<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Goutam Biswas on Medium]]></title>
        <description><![CDATA[Stories by Goutam Biswas on Medium]]></description>
        <link>https://medium.com/@gkbiswas?source=rss-77b905811572------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*4CRtI8mEsNQPLxVS.jpg</url>
            <title>Stories by Goutam Biswas on Medium</title>
            <link>https://medium.com/@gkbiswas?source=rss-77b905811572------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 01:30:25 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@gkbiswas/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Business Metrics for Data Driven Companies that can lead you to profitability whether you are a…]]></title>
            <link>https://medium.com/@gkbiswas/business-metrics-for-data-driven-companies-that-can-lead-you-to-profitability-whether-you-are-a-411d4592cc0?source=rss-77b905811572------2</link>
            <guid isPermaLink="false">https://medium.com/p/411d4592cc0</guid>
            <category><![CDATA[marketing]]></category>
            <category><![CDATA[checklist]]></category>
            <category><![CDATA[ecommerce]]></category>
            <category><![CDATA[business-metrics]]></category>
            <category><![CDATA[profitable-company]]></category>
            <dc:creator><![CDATA[Goutam Biswas]]></dc:creator>
            <pubDate>Fri, 21 Oct 2016 03:50:08 GMT</pubDate>
            <atom:updated>2016-10-21T03:50:08.241Z</atom:updated>
            <content:encoded><![CDATA[<h3>Business Metrics for Data Driven Companies that can lead you to profitability whether you are a product or service company</h3><ol><li>Mobile-optimized version for web site (responsive design). Worldwide mobile loading times under 2 seconds. If High traffic from remote locations, uses Content Delivery Networks for Edge-Cashing for rapid page-load times. Direct purchase available from Android and iOS applications. <strong>~For products and services</strong></li><li>Web site visitors’ full click-stream is recorded for later analysis. <strong>~For products and services</strong></li><li>Visitor “Conversion” on web site is tracked at two levels — voluntary registration and first sale. <strong>~For products and services</strong></li><li>Ongoing program of A/B testing of web site features and text to optimize conversion rates and revenues. <strong>~For products and services</strong></li><li>Achieve and maintain high Google Ad Rank (ad content and web site landing page text factual and relevant to AdWords). <strong>~For products and services</strong></li><li>Realistic Estimates for FLV (Future Lifetime Value of a new customer), click-through rates, and conversion-to- revenue rates permit setting an upper bound on profitable Max CPC (Cost per clickthrough) to bid for AdWords. <strong>~For products and services</strong></li><li>Meaningful incentives for customers to interact with company online between store visits. An example would be targeted promotional offers specifically for those who register online. <strong>~For products and services</strong></li><li>Web site shows accurate local store inventory with images and prices. <strong>~For products only</strong></li><li>Customers can order-ahead online and pick up at the most convenient physical store. <strong>~For products only</strong></li><li>Collaboration with third-party shopping and delivery services to allow same-day (or faster) door-to-door delivery of online purchases. <strong>~For services only</strong></li><li>Service (such as evaluating credit and making a loan) is completed as fast or faster than online competitors. <strong>~For services only</strong></li><li>All interactions with a single customer — web visits, purchases at the SKU level online or in store, support calls and complaints, returns, etc — are stored in a common database and available for just-in-time viewing by Company representatives when talking to a customer. <strong>~For products and services</strong></li><li>Company has developed a paid recurring-revenue “membership” program that is attractive to a significant portion or to all customers. Examples: mandatory Costco Membership, Amazon Prime. <strong>~For products and services</strong></li><li>Point-of-sale customizations. When completing a purchase, customers are offered coupons or other incentives to purchase again in future — and offer is customized based on their unique purchase history. <strong>~For products and services</strong></li><li>Special programs to distinguish, reward, and retain the