether there is a context that needs to be excluded before passing to the daemon and thus preventing any large or sensitive information from reaching daemon Docker ContainersContainers are isolated processes which run on a single host machine Containers consists of packages and dependencies required by your application to run on your system We can create start stop delete move and modify containers which is performed inside a thin writable layer know as Container Layer built on top of immutable read only image layers Containers therefore are running instances of an image Each container has its own binaries dependencies and container layer therefore each container being an isolated process make them fast light weight and more efficient to work with Below is an example of Image layers structure P S If user downloads two or more version of the same image docker only builds the layers which are new from previous versions and all the layers which are mutual won t be installed again to build a container from installed imagedocker run lt image name version gt to get list of all running containers metadatadocker ps to get list of all containers metadatadocker ps a to stop a containerdocker stop lt container name or container id gt to restart a containerdocker start lt container name or container id gt to remove a container docker container rm lt container name or container id gt to rename a containerdocker stop lt container name or container id gt docker run name lt your desired container name gt lt image name gt to navigate inside a container s terminaldocker exec it lt container id or container name gt bin bashContainers helps in setting up environment on any OS without worrying about configurations and dependencies and actually focusing on building applications So what s the catch Containers are not persistent Container when deleted or restarted loses all its data and starts again from its image definition So when we delete restart a container the data inside container layer is lost and it starts from a fresh state Now if we try to install restart it again the container layer is created from scratch not having any history of operations we did inside previous one How do we solve this Persistent dataPersisting database is the solution to the above conflict The idea is to store the data inside the hosts filesystem apart from dockers virtual filesystem Basically there are ways to persist the data Volume Bind mounttmpfs mount Linux However Docker recommends the use of Volume as your primary choice for persisting database because of its fair advantages over the other which includes easy and secured migrating and backup between containers and systems flexibility on working on linux and windows user can easily communicate to volumes using Docker Client CLI etc to create a volume docker volume lt volume name gt to list all volumesdocker volume ls to delete a volumedocker volume rm lt volume name gt What next Let s consider you have more than containers on your machine and you want to scale them up configure their networks test deploy and manage them Some organisations even have thousands of containers to manage Managing containers becomes difficult as they scale up So we need a tool that take care of our containers That s where Container Orchestration comes in picture Container Orchestration deploys scales removes checks containers health load balances the traffic and manages all the containers And the most widely used Container Orchestration tool is Kubernetes So now you know what to explore next Connect with me here	2021-12-01 19:49:53
海外TECH	DEV Community	Moving Up In Technical Leadership	https://dev.to/shariq/moving-up-in-technical-leadership-5egl	Moving Up In Technical LeadershipMany paths lead to the top of the mountain that is technical leadership and it all starts with the definition of the term We will define technical leadership as a position where an individual leads a team and focuses on the details of a project along with the overall implementation Today s leaders need to pay attention to the part where a technical leader is still focusing on the details The nitty gritty details are if anything even more so the responsibility of a technical leader than they were of an independent contributor An individual in the role will need to take the on responsibilities of Making sure their team is up to speed on defined best practices Ensuring the quality of code meets defined expectations Training their team to prepare them for the responsibilities that will help progress their careers over a defined career lattice Managing defined stakeholder expectations Delivering a full scale implementation of a feature or product as defined by