<?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 Daniel Tapia on Medium]]></title>
        <description><![CDATA[Stories by Daniel Tapia on Medium]]></description>
        <link>https://medium.com/@danielftapiar?source=rss-de2914d425d2------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*beY0dd7n3uGIZRPZdR3cjA.jpeg</url>
            <title>Stories by Daniel Tapia on Medium</title>
            <link>https://medium.com/@danielftapiar?source=rss-de2914d425d2------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 08:43:29 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@danielftapiar/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[Linkerd HAZL: How to save costs on cross zone traffic without sacrificing performance.]]></title>
            <link>https://tech.loveholidays.com/linkerd-hazl-how-to-save-costs-on-cross-zone-traffic-without-sacrificing-performance-f8f7d8af7320?source=rss-de2914d425d2------2</link>
            <guid isPermaLink="false">https://medium.com/p/f8f7d8af7320</guid>
            <category><![CDATA[linkerd]]></category>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[platform-engineering]]></category>
            <category><![CDATA[cost-optimization]]></category>
            <dc:creator><![CDATA[Daniel Tapia]]></dc:creator>
            <pubDate>Fri, 03 Jul 2026 10:36:33 GMT</pubDate>
            <atom:updated>2026-07-13T13:05:21.636Z</atom:updated>
            <content:encoded><![CDATA[<p>At loveholidays we get to build a lot of cool stuff, but some quarters, after we wipe the sweat off our brows, and look at what we have built, we also have to look at our costs. And for our <strong>SKU: Network Inter Zone Data Transfer Out</strong> those costs looked like this.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fgcTBCSb14DbpqG4rM9Z5g.png" /></figure><p>We were spending around <strong>£15,000–£19,000</strong> per month on this SKU, which prompted my team to get in the cave and start figuring this one out.</p><p>This particular SKU is only billed if you have deployed workloads on multiple zones within a region. For example, zone-a, and then send traffic to some other workload on zone-b. That counts as Inter Zone Transfer Out, since you are sending network traffic over the network to/from a different zone, and get billed for it. This is common on most Kubernetes high-availability deployments that use multiple replicas and zones.</p><h3>The Problem</h3><p>So let’s imagine we have 3 zones and 2 services, service A and service B, each zone has 1 pod of each service. The general problem is that by default, the behavior when service-A talks to another service-B it does so in round robin fashion. So each service on each zone has 3 connections, if we have 1 pod in each zone that is 9 connections.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xa-bxgcHWwUSsLsJZAPI_Q.png" /></figure><p>Out of those 9 connections 6 of them will be cross zone, so that means 66.6% of the traffic will be cross zone traffic.</p><h3>The Solution</h3><p>We use <a href="https://linkerd.io/">Linkerd</a> as our service mesh for our platform. Linkerd is an open-source, ultra-lightweight service mesh designed specifically for Kubernetes. It acts as a dedicated infrastructure layer that manages, secures, and monitors communication between your microservices — all without requiring any changes to your application code. They have an enterprise feature called <strong>High Availability Zonal Load Balancing</strong> (<a href="https://docs.buoyant.io/buoyant-enterprise-linkerd/latest/features/hazl/">HAZL</a>), which we took for a ride to see how well it performed and if it really does what it claims to do.</p><p>To enable HAZL just get the latest version of Linkerd and deploy it with these flags</p><pre>destinationController:<br> additionalArgs:<br> — -ext-endpoint-zone-weights<br><br>proxy:<br>  additionalEnv:<br>    - name: BUOYANT_BALANCER_LOAD_LOW<br>      value: &quot;0&quot;<br>    - name: BUOYANT_BALANCER_LOAD_HIGH<br>      value: &quot;0&quot;</pre><p>This will enable HAZL globally on the control plane, the additional env vars will disable your data plane proxy services from adopting HAZL so we can enable them on a service by service basis. This makes it a safer rollout, especially when we have over 500+ services.</p><p>And to enable on a service by service basis, just add this annotation to your deployment and done! You are now HAZL enabled!</p><pre>  template:<br>    metadata:<br>      annotations:<br>        balancer.buoyant.io/load-band: 1,2</pre><h4>What the load band is?</h4><p>This question puzzled me at first, but it’s actually quite simple. From the Buoyant HAZL docs.