highest-level recurring revenue customers — “Whales” as they are called in the gambling industry. <strong>~For products and services</strong></li><li>Track churn rates and have a program to contact former customers who have “gone quiet” and are potentially lost — and provide incentives for them to return. <strong>~For products and services</strong></li><li>Track all SKU deliveries and sales patterns at the individual store level to prevent excessive days inventory. <strong>~For products only</strong></li><li>Track all SKU deliveries and sales patterns at the individual store level to identify “zeroinventory” items and avoid the opportunity cost of lost sales. <strong>~For products only</strong></li><li>Use a model that modifies inventory levels at the store level by region, season, and day of the week to optimize days inventory against opportunity cost. <strong>~For products only</strong></li><li>Offer effective price reduction and last-minute promotions to established customers to clear any inventory — such as hotel rooms and airline seats — that would otherwise expire worthless. <strong>~For products and services</strong></li></ol><p>Credits: Daniel Egger, Duke University</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=411d4592cc0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Design of an Scalable BI and Data Visualisation Framework]]></title>
            <link>https://medium.com/@gkbiswas/design-of-an-scalable-bi-and-data-visualisation-framework-5c1bf7a28c5b?source=rss-77b905811572------2</link>
            <guid isPermaLink="false">https://medium.com/p/5c1bf7a28c5b</guid>
            <category><![CDATA[business-inte]]></category>
            <category><![CDATA[scalability]]></category>
            <category><![CDATA[aws]]></category>
            <category><![CDATA[data-visualization]]></category>
            <category><![CDATA[docker]]></category>
            <dc:creator><![CDATA[Goutam Biswas]]></dc:creator>
            <pubDate>Tue, 18 Oct 2016 16:09:51 GMT</pubDate>
            <atom:updated>2016-10-18T16:09:51.106Z</atom:updated>
            <content:encoded><![CDATA[<p>Problem — To design a realtime, highly available and scalable business intelligence system that runs as a web application.</p><p><strong>What is Business Intelligence system?</strong></p><p>It is a tool with which users can define and query various datasources (SQL, csv, google sheets, APIs, etc.) and make meaningful insights in the form of reports and then combine them into dashboards which can be shared.</p><p>In essence it is a data exploration and reporting tool. Eg. Tableau, caravel, chartio, etc.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*dH7WFpITNJyUqHqJ." /></figure><p>[Fig 1: Logical Component View Mind Map]</p><h3>Component Design</h3><h3>Widget</h3><p>A Widget is the Core of BI Tool. A widget can exist as a standalone application or as part of the Dashboard, which is a set of Widgets arranged in some layout.</p><p>A Widget can comprise of static data/text and charts or dynamic and interactive visualizations like [6][7] or a pivot table [8] where users can slice and dice data interactively.</p><p>A Widget shall have a REST URI which identifies it uniquely and which can be shared and bookmarked. For example: <a href="https://awesomebi.mediaiq.com/widgets/v1.0/%7Bunique-widget-Id%7D/[%7Bsummary|detail%7D]"><strong>https://awesomebi.mediaiq.com/widgets/v1.0/{<em>unique-widget-Id</em>}/[{<em>summary</em>|<em>detail</em>}]</strong></a></p><p>A widget has a Widget State which defines the precise state of the widget like specific filters that are applied, and Widget Metadata which describes the type of the widget and its associated visualization and interaction grammar for summary and detail views.</p><h3>Widget State</h3><p>A widget state is a JSON document which completely defines the state of a widget at any given time. A widget state is an immutable object and any changes to the state creates a new state object along with a widget state change event which when applied to the old state of the widget results in the new state of the widget. When a user interacts with a widget by applying or changing a filter for example the change event is published which gets serialized over a websocket connection to the widget server. The server then reconciles the widget state on the server side master copy and then broadcasts the changes to all connected clients which are using that widget. The technique used is popularly known as “operational transformation” [1][3] which is a technology developed at Google for resolving conflicts when documents are edited collaboratively by multiple clients and more formally described in this paper [2].