<?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 Ivan on Medium]]></title>
        <description><![CDATA[Stories by Ivan on Medium]]></description>
        <link>https://medium.com/@arcupcakes?source=rss-1d905a9076f4------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/2*fOM40YIKvunAqwD7oB0CBw.png</url>
            <title>Stories by Ivan on Medium</title>
            <link>https://medium.com/@arcupcakes?source=rss-1d905a9076f4------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 20:02:40 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@arcupcakes/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[Applications performance general overview]]></title>
            <link>https://medium.com/@arcupcakes/applications-performance-general-overview-601507411a34?source=rss-1d905a9076f4------2</link>
            <guid isPermaLink="false">https://medium.com/p/601507411a34</guid>
            <dc:creator><![CDATA[Ivan]]></dc:creator>
            <pubDate>Mon, 13 Apr 2020 05:38:18 GMT</pubDate>
            <atom:updated>2020-04-19T16:54:48.856Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Discaimer: it’s long read, with just genral overview, “technology specific” and “guideline articles” will be published in separate posts.</strong></p><p>Performance is an important characteristic of client applications. It might be considered as an UX summary and value for user and service provider.</p><p>Main goal of this article is to help to classify performance optimization areas, to determine feasibility of optimizations and to find possible ways to make improvements in areas where performance optimization will make sense. In the summary you can find a cheat-sheet (tbd) and performance assessment canvas.</p><p>Performance might be considered as equity with multiple variables, so I decided to create a mind map with short set of main characteristics.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*bthJmgREcCzUkTxH" /></figure><p>First of all let me briefly go over all characteristics and give a short introduction for every item:</p><ul><li><em>Business impact</em> — how critical for business performance is, what the main metrics for business are and how we can help to improve existing solution for business.</li><li><em>Technological environment</em> — it’s where our application will work, for example it is an embedded platform or a general purpose system; which networks are used, fixed LAN or Wireless networks. What its limitations and features are, like background processes limitations, ability to go to energy-save mode.</li><li><em>Implementation aspect</em>s — it’s everything about concrete implementation. Which technologies are used, like cross-platform design or native implementation. Any contracts/agreements in architecture design, within memory or persistence cache.</li><li><em>User engagement </em>— it’s summary of UI/UX system and the way how user feels about and experience application usage. Real user expectations and experience result in measurable metrics such as DAU, MAU, WAU and user retention.</li><li><em>Trade-offs</em> <em>and Expectations</em> — agreements to achieve the expected performance, also dev team efforts and budget for it. Set of assumptions for planning and key priorities for team.</li><li><em>Monitoring &amp; loopback </em>— analytics collection for performance, user operations, rate of errors and unsuccessful operations. Set of monitoring dashboards and notification triggers.</li><li><em>Adjustment mechanics </em>— I truly believe there might be a toggle to turn on/off features in assembled and distributed client application to prevent load rise in case of broken functionality or irregular situation.</li></ul><p>Secondly let me write about every characteristic in a more detailed way, with some examples.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Q_6gJ5mRjZuFotQMn1FwvQ.jpeg" /></figure><h3>Business impact</h3><p>Business needs fastest processing for client’s operations, the faster we process every user the faster we can close deal.</p><p>If business isn’t interested in performance optimization, probably there is no need to do it. Or business just does not know that it’s possible to boost its operations. So it’ll be better to start with a talk with product managers about performance impact and characteristics which are critical for closing deals. It is crucial to determine metrics helping to avoid losing users or preventing them from necessary actions.</p><p>So there should be performed an assessment what business and user operations are critical and to set tasks to work on.</p><p>Without input from business and users your development team will only play with “engineering performance muscles” and that’s it.</p><p>Devs should know business and user critical values, so make sure your development, product and support team are aware of that metrics.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oCHCXyLpawmG96IUegHk-g.jpeg" /></figure><h3>Technological environment</h3><p>Let’s consider everything “around” our solution, probably it will be possible to improve performance without any change</p><p><strong>Network factor</strong></p><ul><li>Bandwith — max throughput on “packet” route</li><li>Latency — time to transfer “packet” from src to dst</li><li>Last mile latency — all your optimizations might be ruined by a very complex corporate network or ISP provider network infrastructure in user’s neighborhood</li><li>Protocol overhead — UDP vs TCP/IP, HTTP vs HTTP 2.0, TLS</li><li>Geography factor, how far our servers are</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/464/0*fnczzhoLXFa4_acs" /></figure><ul><li><strong>Computation host.