DEV Community: Rupesh Mangal The latest articles on DEV Community by Rupesh Mangal (@rupesh). https://dev.to/rupesh https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F547847%2F5f8c0c8e-58f3-449c-9933-c4660661e7cf.jpeg DEV Community: Rupesh Mangal https://dev.to/rupesh en Sync vs Async Django servers - Load testing Rupesh Mangal Sun, 27 Dec 2020 18:06:52 +0000 https://dev.to/rupesh/sync-vs-async-django-servers-load-testing-35lh https://dev.to/rupesh/sync-vs-async-django-servers-load-testing-35lh <p>If you are using <strong>Gunicorn</strong> as <strong>API server Gateway</strong> and you want to reduce the <strong>response time</strong> drastically. then you are in the right place.</p> <p>A while ago our application experienced a surge in the number of active users and our API <strong>response time</strong> started to rise up and then <strong>Asynchronous server</strong> came to our rescue.</p> <p>This post covers the <strong>Load testing</strong> result comparison between <strong>Sync and Async Gunicorn servers</strong>.</p> <p><strong>Load testing</strong> was done using <a href="https://locust.io/" rel="noopener noreferrer">Locust</a>. one of the best tools out there IMO</p> <h1> Synchronous Gunicorn Server </h1> <p><a href="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2F50qhk61t0jpvne6j4sdx.jpeg" class="article-body-image-wrapper"><img src="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2F50qhk61t0jpvne6j4sdx.jpeg" alt="Sync Gunicorn server"></a></p> <h6> <a href="https://medium.com/@nhudinhtuan/gunicorn-worker-types-practice-advice-for-better-performance-7a299bb8f929" rel="noopener noreferrer">credits</a> </h6> <h2> 100 concurrent Users </h2> <h3> RPS(Requests per second): <strong>12</strong> </h3> <h3> Response time: </h3> <p>--&gt; median: <strong>11000</strong> ms<br> --&gt; 95 %tile: <strong>17000</strong> ms (95% requests finish before this time)</p> <h4> RPS </h4> <p><a href="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2Fkljtptd43uxoffr146gv.png" class="article-body-image-wrapper"><img src="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2Fkljtptd43uxoffr146gv.png" alt="RPS 100 users Sync Gunicorn"></a></p> <h4> Response Time </h4> <p><a href="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2F3lgvj3l2dyfah5e5uqs3.png" class="article-body-image-wrapper"><img src="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2F3lgvj3l2dyfah5e5uqs3.png" alt="Response time 100 users Sync Gunicorn"></a></p> <h2> 350 concurrent Users <em>(max supported with synchronous server)</em> </h2> <h3> RPS: <strong>6</strong> </h3> <h3> Response time: </h3> <p>--&gt; median: <strong>25000</strong> ms<br> --&gt; 95 %tile: <strong>30000</strong> ms (95% requests finish before this time)</p> <h4> RPS </h4> <p><a href="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2Fof8zf6t2ods6l86i8lal.png" class="article-body-image-wrapper"><img src="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2Fof8zf6t2ods6l86i8lal.png" alt="RPS 350 users Sync Gunicorn"></a></p> <h4> Response time </h4> <p><a href="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2F02o6s07hm937vwk7o9sp.png" class="article-body-image-wrapper"><img src="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2F02o6s07hm937vwk7o9sp.png" alt="Response time 350 users Sync Gunicorn"></a></p> <h2> 400 concurrent users </h2> <p><em>server started to respond with **ConnectionError</em>* for 100% for requests*</p> <h1> Asynchronous Gunicorn Server(<em>Gevent</em>) </h1> <p><a href="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2F0gi1wvmav0snlos99djb.jpeg" class="article-body-image-wrapper"><img src="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2F0gi1wvmav0snlos99djb.jpeg" alt="Asynchronous Gunicorn Server"></a></p> <h6> <a href="https://medium.com/@nhudinhtuan/gunicorn-worker-types-practice-advice-for-better-performance-7a299bb8f929" rel="noopener noreferrer">credits</a> </h6> <h2> <em>100 concurrent Users</em> </h2> <h3> RPS: <strong>50</strong> </h3> <h3> Response time: </h3> <p>--&gt; median: <strong>2200</strong> ms<br> --&gt; 95 %tile: <strong>5000</strong> ms (95% requests finish before this time)</p> <h4> RPS: </h4> <p><a href="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2Fa8sk2e2jot7sds3ug5dv.png" class="article-body-image-wrapper"><img src="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2Fa8sk2e2jot7sds3ug5dv.png" alt="RPS 100 users Async Gunicorn"></a></p> <h4> Response time: </h4> <p><a href="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2F22beuhk65fgszuv5mp9i.png" class="article-body-image-wrapper"><img src="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2F22beuhk65fgszuv5mp9i.png" alt="Response time 100 users Async Gunicorn"></a></p> <h2> <em>500 concurrent Users</em> </h2> <h3> RPS: <strong>50</strong> </h3> <h3> Response time: </h3> <p>--&gt; median: <strong>6500</strong> ms<br> --&gt; 95 %tile: <strong>18000</strong> ms (95% requests finish before this time)</p> <h4> RPS: </h4> <p><a href="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2Fqy3sz0z1zrvzjjtwr11q.png" class="article-body-image-wrapper"><img src="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2Fqy3sz0z1zrvzjjtwr11q.png" alt="RPS 500 users Async Gunicorn"></a></p> <h4> Response time: </h4> <p><a href="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2F24e38x2391a7rurltapr.png" class="article-body-image-wrapper"><img src="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2F24e38x2391a7rurltapr.png" alt="Response time 500 users Async Gunicorn"></a></p> <h2> <em>1000 concurrent Users</em> </h2> <h3> RPS: <strong>~25-30</strong> </h3> <h3> Response time: </h3> <p>--&gt; median: <strong>20000</strong> ms<br> --&gt; 95 %tile: <strong>50000</strong> ms (95% requests finish before this time)</p> <h4> RPS: </h4> <p><a href="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2Fk0f33ylb7plnyp404lry.png" class="article-body-image-wrapper"><img src="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2Fk0f33ylb7plnyp404lry.png" alt="RPS 1000 users Async Gunicorn"></a></p> <h4> Response time: </h4> <p><a href="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2Fyb23y10ygryps7ycvcyu.png" class="article-body-image-wrapper"><img src="https://media.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fi%2Fyb23y10ygryps7ycvcyu.png" alt="Response time 1000 users Async Gunicorn"></a></p> <h1> Observations: </h1> <p>1) <strong>RPS</strong> starts to <strong>decrease</strong> with <strong>increased load</strong>.<br> 2) <strong>RPS</strong> <strong>increased</strong> by at least <strong>400%</strong> in <strong>Async</strong> server.<br> 3) With the <strong>same</strong> number of <strong>users</strong> response time was just <strong>20%</strong> in the case of the async server (<strong>500%</strong> better performance).<br> 4) with the <strong>same</strong> response time <strong>async</strong> server was able to serve <strong>400%</strong> more users than <strong>sync</strong> server. (<strong>500</strong> users in async vs <strong>100</strong> users in sync).</p> django gunicorn server testing