<?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 Prabha on Medium]]></title>
        <description><![CDATA[Stories by Prabha on Medium]]></description>
        <link>https://medium.com/@bhavyakm?source=rss-4277a547bb------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*mTifFUP_u_sciCg8HI-njA.png</url>
            <title>Stories by Prabha on Medium</title>
            <link>https://medium.com/@bhavyakm?source=rss-4277a547bb------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 11:27:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@bhavyakm/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[Observability — What is the deal here?]]></title>
            <link>https://medium.com/@bhavyakm/observability-what-is-the-deal-here-6e205ce6cc?source=rss-4277a547bb------2</link>
            <guid isPermaLink="false">https://medium.com/p/6e205ce6cc</guid>
            <category><![CDATA[big-data]]></category>
            <dc:creator><![CDATA[Prabha]]></dc:creator>
            <pubDate>Sun, 01 Sep 2019 01:00:52 GMT</pubDate>
            <atom:updated>2019-09-01T01:00:52.051Z</atom:updated>
            <content:encoded><![CDATA[<h3>Observability — What is the deal here?</h3><p>Someone saw my resume and asked why did I work on next-gen observability toolset!</p><p>I was fumbling —</p><blockquote>To reduce Op-ex?</blockquote><blockquote>To introduce standards, best and better practices of observability?</blockquote><blockquote>For better failure detection and recovery?</blockquote><p>Hmmm…..Is it?</p><p>What I hear…What I learnt…is…</p><blockquote>Operationalizing open source is expensive than operationalizing vendor product — Operationalizing does not just mean human resources, infra, support. There is more to it with key thing being what is the right open source to operationalize and what is the value? Does just operationalizing provide value or is there more to it? I will use two open source projects as examples in this space -&gt;</blockquote><p><strong>ElasticSearch</strong> — Underlying technology is widely needed for search use case. Operation is hard on many aspects and many vendor companies build business on top of it. I recently learnt SignalFx core is based on OSS ElasticSearch. Probably just operationalizing is value by itself here!</p><p><strong>Prometheus</strong> — Good concept for specific problem statement and easy to operate. Naturally…why operationalize when there is no pain to operate…:) Only reason could be to have a centralized team manage and maintain it for that problem statement. However, to get the full benefit this is not the only open source project you have to operationalize. You need things like Grafana, Alert manager, etc to be operationalized as well.</p><blockquote>There is never an “end solution” for operations. It evolves always. You identify it along the way and retire old ones and adopt newer ones. Whether it is your good old grep or sql query or lucene query or profiling or whatever</blockquote><blockquote>Debugging data is always needed in some form whether on the file system or a database with easy access to it.</blockquote><p>With all that ever changing need, what evolved?</p><ol><li><em>Debugging data in multiple servers was painful, so we centralized.</em></li></ol><p><em>2. Centralized storage infrastructure is hard to operate, so we de-centralized storage with a need for federated search.</em></p><p><em>3. Single infrastructure type is not good for all debugging data, so we use different types for different data with a need for correlation.</em></p><p><strong>Reality</strong> — all this involves $$$ and we need to generate quality debugging data by good instrumentations.</p><p><strong>Myths</strong> —</p><ol><li>If only I had known what is right and needed…:). So 100% quality data is a myth.</li><li>Single interface to access the debugging data quickly for better debugging experience is a myth. Latency and point of failures involved with this approach will over shadow other benefits.</li><li>Things I could not think of now…</li></ol><p>So….</p><blockquote>#1 — We need to strive for (mythical) quality and standardized debug data -&gt; depends on developers -&gt; Advocacy and Evangelism</blockquote><blockquote>#2 — We need to maintain and manage multiple database infrastructure -&gt; depends on infrastructure engineers -&gt; SysAdmins or Vendors?</blockquote><blockquote>#3 — We need front end interfaces -&gt; depends on product designers, product management and front end engineers -&gt; Build or Buy?</blockquote><p>There are zillion vendors and start up companies already addressing #2 and #3 in multiple ways</p><p>There is a stir to focus on #1, but that is a myth to some extent. I have practically seen many cases where instrumentation is overhead if immediate value is not realized.</p><p>How much are we over engineering this space?!</p><p>I will leave with this question and look forward to some thoughtful discussions.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6e205ce6cc" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>