投稿時間:2020-12-18 05:37:49 RSSフィード2020-12-18 05:00 分まとめ(42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 Big Data Blog	Accelerating Amazon Redshift federated query to Amazon Aurora MySQL with AWS CloudFormation	https://aws.amazon.com/blogs/big-data/accelerating-amazon-redshift-federated-query-to-amazon-aurora-mysql-with-aws-cloudformation/	Accelerating Amazon Redshift federated query to Amazon Aurora MySQL with AWS CloudFormationAmazon Redshift federated query allows you to combine data from one or more Amazon Relational Database Service Amazon RDS for MySQL and Amazon Aurora MySQL databases with data already in Amazon Redshift You can also combine such data with data in an Amazon Simple Storage Service Amazon S data lake This post shows you how …	2020-12-17 19:56:36
AWS	AWS Machine Learning Blog	Monitoring in-production ML models at large scale using Amazon SageMaker Model Monitor	https://aws.amazon.com/blogs/machine-learning/monitoring-in-production-ml-models-at-large-scale-using-amazon-sagemaker-model-monitor/	Monitoring in production ML models at large scale using Amazon SageMaker Model MonitorMachine learning ML models are impacting business decisions of organizations around the globe from retail and financial services to autonomous vehicles and space exploration For these organizations training and deploying ML models into production is only one step towards achieving business goals Model performance may degrade over time for several reasons such as changing consumer …	2020-12-17 19:56:38
Google	Official Google Blog	Redesigning Search would harm American consumers and businesses	http://feedproxy.google.com/~r/blogspot/MKuf/~3/nmTaMtvtEJ0/	Redesigning Search would harm American consumers and businessesOur response to today s lawsuit about the design of Google Search by state attorneys general Google Search is designed to provide you with the most relevant results We know that if you don t like the results we re giving you you have numerous alternativesーincluding Amazon Expedia Tripadvisor and many others just a click away So we keep working to improve our results designing and rolling out helpful features in Searchーincluding maps links to products and services you can buy directly flight and hotel options and local business information like hours of operation and delivery services Look at how our search results have evolved and improved over the years This is what our search results looked like in ー blue links but no other useful features And this is what they look like todayーmore useful information more direct connections to businesses more links to websites Our rigorous testing tells us that you far prefer these types of rich results Other search engines like Microsoft s Bing seem to have heard the same feedback because they have also evolved to provide these kinds of direct results To get more specifically to the issues raised in today s lawsuit it suggests we shouldn t have worked to make Search better and that we should in fact be less useful to you When you search for local products and services we show information that helps you connect with businesses directly and helps them reach more customers This lawsuit demands changes to the design of Google Search requiring us to prominently feature online middlemen in place of direct connections to businesses   Redesigning Google Search this way would harm the quality of your search results And it would come at the expense of businesses like retailers restaurants repair shops airlines and hotels whose listings in Google help them get discovered and connect directly with customers They would have a harder time reaching new customers and competing against big commerce and travel platforms and other aggregators and middlemen  The data shows that our local results in Search drive more than billion direct connections for businesses every month such as visits to businesses websites people calling merchants getting directions to stores ordering food from restaurants  Even as we have added content and features to our search results the volume of traffic we send to non Google sites has increased every year since Search was created Our search results page which used to show links now shows an average of outgoing links on mobile devices  The claims being made have been closely examined and rejected by regulators and courts around the world including the U S Federal Trade Commission competition authorities in Brazil Canada and Taiwan and courts in the United Kingdom and Germany who all agreed that our changes are designed to improve your search results It s also well