</p><h3>Widget Metadata</h3><p>A widget metadata encapsulated the grammar which describes the the type of visualizations and interactions using a set of encodings as a JSON document that map data attributes to visual channels like color, shape, and size and may include common binning transformations such as binning, aggregation, sorting, and filtering [4][5][6].</p><p>Each Widget can have two views defined. The summary view is the default. User’s can additionally define a detail (full screen) view of the widget which will allow the user to engage and interactive data exploration/visualization.</p><h3>Widget Type</h3><p>The frame supports many types of visualizations so long as there a specification attached. The framework also has a Widget Designer which can be used to create and modify widget specifications and view the existing widget gallery. To start with the framework shall come with the visualizations depicted here [6]. In addition we can look at Mike Bostocks library of interesting data visualizations for inspiration [7]. For Javascript Pivot Table (aka Pivot Grid, Pivot Chart, Cross-Tab) implementation with drag and drop refer PivotTable.js by Nicolas Kruchten [8].</p><h3>Widget State Manager</h3><p>The Widget State Manager is the orchestrator which helps the widget interact with its surrounding contexts and perform operations such as described below.</p><h4>Data Source Mapping</h4><p>Widget State Manager uses the Data Source Mapper to bind a widget to a data source. It is also used when the user wants to changes the data source of a widget or change the specific query used to fetch the data from the data source, or provide specific computations or aggregations that must be performed when loading the data from the Data Source.</p><h4>Widget State Change Synchronizer</h4><p>Widget State Manager uses the User Interaction Manager to publish the widget state change event stream which then gets serialized to the server by the Widget State Change Serializer. This enables real time collaborative editing and uses a similar technique that is used by any multiplayer online gaming system to sync its Game State to all connected players [9][10][11].</p><h4>User Interaction Manager</h4><p>User Interaction Manager is an arbitrator and decides which user events are important to be tracked and synchronized with the server as discussed above.</p><h4>Widget Access Manager</h4><p>Widget State Manager delegates to the Widget Access Manager to a) add/ remove user/ groups / permissions (view-only, view-and-change) for a user / group; and b) decide what actions a user can do on a widget provided the user has access to the widget. If the user does not have access then the widget is not rendered for that user and an appropriate message is shown instead.</p><h3>Dashboard</h3><p>A Dashboard is a collection of widgets. A user can add/remove widgets from the Dashboard with ease.</p><h4>Dashboard State</h4><p>A Dashboard State is similar to the Widget State and is a JSON document which tracks the current immutable state of the dashboard. When the user changes something on the Dashboard like a Filter or adds/removes a widget a dashboard state change event is published which gets synchronized over to the server and the server on receiving such events broadcasts the changes to all connected clients who have the dashboard open. This results in live update and enables users the edit Dashboards collaboratively.</p><h4>Dashboard Metadata</h4><p>Dashboard Metadata encapsulates visual elements like layout, themes, or global filters such as Date Range. When Global filters are changed all widgets simultaneously refresh data by making a new query on the data datasource using the modified date range.</p><h3>BI Application</h3><p>BI Application is the web application which the users interact with on their web browsers from Desktop, or Mobile Devices. BI Application provides all functionality related to serving a Dashboard with a set of Widgets (Reports/Visualizations). Each of the Widget is a self contained application and manages the complete state and user interactions by collaborating with other services such as BI Auth Service for User Authentication, BI Widget Service for managing the widget state lifecycle, and BI Data Services for interactively loading the data for the widget on demand.