</strong> CPU vs GPU vs WhatEverSpecializeProcessor Is it good enough to perform calculations on client or probably you should use server instead. There are SaaS services for machine learning in cloud, so there is no need to apply image filters in client app. Also CPU is not always the solution, for different tasks different hardware should be used.</li><li><strong>OS architecture</strong> impact, in some of the cases micro-kernel OS might be more beneficial from performance viewpoint, however maintainability might be a nightmare. Or lightweight OS should be considered</li><li><strong>What-ever-hardware is used limitations</strong> — any hardware which is involved into project is involved, your network and severs might be fast as thunderbolt, but your NFC reader or BLE device will spend tons of time to perform key operation, for example 5 seconds. And will it be 5.2 seconds with super-tuned server or 5.5 seconds with a regular server will not make any sense</li></ul><p>Regarding every item in that list it’s possible to write a separate article or even a book. Let’s just consider multiple factors listed above, as variables in our Performance equity.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-_qw6YCRe7tFZiH2Dyim6g.jpeg" /></figure><h3>Implementation aspects</h3><p>It’s very interesting and the most flexible part in our performance equity, since we are responsible for that part. So here are the key factors we can consider in our solution implementation</p><p><strong>Architecture. </strong>Solution architecture makes one of the most important impacts:</p><ul><li><em>Scalability</em> — is it easy scalable or not, could it be done in real time or not.</li><li><em>Patterns</em> — sorry for obvious words, but patterns matter. Are your jobs done in pub-sub/push-pull/cron manner or not. The way which patterns are used will have an effect on application performance.</li><li><em>Cache ready</em> — is your solution cache ready or not. It’s important keep in mind that cache is not always good. You can maintain cache system, which in reality will be useless, because of data should be fresh and should be always loaded from scratch so cache is redundant.</li><li><em>Reliability</em> — is solution ready to work with absent functionality or not. If some features do not work, it should not produce additional impact on UI rendering or application work. Broken features should not produce any impact on system. There should not be any infinite request which will just brute-force already dead component and freeze entire UI..</li></ul><p><strong>Technology stack. </strong>Which tools, frameworks, services and technologies are used for solution implementation.</p><ul><li><em>Performance over simplicity</em>. Should C++ be considered over NodeJS, since C++ is compiled directly into processor instructions. C++ require less RAM, so no GC calls and you have best performance, but you’ll have to have C++ unique developers.</li><li><em>Common behavior.</em> Should we use any cross-platform solution to have the same behavior everywhere to prevent next data processing due to host platform specific.</li><li><em>Protocol</em> — which protocol will be used. Are there any requirements that data should be delivered, and delivered in proper order. Are there any requirements to data overhead, should TCP be considered over HTTP or UDP should be used</li><li><em>Target platform and its tools limitations.</em> For example QNX is very productive, but I’m not sure if it’s possible to launch latest Java JRE there.</li></ul><p><strong>Security</strong>. Will security add any additional impact on application speed. Should application communicate via any secure channels which may have limited bandwidth. Or is it dependent on any Trusted Center to monitor application and will make down crease of performance. Also there might be a special cryptographic operations usage which consume a lot of CPU or memory, so it will affect performance.</p><p><strong>Third party dependencies</strong> 3rd party libraries and services performance matters. The moment when 3rd party is included into solution, there should be a formed set of If something goes wrong, it’s better to have ability to turn off the feature or keep it with limited functionality, or even consider an option to switch to alternative solution.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vFfhOzaQvV1GeyaFrTbNXg.jpeg" /></figure><h3>User engagement</h3><p>Everyone loves to get everything fast, so user engagement is very important. To make an application interesting and to retain users it might be useful to measure metrics like (I did not invent anything, just took them from <a href="https://web.dev/user-centric-performance-metrics/">here</a>):</p><ul><li><em>FCP</em> — First Content Paint, it measures how fast UI started be informative to user</li><li><em>LCP</em> — Largest Contentful Paint, it measures overall time to fully draw largest UI element</li><li><em>FPS </em>— Frames Per Second, it’s mostly used in entertainment industry, but it’s applicable for client UI, it’s how smooth your application UI is. The bigger FPS the better, in ideal case you should have around 50–60 FPS.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xXiufnjwfoevmK86.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wMx_bO0nZTIk3O2GRHtAxA.png" /></figure><p>Metrics listed above create overall/summary type characteristics:</p><ul><li><em>Load responsiveness</em> — how fast a user is able to start use your solution</li><li><em>Runtime responsiveness</em> — how responsible an application is in runtime</li><li><em>Visual stability</em> — there should not be any “jumps”/”movents” in UI layout displayment</li><li><em>Smoothness</em> — interaction, animations and transitions are sexy without any glitches.