stakeholders Technical Lead Roles and ResponsibilitiesLeaders must provide structure and definition to be able to effectively support their teams and deliver excellent products and features That s quite a lot of responsibility on the shoulders of one individual isn t it But defining technical leadership isn t the penultimate goal of this post This article is for individuals who have started tackling all of these responsibilities and have embraced the role How does a technical leader move up in their role The answer roundaboutedly lies in these initial responsibilities The most surprising aspect of technical leadership is that it is much less technical than one expects in today s technology spheres Each of the individual responsibilities and how they can be used to move up will be covered below Making Sure Your Team Is Up To Speed on Best PracticesThe responsibility for clearly setting details of a project for the stakeholder fall to the technical lead They will be the main point of contact and will need to be up to date on all the technical measures of said project If a technical lead is not well practiced in the necessary technologies they can not expect any other behavior from their team First the Lead must ensure that they are at an expert level in the skills needed on a project Once that is done ensure that they can constantly work with their team to get them as close to that same level as possible The buck stops with the technical lead If they can maintain their s and their team s expertise it will undoubtedly show that they contain the knowledge to not simply better themselves but those around them There s not a quality I would rank higher than this when it comes to any sort of leadership Ensuring the Quality of Code Meets ExpectationsEnsuring high quality code cannot be overstated and goes hand in hand with the above responsibility of keeping a team up to speed on best practices If a team is well versed in the necessary technologies the chances of higher code quality increase substantially and make the leadership role easier One way in which I learned to tackle these first two responsibilities two was to hold ad hoc tech talks or lunch and learns with an open forum for the team to ask questions relevant to the current project This has a two fold effect in that it helped me learn what the team did not know coupled with allowing me to further support and educate them on those topics It also had the unintended effect of building camaraderie with my team and potentially others at my place of work Leads should always feel free to extend the invitation to other teams that could benefit from the information that you have to share Training Your Team to Prepare Them for the Responsibilities that Will Help Progress Their CareersThe world of software development is increasingly social and career growth also depends on individuals taking full advantage of this fact There are only so many roles that let someone excel at technology and continue onto leadership without embracing a team Technical leaders must learn to individually tailor guidance for each team member The above two topics help with the technical grooming but leads must go further and help team members identify their own weaknesses and strengths and build upon them This is where the most pertinent aspects of leadership are seen Managing Stakeholder ExpectationsCommunication with stakeholders is the most paramount skill that can be attained in the realm of software leadership Whether focusing on an internal project a high profile client or even just an individual project it s key to manage these expectations As developers estimate stories early in their careers they re able to get away with miscalculations and misses on demo day because leaders are taking the heat or because they had already managed the expectations to the stakeholders prior to demo day For leaders in these situations it goes further than estimation they have to begin to look at implementation and delivery on an agreed upon date and timeframe As an expert technical leaders must be able to identify the total effort needed on a project the dates that stories and the final product will be delivered and any roadblocks all at the inception of a project and convince the stakeholder that the team will need the estimated time to deliver Delivering Full Scale Implementations as Agreed Upon with StakeholdersMeeting deadlines with technical excellence in the project code and management all lead to this final responsibility Just like dominos every topic interconnects and can make or break another responsibility The above four responsibilities all tie into the final one of delivering full scale implementations as agreed