</p><blockquote>HAZL is a client-side behavior. Each outbound proxy independently decides which backend endpoints are eligible for traffic. It works by adding and removing IP endpoints from an active pool. Once an endpoint is in that pool, Linkerd’s normal p2c/EWMA balancing selects the specific endpoint based on an observed latency.</blockquote><p>The way linkerd operates is that it has a pool of connections from source to target. What HAZL does is that it you configure a band, with low / high values, if the average load from pod to pod goes over the band, then it makes lower-priority/cross-zone endpoints eligible, analogously, when the traffic goes below the low band, HAZL removes lower-priority/cross-zone endpoints.</p><p>HAZL exposes these metrics</p><pre>outbound_http_balancer_adaptive_load_band_low <br>outbound_http_balancer_adaptive_load_band_high <br>outbound_http_balancer_adaptive_load_average</pre><p>So visually it looks like this.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eg4ax-nenGMns-1yBzio0g.png" /><figcaption>HAZL: Load Band Visualisation</figcaption></figure><p>From this graph you can see the blue band is the high band, yellow is the low band and the green line is the current usage. How remote endpoints are added and removed is further simplified when you view this graph by the different areas it produces.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EFHpafIuxFBNwG_aOAJXag.png" /></figure><p>So to summarise, if the load average goes over the high band, then HAZL will add remote zone endpoints to the pool that the client is able to send traffic to, while the load average is within the green zone it won’t add or remove endpoints, it will just use whatever is available in the destination pool and finally, if the load average goes below the low band it will remove remote endpoints.</p><p>A common point of confusion for me was why if the traffic is in the green zone am I still getting remote zone traffic? The answer to that question was that if the application is highly burstable and initial load average goes over the high band it will add remote zone endpoints to the pool, and if the traffic never goes below the low band, then during execution that pool will have both remote and local traffic, even though after the initial burst of the application has passed and traffic is within the green zone for the reminder of the execution, so choosing a good low band number is also critical here.</p><h3>How it Works and what traffic is eligible</h3><p>The final state of a working HAZL deployment will make all zone traffic go to the same zone, and if there isn’t a pod within the same zone, you will have to pay the price and have it be cross zone traffic.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/882/1*gZ8lpIRcq5LXDUIasQSWUw.png" /></figure><p>One thing to keep in mind is that this will only affect service to service traffic that is meshed within linkerd. Traffic outside of this scope will not be affected. So if you really want to cut down on this cost, considerations on where your databases live and how your pods are replicated across zones and how many pods per zone, as well how the load balancer ingest traffic into your platform, must be taken into consideration.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5zqAMqsM2Mj8y7Q_ePyyIQ.png" /><figcaption>Green Arrow: Traffic Affected by HAZL</figcaption></figure><h3>Results</h3><p>When rolling out our services to HAZL we saw immediate changes, this is our graph of remote v/s local traffic for a particular service.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gBJJdm8AEUIHjtMJ-t3PiQ.png" /><figcaption>Traffic by destination</figcaption></figure><p>If we group by zones we can actually see the exact time HAZL was activated and how the traffic from the same zone spikes up while the cross zone traffic trends down.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6qYRSaeVsAoVGcSnOsKx6A.png" /><figcaption>Traffic By Zone</figcaption></figure><h3>Don’t Sacrifice performance</h3><p>Even though the rollout was smooth with no incidents, our teams flagged increased latency on some endpoints that were enabled with HAZL. Since loveholidays.com is the fastest package holiday provider in Europe, every millisecond counts!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*d_YGEdmWSIp6xuJHfuslnA.png" /></figure><p>To analyze these things it’s always better to use a heatmap, in this case by zone at around the time of deployment to debug accordingly. In this case the issue was manifesting itself in the relationship between our service <strong>aurora</strong>-&gt;<strong>content-repository-api</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Pw_3Wb4WmP02iBxVTNh2lQ.png" /></figure><p>Aurora is our backend for frontend application, that talks downstream to one of our content services api.