established that the most important driver for our search results is the specific queryーnot your personal data We know that scrutiny of big companies is important and we re prepared to answer questions and work through the issues But this lawsuit seeks to redesign Search in ways that would deprive Americans of helpful information and hurt businesses ability to connect directly with customers We look forward to making that case in court while remaining focused on delivering a high quality search experience for our users  The lawsuit also contains allegations that have previously been made about how we distribute Search and about our advertising technologies On those topics you can read our blog post and see more specifics on our competition site	2020-12-17 19:17:00
python	Pythonタグが付けられた新着投稿 - Qiita	既存ボーンにスクリプトで補助ボーンを生やしてみる	https://qiita.com/yukimituki11/items/728e803a9a01dccae541	既存ボーンにスクリプトで補助ボーンを生やしてみるBlenderAdventCalendarに空きが目立ったのもあってTwitterのタイムラインで話題を見かけた既存のボーンに補助ボーンを作成するスクリプトを即興で書いてみた。	2020-12-18 04:12:49
js	JavaScriptタグが付けられた新着投稿 - Qiita	ExpoとReact Native Webを使ってコードを書かずに作ってみるPWA入門	https://qiita.com/takagimeow/items/11b9e7f7c1052ada6995	上記の画面が表示されれば成功です過去にlocalhostで何かを実行していてServiceWorker等が動いていた場合はその時のアプリが表示されている可能性があるので、devtoolsを開いて、ApplicationタブからClearSiteStorageを選んでClearSiteDataをクリックする。	2020-12-18 04:18:12
js	JavaScriptタグが付けられた新着投稿 - Qiita	Reactの状態管理ライブラリまとめ	https://qiita.com/ape/items/7873a3bfa1a64b36f462	Reactの状態管理ライブラリまとめはじめに今回はReactのReactの状態管理ライブラリについてまとめていきます。	2020-12-18 04:09:46
Program	[全てのタグ]の新着質問一覧｜teratail（テラテイル）	bmpファイルのrbg値を書き換えるプログラム	https://teratail.com/questions/310938?rss=all	bmpファイルのrbg値を書き換えるプログラムinbmpを読み込んでrbg値を書き換え、outbmpで出力するプログラムです。	2020-12-18 04:20:36
海外TECH	Ars Technica	Looking into the genetics of severe COVID-19	https://arstechnica.com/?p=1730434	people	2020-12-17 19:55:20
Apple	AppleInsider - Frontpage News	Parallels 16 for Apple Silicon M1 Mac launches in beta - minus Intel OS support	https://appleinsider.com/articles/20/12/17/parallels-16-for-apple-silicon-m1-mac-launches-in-beta---minus-intel-os-support	Parallels for Apple Silicon M Mac launches in beta minus Intel OS supportPopular virtualization software Parallels now supports Apple Silicon in an invite only Technical Preview ーwith notable caveats Parallels is inviting beta testers to try out its Technical Preview for M MacsParallels showcased an Apple Silicon version of its software at WWDC but this Technical Preview is the first time the company has made its virtualization app available to M Mac users Parallels is sending invite emails viewed by AppleInsider with installation instructions Read more	2020-12-17 19:25:49
Apple	AppleInsider - Frontpage News	Apple has stopped providing standalone updaters in macOS Big Sur	https://appleinsider.com/articles/20/12/17/apple-stops-providing-macos-standalone-updaters-in-big-sur	Apple has stopped providing standalone updaters in macOS Big SurApple has stopped providing standalone updater versions of its macOS system software starting with macOS Big Sur ーbut is still providing them for Catalina and Mojave as recently as December Credit Andrew O Hara AppleInsiderFor decades Apple has offered a users the ability to download standalone smaller update files without needing to download a full OS install image or the same update in the macOS Software Update feature multiple times across a rollout of Macs In macOS Big Sur that download has not been made available for any of the updates Read more	2020-12-17 19:23:07
Apple	AppleInsider - Frontpage News	Testing the new HomePod mini music Handoff feature in iOS 14.4	https://appleinsider.com/articles/20/12/17/testing-the-new-homepod-mini-music-handoff-feature-in-ios-144	Testing the new HomePod mini music Handoff feature in iOS Apple has released the first beta of iOS to developers and in it is the long awaited U based music Handoff experience for the new HomePod mini iOS for iPhone improves Handoff with HomePod miniWhat s new in iOS beta Read more	2020-12-17 19:30:07
Apple	AppleInsider - Frontpage News	Lowest price: save $400 on Apple's 32" Pro Display XDR, free expedited shipping	https://appleinsider.com/articles/20/12/17/lowest-price-save-400-on-apples-32-pro-display-xdr-free-expedited-shipping	Lowest price save on Apple x s quot Pro Display XDR free expedited shippingIf you re shopping for the ultimate holiday gift for the content creator in your life ーor if you re a business owner looking to invest in premium equipment for tax year ーB amp H s latest deal on the inch Apple Pro Display XDR should be on your radar Apple Pro Display XDR discountFor a limited time only AppleInsider readers can save on Apple s K display in your choice of Standard Glass or Nano Texture Glass These deals deliver the lowest prices available on the high end monitors according to our Display Price Guide when used with the special activation instructions below Read more	2020-12-17 19:17:50