up with stakeholders If leaders have successfully managed the responsibilities above they will be able to focus on the larger implementation and delivery What it all comes down to is playing these dominos correctly so that each of these responsibilities can be handled expertly But “handling does not mean that leads get to solo the effort and only rely on themselves At this point in a developer s career they ve ideally built up a network and strong team to support their responsibilities Leads manage those around them and make them better all while managing the other flip side of the coin with the stakeholders If leads can trust their teams and thus trust their own ability in keeping their team motivated and skilled and if they can effectively manage client expectations and delivery they will have mastered the tools needed to move up in technical leadership	2021-12-01 19:21:34
海外TECH	DEV Community	TypeScript used to be on my $#!% 💩 list	https://dev.to/shecodez/typescript-used-to-be-on-my-list-42l9	TypeScript used to be on my list JavaScript JS JS a high level ECMAScript interpreted just in time compiled programming language developed by Brendan Eich of Netscape Not to be confused with Java JavaScript was initially created to “make web pages alive well known JS is almost as old as the internetJS is an OOP object oriented programming languageJS is single threadedbut JS is weakly and dynamically typed‍ me Good ol JavaScript what more can I say about one of the worlds most popular scripting language world Learn to TypeScript or GTFO So I started learning TypeScript Not gonna lie it was so painful in the beginning I hadn t been working with JS long enough understand JS well enough or become annoyed enough by all the silly JS errors browser compatibility issues to applicate something like TS Thus TS was very much so on my list but I have since seen the light TypeScript TS TS a strongly typed programming language developed by Microsoft that builds on JavaScript giving you better tooling at any scale Editor ChecksSuddenly debugging was less of a nightmare with editor checks because if TS doesn t like it it wont compile const user id l firstName Nicole lastName Nobles job Web developer console log user name JS doesn t know to alert me that user doesn t have a name property but with TS I get an error Property name does not exist on type firstName string lastName string job string Auto completeNo more guessing what properties are on an object InterfacesIf I made a User interface I can t add remove any required properties to the object Note adding after a property makes it optional interface User id number firstName string lastName string job string remote boolean function updateUser id number update Partial lt User gt const data getUser id const user data update city Hachiōji saveUser id user TS will rightfully complain that city does not exist on interface User This only the tip of the TS iceberg Don t take my word for it you can check out the documentation here ConclusionOnce given a proper chance TypeScript has been amazingly helpful at saving me time and hard ache with features Editor Check Auto complete and Interfaces I won t lie sometimes I just slap any on things and call it a day but it eats away at me until I get it typed like the rest of the project So Keep calm and Type Script on Thanks for reading	2021-12-01 19:20:44
海外TECH	DEV Community	Welcome Thread - v153	https://dev.to/thepracticaldev/welcome-thread-v153-36c0	Welcome Thread v Welcome to DEV Leave a comment below to introduce yourself You can talk about what brought you here what you re learning or just a fun fact about yourself Reply to someone s comment either with a question or just a hello Great to have you in the community	2021-12-01 19:15:04
海外TECH	DEV Community	AWS Just Announced that the Future is Serverless at re:Invent — Day 3	https://dev.to/aws-builders/aws-just-announced-that-the-future-is-serverless-at-reinvent-day-3-14pe	AWS Just Announced that the Future is Serverless at re Invent ーDay We just wrapped up Day of Amazon Web Services AWS re Invent and there were a ton of announcements AWS CEO Adam Selipsky delivered the opening keynote and announced a number of exciting updates and announcements including Amazon Private G imagine building your own G network how cool is that Amazon SageMaker Canvas for low code machine learning in the cloud and a great analysis of what a modern data strategy looks like Additionally I share insights from AWS David Yanacek about how he builds operational excellence for his teams in an agile way for developers and DevOps engineers at Amazon Finally I go through a theme park app I built at one of the workshops In this episode