</p><p>So from here it gets a bit confusing, traffic from aurora[B] -&gt; content-repository-api[B] is the cause of all increased latency. Huh? Isn’t the opposite supposed to happen? traffic within the same zone should be lower latency, not higher, what the heck? So in order to understand what is happening it’s always a good idea to draw the current state of the system. So this is the relationship graphed by pods.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XPp3e2N9zJVT6OZDdGCkAw.png" /></figure><p>Hmm… this is a bit odd, judging from the heatmap, C-&gt;A Traffic is faster than B-&gt;B? . Now this is where the junior is separated from the senior. What do you do? Where do you check next? Can you find the answer? A common approach I take (as you can probably guess) is, more diagrams.</p><p>Probing in different dimensions of the platform and analyzing correctly will come with experience. My initial thought was, let’s take a look at the databases and diagram those as well so we can get a better picture. This is what I came up with after adding the database/storage components to it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PHlJXvbtqd7PALc5FB1pEg.png" /></figure><p>Ah…. the joy of platform engineering. So we’ve just discovered that our content-repository-api has a rapid bucket on zone-b and a database on zone-c. There was a proverbial tick in my eye when I first saw this, which almost caused a stroke, but I carried on.</p><p>What we have discovered is that we are optimized for cost spending! Every route aurora takes to communicate with content-repository-api will be cross zone traffic! I’m sure Google Cloud is happy, but we are not. Although this is an interesting find, it doesn’t answer the question we had initially, since the database is in zone-c we would expect A-C traffic to also be high in latency but it isn’t, only B-&gt;B traffic. But goes to show you when you diagram and understand the system, you’ll uncover new and exciting legacy consequences of decisions that were made years ago. This got flagged and we fixed it by modifying the affinity of pods to zone-c as well as the rapid bucket to zone-c, but still, didn’t explain the reason why the latency went up.</p><p>Another thing that came up to me was, the pod count, if we had too little services on zone-b then maybe that was the cause? with HAZL enabled we are now forcing traffic to a specific zone after all.</p><p>Let’s look first at the pod distribution for each service.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JlbDHNBXXBhy3LjPdJxN1A.png" /></figure><p>So aurora on zone-b has 41 pods hitting one pod of content-repository-api, bingo! . That would explain the latency. We are DDOSing our lonely pod in zone-b.</p><p>I won’t go into details as to how we got into this state, but the TLDR is that we move / upgrade our infrastructure all the time and sometimes some machine types are available on some zones and not others, as well as we simply didn’t take cross zone traffic as something to think about when designing our solutions, because the cost was always so low. But as you scale bad decisions tend to multiply by orders of magnitude.</p><p>Because of this, we tend to end up with inconsistent deployment of services like this, that no matter the route, we will always pay cross zone traffic costs.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*N7yzBUADxzGAUuhPTYxn8A.png" /></figure><p>Whereas a more cost efficient solution when controlling for cross zone traffic costs would look more like this.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ChSCyWqtIJH5syQZlebQ2Q.png" /></figure><p>Here at least, within zone-A, you would have savings relative to the other configuration.</p><h4>Conclusion</h4><p>So this was a fun exercise, but not a silver bullet. Although we are seeing massive savings there are limitations and tradeoffs we have found that prevent us from rolling this out cluster wide, as well as new features that we would like to see to further enhance cost savings on cross zone traffic.</p><ul><li>In services which have a database attached: There is a tradeoff in cost in how you distribute pods v/s how many replicas you have of the database in each zone. It is more costly to have read replicas in each zone than having one instance of the db in a zone and paying for cross zone traffic. This is a decision you, as a platform engineer, must weigh up if reliability is more important or cost savings and adjust accordingly.