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*GE3Js9XFrKRX_8oM." /></figure><p>[Fig 2: BI Application with Typical Widget Loading Sequence]</p><h3>BI Auth Service</h3><p>BI Auth Service fronts all BI Application requests. When the user requests for a Dashboard or a Widget URI, the web browser prompts the user to login, and requests the Auth Service to generate an authentication token given the user’s login credentials. The BI Application then embeds this user authentication token in all subsequent requests and interactions (http requests) to other services.</p><h3>BI Widget Service</h3><p>BI Widget Service on receiving a request containing a valid authentication token from the BI Application serves the BI Widget Metadata and state for the given user by retrieving the data from the BI Widget Database.</p><h3>BI Data Service</h3><p>BI Data Service handles all data serving requests from the BI Application. The Services can serve data from a vast array of underlying data stores such as MySQL, Postgresql, Redshift, Hive, Spark, Druid. It can also load data from files locally or over an HTTP connection (REST API). The framework itself is built using extensible components and we can easily add support for any new data sources by providing the database driver, query parser, query executor, and any user defined functions for aggregation, binning, mapping, reducing, joining, etc.</p><h3>BI Cache Service</h3><p>BI Cache Service is used as an LRU Cache to cache most recent user query and custom aggregation results which boosts the query performance, especially when multiple concurrent users requests for the same data.</p><h3>Non Functional Requirements</h3><h3>Scalability</h3><p>The BI Application as depicted in Fig. 2 uses a number of collaborating services, and as such each service are independently deployed in their own clusters using dockerized containers and kubernetes. Google has been using container technology <a href="http://www.infoq.com/news/2014/06/everything-google-containers">for over ten years</a>, starting over 2 billion containers per week. With Kubernetes it shares its container expertise creating an open platform to run containers at scale. Kubernetes also supports auto scaling so nodes can be automatically added to or removed from the cluster in response to increase or decrease in load (CPU, Memory, I/O throughput) [12][13].</p><h3>High Availability</h3><p>BI Application achieves 99.99% HA by ensuring that all collaborating services are set up in auto load balancing and fault tolerant clusters using kubernetes. A lot of in-depth discussions on setting up high availability clusters can be found here [14][15][16][17].</p><h3>Zero Downtime Deployments</h3><p>It is possible to do zero downtime or Blue-Green deployments as described here [18][19]. Essentially a Blue Green deployment [Fig. 3] involves the following steps:</p><ol><li>Scale up a new Cluster</li><li>Wait until healthy</li><li>Switch Backend in Load Balancer</li><li>Dispose Old Cluster</li></ol><p>We can also perform Canary deployment [Fig. 4] which involves the following steps:</p><ol><li>Scale up a new Canary Cluster</li><li>Wait until healthy</li><li>Add the Canary Cluster to the Load Balancer</li><li>Gradually increase the size up the Canary cluster while reducing the size of the Old cluster</li><li>Repeat step 4 until all traffic has moved over to the Canary Cluster</li><li>Remove the Old cluster from Load Balancer and dispose it</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/837/0*3-cOvOsZyUaE4Yof." /></figure><p>[Fig. 3: Blue Green Deployment setup using Kubernetes]</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/659/0*cV8YlkBMbPM0snZf." /></figure><p>[Fig. 4: Canary deployment setup using Kubernetes]</p><h3>Monitoring</h3><p>Since all our applications and services are running in a distributed and clustered setup we need to use a central logging and event monitoring platform. Understanding how an app/service behaves after deployment is crucial for scaling the application and providing a reliable service. To do that, we need to measure and analyze the performance of our apps. This post [20] describes how we can do that with Heapster, InfluxDB, and Grafana.</p><h3>User Auditing &amp; Analytics</h3><p>We can publish user clickstream using beacon technology like Snowplow analytics’ event data pipeline [21]. We can event integrate the data collected through this event pipeline and setup our own custom Dashboard to track events that are important to us [22].</p><h3>Other Technical Design Considerations</h3><ul><li><strong>What Front End Language should we use for our Awesome BI Application?