of the Tech Stack Playbook I ll be sharing more about AWS re Invent the year s largest global tech conference that brings together leaders builders innovators and creators from cloud DevOps serverless app development and software engineering from all around the world In this blog post and the blog posts to come this week my hope is to vlog through the entire week and share all that I m learning and building with you all so whether you are attending virtually or in person this AWS re Invent series on the Tech Stack Playbook will help you gain skills level up in tech and boost your software engineering abilities from the sessions I m attending throughout the conference Here s a glance at what you ll learn in this blog post What AWS re Invent is all aboutTraveling to Las Vegas for the year s largest tech conferenceWhat I ll be learning about and building this weekIt s going to be a major week for cloud and tech in general so I am so excited to bring you along for the ride What do you want to know about AWS serverless and DevOps Let me know in the comments below ️ Check out the full recording below Let me know if you found this post helpful And if you haven t yet make sure to check out these free resources below Follow my Instagram for more BrianHHoughWatch my latest YouTube video for moreListen to my Podcast on Apple Podcasts and Spotify Join my FREE Tech Stack Playbook Facebook GroupLet s digitize the world together Brian	2021-12-01 19:10:27
海外TECH	DEV Community	Build 30 JavaScript projects in 30 days:Day 0	https://dev.to/prachigarg19/build-30-javascript-projects-in-30-daysday-0-2m8j	Build JavaScript projects in days Day This is building projects in days series I ll be following Wes Bos Javascript course for this challenge the link to which is I ll be sharing my approach in every article which may or may not be same as the one mentioned in his videos Before starting with this series I d recommend you to download the starter files in your device the link to which is hereProcess to download starter files gt Fork the repo by clicking the option in the top right Clone repo a If you have github desktop installed which I recommend as it is easier to work with click on the green colored code option and then click open with gitub desktop Now the application will automatically clone the repo to your device b Using git bash Refer to this article Now go to the folder where repo has been cloned and congrats there is your starter package You can copy and paste all the sounds images etc to your project folder as per your requirements and use them in your projects This is it for today In my next article I will discuss the first project which is a DrumKit If you have any doubts or suggestions please do let me know in the comment section I ll be more than happy to interact with you If you like this series and want to be a part if it do consider following me at prachigargThanks for reading	2021-12-01 19:09:51
海外TECH	DEV Community	🌟 Biggest “Online Tools” Collection: 10015.io 🧰	https://dev.to/fatihtelis/biggest-online-tools-collection-10015io-3gbf	Biggest “Online Tools Collection io Hi I m Fatih a frontend developer based in Istanbul and this is the short story of my project io io is an online toolbox project I ve been working on for a while Since I m a frontend developer I ve been using many online tools while I m working I m also a bookmark type of guy where I organize all my bookmarks under specific folders The longest folder I had on my bookmarks toolbar was Tools folder previously At some point I asked myself What is going on Clearly there is something wrong with that folder This is the point where I decided to create an online toolbox There are some reasons behind this decision Why I decided to create an online toolbox There isn t any nice toolbox which contains every tool I need UI UX of almost all online tools are really bad It is hard the find the tool I need instantly It will be a good challenge for me to discover how this types of tools are developed After I made my decision I started to think about the brand name of the project As many of you experienced before almost all generic domains were already bought by domain brokers So I need to find a short amp catchy as well as smart brand name Then a light bulb went on my head and I found If you look closer to the logo you will see why I chose this brand name If you have a prediction about why I chose as brand name just comment and I will send you a year free membership Ohh sorry io is already free I developed extension for io one is for Chrome and one is for Firefox You can use these extensions for accessing the tools instantly Chrome ExtensionFirefox ExtensionThere are online tools in