</li><li>For best performance, you must have pods evenly distributed by zone: This requires some tuning as you may not have replica count to match the zones your platform is deploying to, if you have massive pods with low replicas, it’s a good approach to reduce resource on each pod and allow for more pods to be spread out across the cluster, carefully monitoring how many source / destination pods you have.</li><li>One important limitation of HAZL is that it is an all or nothing enablement of the configuration. If you have one pod that talks to 3 other services / endpoints, then they all share the same load band configuration. Which means that you can’t roll out HAZL for a single service to service connection, but rather, all of the connections the client pod has. Being able to be more precise and have HAZL be rolled out exclusively on A-&gt;B but not A-&gt;C, would allow for smoother rollouts on critical systems instead of the YOLO approach that it currently presents us with.</li><li>Another feature that is not present, but would be quite useful, is the ability of HAZL to prefer same node traffic, this isn’t configurable at the moment but having different hierarchies of preference to where to send traffic to would be a great addition to HAZL.</li><li>Finally, there are times (roughly 66%) of the time when you have to go cross zone, having the ability for linkerd to compress traffic using something like <strong>zstd</strong> would be incredibly cost efficient when you have to take the hit on cross zone traffic.</li></ul><p>After rolling out HAZL and performing our due diligence to better understand where our infrastructure is deployed, we managed to come from a peak of £620 on June 1st to £280-£290 on June 17th for a massive 55% decrease in this SKU, leading to savings of £120,000 per year!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6OEYQZl1t3IV7akL_Bl5hw.png" /></figure><p>This wasn’t something we paid attention to as part of our architecting and design phases of new services, we are now rolling out new ways to catch and mitigate bad configuration causing cross zone traffic. It’s near impossible to have this cost be 0 in multi region, multi zone clusters. But knowing the tradeoffs when making architectural decisions goes a long way in saving your budget.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f8f7d8af7320" width="1" height="1" alt=""><hr><p><a href="https://tech.loveholidays.com/linkerd-hazl-how-to-save-costs-on-cross-zone-traffic-without-sacrificing-performance-f8f7d8af7320">Linkerd HAZL: How to save costs on cross zone traffic without sacrificing performance.</a> was originally published in <a href="https://tech.loveholidays.com/">loveholidays tech</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Past and Future of AI in Tech and Business]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://danielftapiar.medium.com/the-past-and-future-of-ai-in-tech-and-business-2d7495487da2?source=rss-de2914d425d2------2"><img src="https://cdn-images-1.medium.com/max/1049/1*JJ8_O5n3OkdbRzNpNCdGAg.png" width="1049"></a></p><p class="medium-feed-snippet">Over the past few years AI has taken over the world not just in tech but in business, but often times I see people all over the internet&#x2026;</p><p class="medium-feed-link"><a href="https://danielftapiar.medium.com/the-past-and-future-of-ai-in-tech-and-business-2d7495487da2?source=rss-de2914d425d2------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://danielftapiar.medium.com/the-past-and-future-of-ai-in-tech-and-business-2d7495487da2?source=rss-de2914d425d2------2</link>
            <guid isPermaLink="false">https://medium.com/p/2d7495487da2</guid>
            <category><![CDATA[business]]></category>
            <category><![CDATA[future-technology]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Daniel Tapia]]></dc:creator>
            <pubDate>Mon, 19 Jan 2026 19:21:27 GMT</pubDate>
            <atom:updated>2026-01-19T19:21:27.786Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Why aren’t my pods distributing evenly? Debugging Load Testing]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://danielftapiar.medium.com/why-arent-my-pods-distributing-evenly-debugging-load-testing-0f23b7fd4d2e?source=rss-de2914d425d2------2"><img src="https://cdn-images-1.medium.com/max/2356/1*KsE3m_fz45WV1sWe7Ot0uw.png" width="2356"></a></p><p class="medium-feed-snippet">During one of my support weeks at my job, an interesting problem came across my desk. We have an nginx service that has over 50 pods to&#x2026;</p><p class="medium-feed-link"><a href="https://danielftapiar.medium.com/why-arent-my-pods-distributing-evenly-debugging-load-testing-0f23b7fd4d2e?source=rss-de2914d425d2------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://danielftapiar.medium.com/why-arent-my-pods-distributing-evenly-debugging-load-testing-0f23b7fd4d2e?source=rss-de2914d425d2------2</link>