</strong></li></ul><p>As far as Front End Web Application frameworks go we are spoilt for choice [23] and with the advent of ES5 (ES2015) and Reactive frameworks being embraced by some of world’s leading websites we should be inclined to use a library that allows us to easily build template drive UI library with reactive updates. Reactive was originally created at theguardian.com to produce news applications. A typical news app is heavily interactive, combines HTML and SVG, and is developed under extreme deadline pressure. It has to work reliably across browsers, and perform well even on mobile devices.</p><p>Many tools exist to help you build web apps, but very few deal with the fundamental problem: HTML, while excellent at describing static documents, wasn’t designed for interactivity. Reactive changes that. It takes your <a href="http://mustache.github.com/">Mustache</a> templates and transforms them into a lightweight representation of the DOM — then when your data changes, it intelligently updates the real DOM [24].</p><p>Of course React.js, the framework invented at Facebook is arguably the most popular front end framework out there so that should deserve a special mention. Here’s more reason why most front end developers just love React.js [25][26]. Facebook uses some React.js, and Instagram’s entire website was built on React.js after the two companies joined forces. Other successful projects using React.js include Khan Academy and New York Times.</p><ul><li><strong>How do you ensure that the BI application is realtime and responsive?</strong></li></ul><p>The main tenets of a real time and responsive application is that it should respond to any user action at interactive speeds. This is especially important to drive user productivity as a BI tool helps users ask many questions on the data interactively and then arrive at the view that helps them answer their question, or sometimes helps them even frame the right question. Many things work together to make the whole application appear real time.</p><p>Starting with the database server should ensure that the data is aggregated at the right granularity for example all metrics are rolled up hourly instead of loading all the raw data into the application. When dealing with real time Big Data (a few terabytes to petabytes) and streaming fast data (a few thousand records/ second) it is essential to have a real time data ingestion pipeline using Kafka and Samza (for any pre-processing like Geo Coding, Sessionization, etc.) and storing the data in Druid for serving. Druid promises<a href="https://en.wikipedia.org/wiki/Online_analytical_processing"> Online Analytical Processing (OLAP)</a> capability in fast data contexts with real time stream ingestion and sub-second queries [27].</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/673/0*-Jzi7nQgqSl4TCOk." /></figure><p>[Fig. 5: Architecture of a Druid Platform]</p><ul><li><strong>How can we make the communication between BI Application and the BI Data Service more efficient?</strong></li></ul><p>The BI Data Server should use efficient protocol to stream the data to the BI Application so that the BI Application can progressively render the data as they arrive. <strong>Oboe.js</strong> is a framework which shows how streaming can speed up applications [28].</p><p>In addition to this BI Application should use standard data caches and reverse proxies to cache the most frequently requested data such that when multiple users request for the same data computing bandwidth and latency can be improved by serving the data from the cache.</p><h3>Final Thoughts</h3><p>The rise of modern business intelligence (BI) has seen the emergence of a number of component parts designed to support the different analytical functions necessary to deliver what enterprises require.</p><p>Perhaps the most fundamental component of the BI movement is the traditional frontend or visualization application. Companies like Tableau, Qlik, Birst, Domo and Periscope provide these. There are dozens more — all with essentially equivalent capabilities: the ability to make spreadsheets look beautiful. Some of these companies have been tremendously successful, primarily differentiating themselves on the axis of usability.</p><p>Another, equally critical component of the BI equation is the database. Here, too, there are the usual suspects: Redshift, Impala, Vertica, Netezza and others. Some of these databases are fully featured, system-of-record worthy solutions, while others focus on a particular performance axis, streaming, for instance, and do it well.