io right now and I m adding new tools regularly I have a list on Notion for upcoming tools which are added as a result of either my needs or requests I get via contact page I set some goals and milestones for io Here is my roadmap ️Roadmap Tools gt Write some articles and post stuff on social media Tools gt Post on Product Hunt Tools gt Post on Product Hunt as v Tools gt Post on Product Hunt as vMy aim is to make io the biggest free online toolbox in web so I decided to use a pretentious title like this for my article as a motivation Tech StackI m using Next js A React Framework along with styled components for styling amp theming I m not using a UI library for components all are written specifically for this project Almost all tools are working on client side only There are very few of them making API requests to server If you have any comments about io feel free to write on comment section or you can directly contact with me I will post some articles about the tools individually how I made them and what I ve learnt while I was creating the tools If you want to get notified just follow me	2021-12-01 19:09:05
海外TECH	DEV Community	I broke production 3 times in 3 weeks - Part II	https://dev.to/sandordargo/i-broke-production-3-times-in-3-weeks-part-ii-4c4a	I broke production times in weeks Part IILast week I shared with you that despite that I consider myself a careful coder I managed to break production several times in a row It can happen to anyone though one shouldn t forget about his responsibility leading to such events We can complain about how useless the test systems are we can blame the reviewers but at the end of the day the code was written by one person In these cases by me Last week I shared how I slipped and introduced undefined behaviour by not initializing a pointer correctly Now let s continue with two other stories with two other bugs A memory leakAnother issue I introduced was once again about pointers A very knowledgeable C engineer told me recently for a good reason that for dynamic memory management you should always use smart pointers but it s even better if you can avoid using dynamic allocations at all So in one of another monster classes I found a pointer that was initialized to nullptr in the initializer list some objects were assigned to it at many different places and at the end in the destructor it was not deleted and I couldn t find where it was cleaned up A memory leak unless I missed the cleanup The pointer was passed to another object several times it updated the pointed object and then it was taken back Somehow like this auto aStatus STATUS UNDEFINED auto aService MyService aService setAdapter m serviceAdapter try aStatus aService resume catch std exception amp e should now contain the right data m serviceAdapter static cast lt MyServiceAdapter gt aService getAdapter All problems can be avoided by using smart pointers A very easy option could have been using a shared pointer but I didn t want to do it for two reasons MyService lives in another repository and it takes about a day to change review and deliver a new version and this is such a lousy reason in most cases where you use a shared pointer it s not necessary It s simply the easier road to take I didn t want to take the easier road So I went on using a unique pointer m serviceAdapter became a std unique ptr lt MyServiceAdapter gt instead of MyServiceAdapter and I changed the code like this auto aStatus STATUS UNDEFINED auto aService MyService aService setAdapter m serviceAdapter release try aStatus aService resume catch std exception amp e should now contain the right data m serviceAdapter reset static cast lt MyServiceAdapter gt aService getAdapter My reasoning was that at we don t want to own the service adapter anymore we pass the ownership to the service even if we happen to know that it won t delete it it ll just give back the ownership a few lines later At step we just reset the local adapter from the other service s adapter All is fine right At step we released the ownership and at step we got it back What can go wrong What if MyServiceAdapter assigns another pointer without deleting what it got It s a memory leak but it s a problem in MyServiceAdapter not at the call place So we could argue that all is fine There were about functions following the above pattern But there was another one where there was only the release part there was no reset And with this I clearly introduced a memory leak and it required a fallback So how it is possible that from a small memory leak we went to a bigger one That s something I still don t understand I think that with the above change I should have reduced the memory leak because in most cases the