海外TECH	Engadget	Twitter will reopen verification requests in 2021	https://www.engadget.com/twitter-will-open-verification-requests-in-2021-195624021.html	Twitter will reopen verification requests in Twitter will soon let users request verification for the first time in more than three years The company unveiled its revamped verification policy which will go into effect in January And with the new policy Twitter will open back up verific	2020-12-17 19:56:24
海外TECH	Engadget	Judge orders Apple’s iOS and macOS chief to testify in Epic case	https://www.engadget.com/tim-cook-craig-federighi-testify-apple-epic-games-case-195128510.html	Judge orders Apple s iOS and macOS chief to testify in Epic caseApple s iOS and macOS chief Craig Federighi will have to join CEO Tim Cook in testifying in the company s case against Epic Games Apple wanted Cook and App Store head Eric Neuenshwander who reports to Federighi to represent the company in the case	2020-12-17 19:51:28
海外科学	NYT > Science	Biden Will Pick Deb Haaland to Lead Interior Department	https://www.nytimes.com/2020/12/17/climate/deb-haaland-interior-department-native-american.html	cabinet	2020-12-17 19:50:56
海外TECH	WIRED	Russia's Hack Wasn't Cyberwar. That Complicates US Strategy	https://www.wired.com/story/russia-solarwinds-hack-wasnt-cyberwar-us-strategy	solarwinds	2020-12-17 19:44:01
海外ニュース	Japan Times latest articles	France’s Macron tests positive for COVID-19	https://www.japantimes.co.jp/news/2020/12/17/world/macron-positive-covid/	isolate	2020-12-18 04:35:43
海外ニュース	Japan Times latest articles	Nissan decides against making EV in U.K. over Brexit worries	https://www.japantimes.co.jp/news/2020/12/17/business/nissan-brexit-ev/	europe	2020-12-18 04:16:02
ニュース	BBC News - Home	Brexit trade talks: 'Big differences remain' - Ursula von der Leyen	https://www.bbc.co.uk/news/uk-politics-55347723	leyen	2020-12-17 19:51:32
ニュース	BBC News - Home	Coronavirus: NI facing six-week lockdown from 26 December	https://www.bbc.co.uk/news/uk-northern-ireland-55349545	neill	2020-12-17 19:53:36
ニュース	BBC News - Home	Rishi Sunak extends furlough scheme for another month	https://www.bbc.co.uk/news/business-55345392	april	2020-12-17 19:33:01
ニュース	BBC News - Home	Best Fifa Football Awards 2020: England's Lucy Bronze wins top prize	https://www.bbc.co.uk/sport/football/55352832	Best Fifa Football Awards England x s Lucy Bronze wins top prizeManchester City defender Lucy Bronze becomes the first English player to win women s player of the year at the Best Fifa Football Awards	2020-12-17 19:50:37
ニュース	BBC News - Home	Lewandowski wins Fifa best men's player award	https://www.bbc.co.uk/sport/football/55354762	Lewandowski wins Fifa best men x s player awardBayern Munich s Robert Lewandowski wins the best men s player at the Best Fifa Football Awards beating Lionel Messi and Cristiano Ronaldo to the top prize	2020-12-17 19:54:14
ビジネス	ダイヤモンド・オンライン - 新着記事	ソニーがアニメ『鬼滅の刃』で見せた、ビジネス・エコシステムの「柱」 - 長内 厚のエレキの深層	https://diamond.jp/articles/-/257607	産みの親	2020-12-18 04:55:00
ビジネス	ダイヤモンド・オンライン - 新着記事	円高進展でドル円100円割れは起きるか？2021年の主要通貨を大胆予測 - 政策・マーケットラボ	https://diamond.jp/articles/-/257431	円高進展でドル円円割れは起きるか年の主要通貨を大胆予測政策・マーケットラボ年の為替市場では、スウェーデンクローナやユーロなどの欧州通貨が上昇する一方で、ノルウェークローナやカナダドルといった産油国通貨が下落した。	2020-12-18 04:50:00
ビジネス	ダイヤモンド・オンライン - 新着記事	中国都市「GDPベスト10」に異変、コロナ禍で注目の武漢市は失地挽回！ - 莫邦富の中国ビジネスおどろき新発見	https://diamond.jp/articles/-/257472		2020-12-18 04:45:00
ビジネス	ダイヤモンド・オンライン - 新着記事	半年後は「アフターコロナ」で日本経済が回復するといえる根拠 - 岸博幸の政策ウォッチ	https://diamond.jp/articles/-/257606	可処分所得	2020-12-18 04:40:00
ビジネス	ダイヤモンド・オンライン - 新着記事	テレワークのコミュニケーションで重要な3つの「きく」とは - News&Analysis	https://diamond.jp/articles/-/257425	newsampampanalysis	2020-12-18 04:35:00
ビジネス	ダイヤモンド・オンライン - 新着記事	ユーザー爆増の「note」「cakes」炎上、他人事ではないメディアへの教訓 - News&Analysis	https://diamond.jp/articles/-/257605	cakes	2020-12-18 04:30:00
ビジネス	ダイヤモンド・オンライン - 新着記事	ブロンコス全試合無料化を断念！それでも「お金はもらいたくない」私の言い分 - 池田純のプロスポーツチーム変革日記	https://diamond.jp/articles/-/257604		2020-12-18 04:25:00
ビジネス	ダイヤモンド・オンライン - 新着記事	米38州がグーグル提訴　反トラスト法違反で - WSJ発	https://diamond.jp/articles/-/257707	反トラスト法	2020-12-18 04:20:00
ビジネス	ダイヤモンド・オンライン - 新着記事	漢方で防ぐ「体内の乾燥」、多くの人が気づきにくい不調のサインとは - ニュース3面鏡	https://diamond.jp/articles/-/257603	そこで今回は、薬剤師でうるおいコンシェルジュである大塚まひささんの著書『うるおい漢方』青春出版社から、体内のうるおい成分が不足したときのサインと原因について、漢方の考え方からひも解いていきます。	2020-12-18 04:20:00