</p><p>Finally, there is the emergence, across BI and database players, of more advanced analytics tools, driven by the explosion of interest in and development of machine learning, deep learning and artificial intelligence. This market has its stars, starting with the big boys — Google, Facebook, Amazon, Microsoft, IBM, Baidu, Tesla — as well as a host of highly credentialed startups, like Sentient, Ayasdi, Bonsai and H2O.ai.</p><p>A successful, fully functional BI system has each of these operating optimally.</p><p>The problem is that none of these systems are operating optimally.</p><p>The reason is the extraordinary growth in data.</p><p><strong><em>These systems are laboring because they are all based on an antiquated, CPU-centric view of the world that is computationally incapable of querying, rendering or learning from data at the scale demanded by the petabyte economy.</em></strong></p><p>There is a solution, however, one that the deep learning folks have already adopted: GPUs.</p><p>With GPUs, you get order of magnitude performance enhancements. There’s a reason why so many supercomputers on the Top500 list use NVIDIA GPUs. It’s because GPUs are far more adept at the mathematical tasks than traditional CPUs.</p><p>But, it’s more than just deep learning. Databases and visualization also benefit significantly from GPUs. A system based on GPUs can deliver the speed and scale required to handle these massive working sets and deliver the functionality required.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*TdB3RlC9HdRnMG36." /></figure><p>[Fig 6: MapD Taxi Demo]</p><h3>What You Need to Know About GPUs and Integrated Analytics</h3><p>First, GPUs offer exceptionally high memory bandwidth — we’re talking terabytes per second across multiple GPUs. This is important since database queries are typically memory bandwidth or I/O bound. Because of the memory footprint, GPUs can scan more data in less time, resulting in faster results.</p><p><strong><em>To put this in context, noted database authority </em></strong><a href="http://tech.marksblogg.com/"><strong><em>Mark Litwintschik</em></strong></a><strong><em> found that a single GPU server is 74x faster than a larger cluster of Redshift over 1.1 billion rows. Not 74% faster, 74 times faster. Against Postgres, that figure was 3,500x faster — milliseconds vs. tens of minutes.</em></strong></p><p>That’s significant because working sets have grown commensurately with data. A few million row dataset used to be big, now it is tiny. The new normal starts at several hundred million and runs to the billions.</p><p>Second, GPUs are not simply about straight-line speed. While, other systems can be optimized for specific tasks and queries, GPUs are also extraordinary at graphics. In fact, the native rendering pipeline of GPUs makes them the ideal engine for data visualization.</p><p>This manifests itself not only in better looking dashboards, but also in more responsive, faster dashboards. The reason is that if you can do the query on the same chip as the render, then you don’t have to move your data around. This might not be a problem when dealing with only a few million rows, but it’s a big problem when you cross a billion, let alone several billion.</p><p>Finally, GPUs deliver supercomputing-class computational throughput. GPUs dominate the machine learning and deep learning ranks because they excel at matrix multiplication. Again, the ability to co-locate the querying and machine learning on the same chip lets you enjoy exceptional efficiency in feeding the machine learning algorithm with the data needed for training and inference.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/566/0*qSDzwr3gMLSiLklX." /></figure><h3>How to Put GPUs into Play in Your Organization</h3><p>True performance will come from an integrated system, one that combines GPU hardware with a GPU-tuned database, a GPU-tuned front end/visualisation layer and a GPU-tuned machine learning layer.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*BZg4vmq-SUDq_BF9." /></figure><p>Upgrading just one component, however, creates the weakest link problem.</p><p>A GPU database feeding a CPU visualization frontend will be faster, but it won’t be as fast as a GPU database feeding a GPU visualization frontend.</p><p>Any potential combination creates the same challenge, introduces the same weak CPU link.</p><p>The optimal system benefits from GPU hardware and GPU-tuned software at every turn [29].