pointer got deleted unlike before Yet the stats from production was very clear The takeaways for this second issue When it comes to memory management be extra cautious Don t go with half solutions If you assume you pass ownership go all way through the chain and fix the whole flow Use valgrind more to understand better what happens to your allocated memory Know your toolsThis last one will be shorter probably a bit less technical Do you know this meme I think this is a great one and what you can see on the bottom half is actually a quite frequent situation Why do I say so You have some piece of code that doesn t work and you have no idea why Then you fix it Do you even think it through Do you understand why it works If so great But what if not Do you keep investigating Or do you simply move on to the next issue I m not here to judge Often we don t have time to continue the investigation and we must take things in the pipe But it has a serious implication Often what works will not function because it s the good solution I wouldn t even call it a good enough solution Sometimes it s just a brittle repair that happens to work under the given circumstances but it can break any time C still doesn t have a build management system that would be the de facto standard many companies have their own one just like us Therefore I will not go into deep technical details of what happened but I give you a more high level view Some dependencies were grouped into a package and I made the necessary updates in our descriptors We were already depending on several packages which were pretty much listed in alphabetical order By the way the two most important packages happened to be at the beginning of this sorting So I updated the dependencies put the package in its alphabetical place then I ran launched the build and the tests They were all fine We loaded into test systems Nobody raised a word Later we loaded into production People started to complain We broke our stats collector We understood quickly that it s coming from the load so we did a fallback But what could it be I was preparing for an important and high risk change and I didn t want to mix it with a routine version update so I performed this version update separately Even that broke our software I changed two things I updated the versions of some of our dependencies And I changed from where we take those exact same dependencies I checked the release notes the source code of our dependencies Nothing I checked with the maintainers of the package they had no idea I checked the documentation of the build management system and it had nothing on the order of the dependencies But as we were out of ideas we changed the order and lo and behold that actually worked The order of the included dependencies matters when it comes to resolving some non matching versions Many things made this investigation more painful than it should have been the problem was not noticed before the production load even though it was visible in test already for the usersit s not reproducible in local so there was only a very limited number of chances to try something new each day the documentation is clearly incomplete on how versions are inferredWhat are the lessons learned Things only work by chance more often than you thinkIf you want to grow to the next level take time to understand your toolsIf you are unsure about your changes take baby steps and validate them as soon as possible ConclusionOften things are accidentally working and they can break anytime You can break them with the best intention even when you think you introduce some changes that you consider technical improvements I d go even further those are the moments when it s the easiest to break the system when you are convinced that you are delivering improvements My advice is to take time to understand what exactly are you doing and don t be afraid of taking baby steps The smaller the iterations the easier it will be to understand and debug And if shit happens don t be discouraged Keep improving the system Connect deeperIf you liked this article please hit on the like button subscribe to my newsletter and let s connect on Twitter	2021-12-01 19:02:49