ビジネス	ダイヤモンド・オンライン - 新着記事	生活保護の申請をよしとしない役所の「水際作戦」に、立ち向かう手立て - 生活保護のリアル～私たちの明日は？ みわよしこ	https://diamond.jp/articles/-/257602	水際作戦	2020-12-18 04:15:00
ビジネス	ダイヤモンド・オンライン - 新着記事	ロシア出場禁止を2年に短縮、ドーピング問題＝スポーツ裁判所 - WSJ発	https://diamond.jp/articles/-/257708	裁判所	2020-12-18 04:14:00
ビジネス	ダイヤモンド・オンライン - 新着記事	重病なのに痛みのない「睾丸の腫れ」を放置した、43歳男性の後悔 - 40代以上の男のカラダとココロの悩み	https://diamond.jp/articles/-/257601	風呂	2020-12-18 04:10:00
ビジネス	ダイヤモンド・オンライン - 新着記事	「コロナ禍におけるリーダーとは？」――ユーグレナ社長・出雲充が『嫌われる勇気』著者・岸見一郎へ問う - 嫌われる勇気──自己啓発の源流「アドラー」の教え	https://diamond.jp/articles/-/257399	「コロナ禍におけるリーダーとは」ーユーグレナ社長・出雲充が『嫌われる勇気』著者・岸見一郎へ問う嫌われる勇気ー自己啓発の源流「アドラー」の教え配信イベント『嫌われる勇気オンラインフェス』で実施した『嫌われる勇気』著者・岸見一郎氏とユーグレナ社長・出雲充氏による対談。	2020-12-18 04:05:00
ビジネス	ダイヤモンド・オンライン - 新着記事	バイデン氏、次期環境長官にNC州環境省トップを起用へ - WSJ発	https://diamond.jp/articles/-/257709	長官	2020-12-18 04:02:00
ビジネス	東洋経済オンライン	ワークマンとカインズが実践する｢真逆｣の経営 ベイシアグループの有力2社の性格は大違い | 百貨店・量販店・総合スーパー | 東洋経済オンライン	https://toyokeizai.net/articles/-/397241?utm_source=rss&utm_medium=http&utm_campaign=link_back	東洋経済オンライン	2020-12-18 05:00:00
ビジネス	東洋経済オンライン	オリンパス､脱･カメラで狙う｢医療の列強｣の座 医療機器の世界大手を目指し､M&Aでも攻勢 | IT･電機･半導体･部品 | 東洋経済オンライン	https://toyokeizai.net/articles/-/396870?utm_source=rss&utm_medium=http&utm_campaign=link_back	mampa	2020-12-18 05:00:00
ビジネス	東洋経済オンライン	中国の半導体大手｢紫光集団｣が破産の崖っ縁 相次ぐ社債デフォルト､過大な合併買収が仇に | 「財新」中国Biz＆Tech | 東洋経済オンライン	https://toyokeizai.net/articles/-/396753?utm_source=rss&utm_medium=http&utm_campaign=link_back	biztech	2020-12-18 04:50:00
ビジネス	東洋経済オンライン	2021年の｢丑年相場｣は相場格言通りつまずく？ ｢辛丑（かのとうし）｣の｢辛｣のほうにも注目 | 市場観測 | 東洋経済オンライン	https://toyokeizai.net/articles/-/396929?utm_source=rss&utm_medium=http&utm_campaign=link_back	topix	2020-12-18 04:30:00
GCP	Cloud Blog	How to automatically scale your machine learning predictions	https://cloud.google.com/blog/products/ai-machine-learning/scaling-machine-learning-predictions/	How to automatically scale your machine learning predictionsHistorically one of the biggest challenges in the data science field is that many models don t make it past the experimental stage As the field has matured we ve seen MLOps processes and tooling emerge that have increased project velocity and reproducibility While we ve got a ways to go more models than ever before are crossing the finish line into production That leads to the next question for data scientists how will my model scale in production In this blog post we will discuss how to use a managed prediction service Google Cloud s AI Platform Prediction to address the challenges of scaling inference workloads Inference WorkloadsIn a machine learning project there are two primary workloads training and inference Training is the process of building a model by learning from data samples and inference is the process of using that model to make a prediction with new data Typically training workloads are not only long running but also sporadic If you re using a feed forward neural network a training workload will include multiple forward and backward passes through the data updating weights and biases to minimize errors In some cases the model created from this process will be used in production for quite some time and in others new training workloads might be triggered frequently to retrain the model with new data On the other hand an inference workload consists of a high volume of smaller transactions An inference operation essentially is a forward pass through a neural network starting with the inputs perform matrix multiplication through each layer and produce an output The workload characteristics will be highly correlated with how the inference is used in a production application For example in an e commerce site each request to the product catalog could trigger an inference operation to provide product recommendations and the traffic served will peak and lull with the e commerce traffic Balancing Cost and LatencyThe primary challenge for inference workloads is balancing cost with latency It s a common requirement for production workloads to have latency lt milliseconds for a smooth user experience On top of that application usage can be spiky and unpredictable but the latency requirements don t go away during times of intense use To ensure that latency requirements are always met it might be tempting to provision an abundance of nodes The downside of overprovisioning is that many nodes will not be fully utilized leading to unnecessarily high costs On the other hand underprovisioning will reduce cost but lead to missing latency targets due to servers being overloaded Even worse users may experience errors if timeouts or dropped packets occur It gets even