</p><h3>References</h3><p>[1] <a href="http://decisionmapper.com/tutorials/derby1">http://decisionmapper.com/tutorials/derby1</a></p><p>[2] <a href="https://neil.fraser.name/writing/sync/eng047-fraser.pdf">https://neil.fraser.name/writing/sync/eng047-fraser.pdf</a></p><p>[3] <a href="https://wiki.dandascalescu.com/essays/why_meteor">https://wiki.dandascalescu.com/essays/why_meteor</a></p><p>[4] <a href="https://medium.com/hci-design-at-uw/introducing-vega-lite-438f9215f09e#.mgcxn7cbx">https://medium.com/hci-design-at-uw/introducing-vega-lite-438f9215f09e#.mgcxn7cbx</a></p><p>[5] <a href="http://idl.cs.washington.edu/files/2017-VegaLite-InfoVis.pdf">http://idl.cs.washington.edu/files/2017-VegaLite-InfoVis.pdf</a></p><p>[6] <a href="http://vega.github.io/vega-editor/?mode=vega&amp;spec=crossfilter&amp;renderer=svg">http://vega.github.io/vega-editor/?mode=vega&amp;spec=crossfilter&amp;renderer=svg</a></p><p>[7] <a href="http://bl.ocks.org/mbostock">http://bl.ocks.org/mbostock</a></p><p>[8] <a href="https://github.com/nicolaskruchten/pivottable">https://github.com/nicolaskruchten/pivottable</a></p><p>[9] <a href="https://www.ibm.com/developerworks/library/ar-powerup1/">https://www.ibm.com/developerworks/library/ar-powerup1/</a></p><p>[10] <a href="http://comp.utm.my/pars/files/2013/04/An-Overview-of-Networking-Infrastructure-for-Massively-Multiplayer-Online-Games.pdf">http://comp.utm.my/pars/files/2013/04/An-Overview-of-Networking-Infrastructure-for-Massively-Multiplayer-Online-Games.pdf</a></p><p>[11] <a href="https://gameserverarchitecture.com/2015/10/pattern-distributed-network-connections/">https://gameserverarchitecture.com/2015/10/pattern-distributed-network-connections/</a></p><p>[12] <a href="https://www.infoq.com/articles/scaling-docker-with-kubernetes">https://www.infoq.com/articles/scaling-docker-with-kubernetes</a></p><p>[13] <a href="http://blog.kubernetes.io/2016/07/autoscaling-in-kubernetes.html">http://blog.kubernetes.io/2016/07/autoscaling-in-kubernetes.html</a></p><p>[14] <a href="http://kubernetes.io/docs/admin/high-availability/">http://kubernetes.io/docs/admin/high-availability/</a></p><p>[15] <a href="http://www.tothenew.com/blog/setup-kubernetes-cluster-on-aws-ec2/">http://www.tothenew.com/blog/setup-kubernetes-cluster-on-aws-ec2/</a></p><p>[16] <a href="http://stackoverflow.com/questions/26705201/whats-the-difference-between-apaches-mesos-and-googles-kubernetes">http://stackoverflow.com/questions/26705201/whats-the-difference-between-apaches-mesos-and-googles-kubernetes</a></p><p>[17] <a href="https://github.com/kubernetes/kubernetes/blob/release-1.4/docs/design/aws_under_the_hood.md#architecture-overview">https://github.com/kubernetes/kubernetes/blob/release-1.4/docs/design/aws_under_the_hood.md#architecture-overview</a></p><p>[18] <a href="http://www.slideshare.net/ArjanSchaaf/zero-downtimejavadeploymentswithdockerandkubernetes">http://www.slideshare.net/ArjanSchaaf/zero-downtimejavadeploymentswithdockerandkubernetes</a></p><p>[19] <a href="https://www.youtube.com/watch?v=7TOWLerX0Ps">https://www.youtube.com/watch?v=7TOWLerX0Ps</a></p><p>[20] <a href="https://deis.com/blog/2016/monitoring-kubernetes-with-heapster/">https://deis.com/blog/2016/monitoring-kubernetes-with-heapster/</a></p><p>[21] <a href="http://snowplowanalytics.com/">http://snowplowanalytics.com/</a></p><p>[22] <a href="https://github.com/snowplow/snowplow/wiki/snowplow-tracker-protocol">https://github.com/snowplow/snowplow/wiki/snowplow-tracker-protocol</a></p><p>[23] <a href="https://colorlib.com/wp/javascript-frameworks/">https://colorlib.com/wp/javascript-frameworks/</a></p><p>[24] <a href="http://www.ractivejs.org/">http://www.ractivejs.org/</a></p><p>[25] <a href="https://www.syncano.io/blog/reactjs-reasons-why-part-1/">https://www.syncano.io/blog/reactjs-reasons-why-part-1/</a></p><p>[26] <a href="https://camjackson.net/post/9-things-every-reactjs-beginner-should-know">https://camjackson.net/post/9-things-every-reactjs-beginner-should-know</a></p><p>[27] <a href="https://blog.codecentric.de/en/2016/08/realtime-fast-data-analytics-druid/">https://blog.codecentric.de/en/2016/08/realtime-fast-data-analytics-druid/</a></p><p>[28] <a href="http://oboejs.com/why">http://oboejs.com/why</a></p><p>[29] <a href="https://www.mapd.com/blog/2016/10/07/the-argument-for-integrated-analytics-database-data-viz-data-science">https://www.mapd.com/blog/2016/10/07/the-argument-for-integrated-analytics-database-data-viz-data-science</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5c1bf7a28c5b" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>