海外TECH	Engadget	Barnes & Noble releases its first Nook GlowLight e-reader in four years	https://www.engadget.com/barnes-and-noble-nook-glowlight-4-e-reader-191526217.html?src=rss	Barnes amp Noble releases its first Nook GlowLight e reader in four yearsBarnes amp Noble hasn t done as much lately to push e readers forward as some of its rivals but it s starting to catch up TechCrunchnotes the bookseller has launched the Nook GlowLight its first entry in the mid tier e reader line since s GlowLight A lot has changed in four years if not as much as you might think ーthis is as much about dragging the device into the modern era as anything The Nook GlowLight is billed as quot sleeker quot than the GlowLight with a smaller overall profile better page turning buttons and a more comfortable hand feel It s the first Nook e reader with USB C and storage has quadrupled to GB However you ll still see a inch DPI e paper display and a claimed one month battery life This is more for Nook newcomers or those upgrading from particularly old devices The GlowLight ships December th for That could make it a tough sell against the inch Kindle Paperwhite the water friendly Kobo Libra and other mid tier e readers It s a viable alternative if you don t want to attach yourself to the Amazon or Kobo ecosystems though Also this is as much a declaration of intent as a new product Barnes amp Noble chief James Daunt said the company planned to quot reinvigorate quot the Nook line in the months and years ahead ーthis may just be one of the opening salvos	2021-12-01 19:15:26
ニュース	BBC News - Home	Covid Omicron: Time to consider mandatory jabs, EU chief says	https://www.bbc.co.uk/news/world-europe-59497462?at_medium=RSS&at_campaign=KARANGA	leyen	2021-12-01 19:18:32
ニュース	BBC News - Home	Women's tennis China tournaments suspended over Peng	https://www.bbc.co.uk/sport/tennis/59498779?at_medium=RSS&at_campaign=KARANGA	shuai	2021-12-01 19:46:33
ニュース	BBC News - Home	Turner Prize 2021: Irish pub installation wins award	https://www.bbc.co.uk/news/entertainment-arts-59490291?at_medium=RSS&at_campaign=KARANGA	ireland	2021-12-01 19:53:36
ニュース	BBC News - Home	Storm Arwen power cuts: 30,000 still waiting to be reconnected	https://www.bbc.co.uk/news/uk-59396135?at_medium=RSS&at_campaign=KARANGA	homes	2021-12-01 19:06:57
ニュース	BBC News - Home	US Supreme Court hears landmark abortion case	https://www.bbc.co.uk/news/world-us-canada-59495210?at_medium=RSS&at_campaign=KARANGA	american	2021-12-01 19:32:21
ニュース	BBC News - Home	'It's fuelling me every day' - Joshua remains 'angry' after Usyk defeat	https://www.bbc.co.uk/sport/boxing/59495503?at_medium=RSS&at_campaign=KARANGA	x It x s fuelling me every day x Joshua remains x angry x after Usyk defeatAnthony Joshua says he remains angry at myself after losing to Oleksandr Usyk but it s fuelling me every day	2021-12-01 19:03:49
ビジネス	ダイヤモンド・オンライン - 新着記事	保険の見直しで絶対してはいけない3つのこと、解約方法・特約の扱いで大損も - 絶対やってはいけないお金の話	https://diamond.jp/articles/-/289217	非常	2021-12-02 04:50:00
ビジネス	ダイヤモンド・オンライン - 新着記事	製薬業界再編の勝ち組に？「第一三共」期待の大型新薬と宿痾の組織風土 - 医薬経済ONLINE	https://diamond.jp/articles/-/287915		2021-12-02 04:45:00
ビジネス	ダイヤモンド・オンライン - 新着記事	伊藤忠商事が「祖業の繊維事業」で飛躍できた理由 - 伊藤忠 財閥系を凌駕した野武士集団	https://diamond.jp/articles/-/283339	伊藤忠商事	2021-12-02 04:35:00
ビジネス	ダイヤモンド・オンライン - 新着記事	ESG実現の“ミッシングピース”、日本が軽視する「役員教育」 - 政策・マーケットラボ	https://diamond.jp/articles/-/289215	代表理事	2021-12-02 04:30:00
ビジネス	ダイヤモンド・オンライン - 新着記事	【クイズ】日本最大の機関投資家は、日本株の1年間のリターンをどうみてる？ - 「お金の達人」養成クイズ	https://diamond.jp/articles/-/288091	機関投資家	2021-12-02 04:25:00
ビジネス	ダイヤモンド・オンライン - 新着記事	「5万円クーポン支給問題」で甦る地域振興券の記憶、日本は23年前より後退 - 情報戦の裏側	https://diamond.jp/articles/-/289214	参議院議員選挙	2021-12-02 04:20:00
ビジネス	ダイヤモンド・オンライン - 新着記事	コンビニ“冬の風物詩”おでんの撤去相次ぐ裏で、台頭する商品とは？ - News&Analysis	https://diamond.jp/articles/-/289084	newsampampanalysis	2021-12-02 04:15:00
ビジネス	ダイヤモンド・オンライン - 新着記事	インフレ高進の欧米とは別世界、日本企業はなぜ価格転嫁できなくなったのか - 政策・マーケットラボ	https://diamond.jp/articles/-/289172	日本企業	2021-12-02 04:10:00
ビジネス	ダイヤモンド・オンライン - 新着記事	ベゾス氏の不倫・ハッキング騒動、静かに幕切れ - WSJ発	https://diamond.jp/articles/-/289329	騒動	2021-12-02 04:07:00
ビジネス	ダイヤモンド・オンライン - 新着記事	銀行や電力「低成長・独占」産業ばかり高賃金な日本の“貧困” - 野口悠紀雄　新しい経済成長の経路を探る	https://diamond.jp/articles/-/289167	人材育成	2021-12-02 04:05:00
ビジネス	東洋経済オンライン	箱根登山ケーブルカー､どこで運転しているのか 開業100年の歴史､一般的な鉄道と何が違う？ | ローカル線･公共交通 | 東洋経済オンライン	https://toyokeizai.net/articles/-/472955?utm_source=rss&utm_medium=http&utm_campaign=link_back	東洋経済オンライン	2021-12-02 04:30:00
IT	IT号外	できるだけ多くのコンテンツを迅速に楽しむための、自分が考え出した50%ルールについて	https://figreen.org/it/%e3%81%a7%e3%81%8d%e3%82%8b%e3%81%a0%e3%81%91%e5%a4%9a%e3%81%8f%e3%81%ae%e3%82%b3%e3%83%b3%e3%83%86%e3%83%b3%e3%83%84%e3%82%92%e8%bf%85%e9%80%9f%e3%81%ab%e6%a5%bd%e3%81%97%e3%82%80%e3%81%9f%e3%82%81/	できるだけ多くのコンテンツを迅速に楽しむための、自分が考え出したルールについて多くの人が既に感じている通り、現代はコンテンツ過多の時代になってしまっていますよね。	2021-12-01 19:05:21