trickier when we consider that many organizations are using machine learning in multiple applications Each application has a different usage profile and each application might be using a different model with unique performance characteristics For example in this paper Facebook describes the diverse resource requirements of models they are serving for natural language recommendation and computer vision AI Platform Prediction ServiceThe AI Platform Prediction service allows you to easily host your trained machine learning models in the cloud and automatically scale them Your users can make predictions using the hosted models with input data The service supports both online prediction when timely inference is required and batch prediction for processing large jobs in bulk To deploy your trained model you start by creating a model which is essentially a package for related model artifacts Within that model you then create a version which consists of the model file and configuration options such as the machine type framework region scaling and more You can even use a custom container with the service for more control over the framework data processing and dependencies To make predictions with the service you can use the REST API command line or a client library For online prediction you specify the project model and version and then pass in a formatted set of instances as described in the documentation Introduction to scaling optionsWhen defining a version you can specify the number of prediction nodes to use with the manualScaling nodes option By manually setting the number of nodes the nodes will always be running whether or not they are serving predictions You can adjust this number by creating a new model version with a different configuration You can also configure the service to automatically scale The service will increase nodes as traffic increases and remove them as it decreases Auto scaling can be turned on with the autoScaling minNodes option You can also set a maximum number of nodes with autoScaling maxNodes  These settings are key to improving utilization and reducing costs enabling the number of nodes to adjust within the constraints that you specify Continuous availability across zones can be achieved with multi zone scaling to address potential outages in one of the zones Nodes will be distributed across zones in the specified region automatically when using auto scaling with at least node or manual scaling with at least nodes GPU SupportWhen defining a model version you need to specify a machine type and a GPU accelerator which is optional Each virtual machine instance can offload operations to the attached GPU which can significantly improve performance For more information on supported GPUs in Google Cloud see this blog post Reduce costs and increase throughput with NVIDIA Ts Ps Vs The AI Platform Prediction service has recently introduced GPU support for the auto scaling feature The service will look at both CPU and GPU utilization to determine if scaling up or down is required How does auto scaling work The online prediction service scales the number of nodes it uses to maximize the number of requests it can handle without introducing too much latency  To do that the service Allocates some nodes the number can be configured by setting the minNodes option on your model version the first time you request predictions  Automatically scales up the model version s deployment as soon as you need it traffic goes up Automatically scales it back down to save cost when you don t traffic goes down Keeps at least a minimum number of nodes by setting the minNodes option on your model version ready to handle requests even when there are none to handle Today the prediction service supports auto scaling based on two metrics CPU utilization and GPU duty cycle Both metrics are measured by taking the average utilization of each model The user can specify the target value of these two metrics in the CreateVersion API see examples below  the target fields specify the target value for the given metric once the real metric deviates from the target by a certain amount of time the node count adjusts up or down to match How to enable CPU auto scaling in a new modelBelow is an example of creating a version with auto scaling based on a CPU metric In this example the CPU usage target is set to with the minimum nodes set to and maximum nodes set to Once the real CPU usage exceeds the node count will increase to a maximum of  Once the real CPU usage goes below for a certain amount of time the node count will decrease to a minimum of  