</li></ul><p>Another factor to think about is overall application reliability and predictability, application should work in any conditions and do not crash. If any features are not available, user should see distinct and user-friendly UI with information what happens. UI should be ready to provide UI as soon as it is possible, block by block</p><p>I don’t like to follow die-fast/crash fast approach on production builds, because it’s better to tell user something instead of just application crash, it will be more informative and useful for user than just crash.</p><p>In general such approaches and metrics will give you ability to focus on service itself and improvement of metrics like DAU, MAU, Retention and etc. With poor UI/UX performance it’s hard to make changes in product.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Bp2yeaGo-aGgBXSV" /></figure><h3>Trade offs and expectations</h3><p>To manage performance inside your solution, it’s always better to have rules declared like:</p><ul><li><em>Set of critical operations </em>to be monitored</li><li><em>Expected performance values</em> for particular operations, with ability to be monitored</li><li><em>Set of trade-off</em>s not supposed to be optimized, with explanation why that decision was made.</li></ul><p>Having agreements for trade-offs and expectations for product set in advance will not ruin “delivery-train” in the last moment.</p><p>Metrics declaration gives ability to make proper architecture decision on product development stage and prevent performance issues. Such approach will help you to get rid of hot fix releases after you rolled-out new release.</p><p><em>NOTA BENE</em>: if any part of your system/service will be used by external/another components, it’s better to get agreement on their expectations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*_amO5GVlPW_LoaXOZAnzjg.jpeg" /></figure><h3>Monitoring loopback</h3><p>It’s very important to include performance and quality monitoring tools into your solution. It will give you a real picture of what happens in your product, will enable you to track changes and impact of newly introduced changes.</p><p>There are multiple solutions for different areas. Not all applicable for everything. There are general purpose analytics tools like Google Firebase, NewRelic, but also specialized for video like Conviva and Adobe Analytics.</p><p>Monitoring is useless without analysis reports and triggers threshold values.</p><p>It’s very important to collect things like FCP, LCP, PAR, FPS. But also it will be very useful to collect:</p><ul><li><em>Failed Operations</em> — it will give you understanding how stable your APIs are</li><li><em>Retry Operation Rate</em> — how often operations fail, so probably there is a potential bottle neck</li><li><em>Network Interaction Metrics</em> — it’s always very interesting to measure how fast API requests are executed, even protocol could be improved</li><li><em>Selected Operations</em> — set of operations you agreed to measure and which threshold values are important to track.</li></ul><p>In one of my pet projects I implemented simple performance measurements collection into Firebase and several functions with threshold analysis, in case of bad behavior I was getting an email.</p><p>It’s possible to integrate such analysis on all stages: development, staging and production. With tools available it will be possible to identify next steps for service improvement and focus on real needs. You’ll always have information which is really needed, without any additional research. With metrics collected you’ll be able to predict your Quality of Service.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*uNFdDa49AXvRURcW" /></figure><h3>Adjustment mechanics</h3><p>Service should be predictable and useful for everyone in any moment, even when something is not available. So it’s possible to consider adjustment mechanics in your solution, like:</p><ul><li><em>Limited functionality</em>. For example, turn off resource consuming features on low end devices, or in moment of high service demand you can turn off secondary/not business critical features/ services, to use more resources on business critical tasks</li><li><em>Have alternative</em> feature/functionality/user-flow to perform the same operation, if general flow is not working for some reason. It will give ability to “adjust” service in timely manner with probably some additional interactions, but without “no service” period.</li></ul><h3>Summary</h3><p>Performance is not a very easy thing to work with. However with systematic approach like environment and technologies assessment, expectations agreements, metrics collection and analysis it’s possible to manage “Performance” in your project in predictable way.</p><p>In the next post I’ll create a case study for several “Android applications” and show how it’s possible to apply Performance Canvas in real life.</p><p><em>Originally published at </em><a href="https://ivanksk.com/2020/04/13/performance-intro/"><em>http://ivanksk.com</em></a><em> on April 13, 2020.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=601507411a34" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The first entry]]></title>
            <link>https://medium.com/@arcupcakes/the-first-entry-134faa6fcd5?source=rss-1d905a9076f4------2</link>
            <guid isPermaLink="false">https://medium.com/p/134faa6fcd5</guid>
            <dc:creator><![CDATA[Ivan]]></dc:creator>
            <pubDate>Wed, 30 May 2018 20:08:30 GMT</pubDate>
            <atom:updated>2018-05-30T20:08:30.489Z</atom:updated>
            <content:encoded><![CDATA[<p>Hello world :)</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=134faa6fcd5" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>