GCP	Cloud Blog	Google showcases Cloud TPU v4 Pods for large model training	https://cloud.google.com/blog/topics/tpus/google-showcases-cloud-tpu-v4-pods-for-large-model-training/	Google showcases Cloud TPU v Pods for large model trainingRecently models with billions or trillions of parameters have shown significant advances in machine learning capabilities and accuracy For example Google s LaMDA model is able to engage in a free flowing conversation with users about a large variety of topics There is enormous interest within the machine learning research and product communities in leveraging large models to deliver breakthrough capabilities The high computational demand of these large models requires an increased focus on improving the efficiency of the model training process and benchmarking is an important means to coalesce the ML systems community towards realizing higher efficiencies In the recently concluded MLPerf v Training round Google submitted two large language model benchmarks into the Open division one with billion parameters and a second with billion parameters These submissions make use of publicly available infrastructure including Cloud TPU v Pod slices and the Lingvo open source modeling framework  Traditionally training models at these scales would require building a supercomputer at a cost of tens or even hundreds of millions of dollars something only a few companies can afford to do Customers can achieve the same results using exaflop scale Cloud TPU v Pods without incurring the costs of installing and maintaining an on premise system  Large model benchmarksGoogle s Open division submissions consist of a billion parameter dense Transformer based encoder only benchmark using TensorFlow and a billion parameter JAX benchmark These models are architecturally similar to MLPerf s BERT model but with larger dimensions and number of layers These submissions demonstrate large model scalability and high performance on TPUs across two distinct frameworks Notably these benchmarks with their stacked transformer architecture are fairly comparable in terms of their compute characteristics with other large language models Figure Architecture of the Encoder only model used in Google s MLPerf submissions Our two submissions were benchmarked on chip and chip TPU v Pod slices respectively We were able to achieve an end to end training time of hours for the B parameter model and hours for the B parameter model Each of these runs achieved a computational efficiency of calculated as a fraction of floating point operations of the model together with compiler rematerialization over the peak FLOPs of the system used  Next generation ML infrastructure for large Model training Achieving these impressive results required a combination of several cutting edge technologies First each TPU v chip provides more than X the compute power of a TPU v chip up to peak TFLOPS Second TPU v chips are networked together into a Cloud TPU v Pod by an ultra fast interconnect that provides x the bandwidth per chip at scale compared to typical GPU based large scale training systems Large models are very communication intensive local computation often depends on results from remote computation that are communicated across the network TPU v s ultra fast interconnect has an outsized impact on computational efficiency of large models by eliminating latency and congestion in the network Figure A portion of one of Google s Cloud TPU v Pods each of which is capable of delivering in excess of exaflop s of computing power The performance numbers demonstrated by our submission also rely on our XLA linear algebra compiler and leverage the Lingvo framework XLA transparently performs a number of optimizations including GSPMD based automatic parallelization of many of the computation graphs that form the building blocks of the ML model XLA also allows for reduction in latency by overlapping communication with the computations Our two submissions demonstrate the versatility and performance of our software stack across two frameworks TensorFlow and JAX Large models in MLPerfGoogle s submissions represent an important class of models that have become increasingly important in ML research and production but are currently not represented in MLPerf s Closed division benchmark suite  We believe that adding these models to the benchmark suite is an important next step and can inspire the ML systems community to focus on addressing the scalability challenges that large models present Our submissions demonstrate computational efficiency cutting edge in the industry This high computational efficiency enables higher experimentation velocity through faster training This directly translates into cost savings for Google s Cloud TPU customers   Please visit the Cloud TPU homepage and documentation to learn more about leveraging Cloud TPUs using TensorFlow PyTorch and JAX The MLPerf name and logo are trademarks of MLCommons Association in the United States and other countries All rights reserved Unauthorized use is strictly prohibited See www mlcommons org for more information	2021-12-01 20:00:00