If no target value is set for a metric it will be set to the default value of REGION us centralusing gcloud  gcloud beta ai platform versions create v model MODEL   region REGION   accelerator count type nvidia tesla t    metric targets cpu usage   min nodes max nodes   runtime version origin gs lt your model path gt machine type n standard framework tensorflowcurl example curl k H Content Type application json H Authorization Bearer gcloud auth print access token https REGION ml googleapis com v projects PROJECT models MODEL versions d version jsonversion jsonUsing GPUs Today the online prediction service supports GPU based prediction which can significantly accelerate the speed of prediction Previously the user needed to manually specify the number of GPUs for each model This configuration had several limitations To give an accurate estimate of the GPU number users would need to know the maximum throughput one GPU could process for certain machine types  The traffic pattern for models may change over time so the original GPU number may not be optimal For example high traffic volume may cause resources to be exhausted leading to timeouts and dropped requests while low traffic volume may lead to idle resources and increased costs To address these limitations the AI Platform Prediction Service has introduced GPU based auto scaling Below is an example of creating a version with auto scaling based on both GPU and CPU metrics In this example the CPU usage target is set to GPU duty cycle is minimum nodes are and maximum nodes are When the real CPU usage exceeds or the GPU duty cycle exceeds for a certain amount of time the node count will increase to a maximum of When the real CPU usage stays below or GPU duty cycle stays below for a certain amount of time the node count will decrease to a minimum of  If no target value is set for a metric it will be set to the default value of acceleratorConfig count is the number of GPUs per node  REGION us centralgcloud Example gcloud beta ai platform versions create v model MODEL   region REGION   accelerator count type nvidia tesla t    metric targets cpu usage   metric targets gpu duty cycle   min nodes max nodes   runtime version origin gs lt your model path gt machine type n standard framework tensorflowCurl Example  curl k H Content Type application json H Authorization Bearer gcloud auth print access token https REGION ml googleapis com v projects PROJECT models MODEL versions d version jsonversion jsonConsiderations when using automatic scalingAutomatic scaling for online prediction can help you serve varying rates of prediction requests while minimizing costs However it is not ideal for all situations The service may not be able to bring nodes online fast enough to keep up with large spikes of request traffic If you ve configured the service to use GPUs also keep in mind that provisioning new GPU nodes takes much longer than CPU nodes If your traffic regularly has steep spikes and if reliably low latency is important to your application you may want to consider setting a low threshold to spin up new machines early setting minNodes to a sufficiently high value or using manual scaling It is recommended to load test your model before putting it in production Using the load test can help tune the minimum number of nodes and threshold values to ensure your model can scale to your load The minimum number of nodes must be at least for the model version to be covered by the AI Platform Training and Prediction SLA The AI Platform Prediction Service has default quotas enabled for service requests such as the number of predictions within a given period as well as CPU and GPU resource utilization You can find more details on the specific limits in the documentation If you need to update these limits you can apply for a quota increase online or through your support channel Wrapping upIn this blog post we ve shown how the AI Platform Prediction service can simply and cost effectively scale to match your workloads You can now configure auto scaling for GPUs to accelerate inference without overprovisioning If you d like to try out the service yourself we have a sample notebook which demonstrates how to deploy a model and configure auto scaling settings The AI Platform Prediction documentation also provides a thorough walkthrough of how to use the service and its configuration options Related ArticleAI Platform Prediction goes GA with improved reliability amp ML workflow integrationAI Platform Prediction goes GA with enhanced reliability amp ML workflow integration Read Article	2020-12-17 20:00:00