            <guid isPermaLink="false">https://medium.com/p/0f23b7fd4d2e</guid>
            <category><![CDATA[load-testing]]></category>
            <category><![CDATA[kubernetes-networking]]></category>
            <category><![CDATA[http-request]]></category>
            <category><![CDATA[golang]]></category>
            <category><![CDATA[http-client]]></category>
            <dc:creator><![CDATA[Daniel Tapia]]></dc:creator>
            <pubDate>Thu, 08 Jan 2026 10:59:38 GMT</pubDate>
            <atom:updated>2026-01-08T10:59:38.502Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Another MCP Servers Tutorial… Internal Mechanics]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/binome/mcp-servers-internal-mechanics-37dba35daab0?source=rss-de2914d425d2------2"><img src="https://cdn-images-1.medium.com/max/700/0*nhw7wXou1eF-koGI.png" width="700"></a></p><p class="medium-feed-snippet">We in the Binome Team have been working on MCP servers recently and after spending a fair amount of time trying to figure out how they&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/binome/mcp-servers-internal-mechanics-37dba35daab0?source=rss-de2914d425d2------2">Continue reading on Binome »</a></p></div>]]></description>
            <link>https://medium.com/binome/mcp-servers-internal-mechanics-37dba35daab0?source=rss-de2914d425d2------2</link>
            <guid isPermaLink="false">https://medium.com/p/37dba35daab0</guid>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[agents]]></category>
            <category><![CDATA[mcp-server]]></category>
            <dc:creator><![CDATA[Daniel Tapia]]></dc:creator>
            <pubDate>Thu, 12 Jun 2025 16:14:58 GMT</pubDate>
            <atom:updated>2025-06-12T16:30:08.436Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Kong DB-less Gateway API with AWS ACM]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/binome/kong-db-less-gateway-api-with-aws-acm-aa764a6ab76c?source=rss-de2914d425d2------2"><img src="https://cdn-images-1.medium.com/max/1601/1*gO5bqjXwT15Axg6Sy4HpSw.png" width="1601"></a></p><p class="medium-feed-snippet">Background</p><p class="medium-feed-link"><a href="https://medium.com/binome/kong-db-less-gateway-api-with-aws-acm-aa764a6ab76c?source=rss-de2914d425d2------2">Continue reading on Binome »</a></p></div>]]></description>
            <link>https://medium.com/binome/kong-db-less-gateway-api-with-aws-acm-aa764a6ab76c?source=rss-de2914d425d2------2</link>
            <guid isPermaLink="false">https://medium.com/p/aa764a6ab76c</guid>
            <category><![CDATA[ingress]]></category>
            <category><![CDATA[gateway]]></category>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[kong]]></category>
            <category><![CDATA[networking]]></category>
            <dc:creator><![CDATA[Daniel Tapia]]></dc:creator>
            <pubDate>Tue, 28 May 2024 17:20:53 GMT</pubDate>
            <atom:updated>2024-05-28T17:20:53.274Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[RAG Enriched Prompt with SQL Table Data]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/binome/rag-enriched-prompt-with-sql-table-data-8baf62339113?source=rss-de2914d425d2------2"><img src="https://cdn-images-1.medium.com/max/2000/1*dddx8wlC_4AGTpUM4UpkBA.png" width="2000"></a></p><p class="medium-feed-link"><a href="https://medium.com/binome/rag-enriched-prompt-with-sql-table-data-8baf62339113?source=rss-de2914d425d2------2">Continue reading on Binome »</a></p></div>]]></description>
            <link>https://medium.com/binome/rag-enriched-prompt-with-sql-table-data-8baf62339113?source=rss-de2914d425d2------2</link>
            <guid isPermaLink="false">https://medium.com/p/8baf62339113</guid>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[rags]]></category>
            <category><![CDATA[langchain]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[bedrock]]></category>
            <dc:creator><![CDATA[Daniel Tapia]]></dc:creator>
            <pubDate>Fri, 19 Apr 2024 13:17:24 GMT</pubDate>
            <atom:updated>2024-04-19T15:22:37.308Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Getting Started: EKS Access Entries]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/binome/getting-started-eks-access-entries-6ae5eec82544?source=rss-de2914d425d2------2"><img src="https://cdn-images-1.medium.com/max/1235/1*r5AZcdgJFKz3QqFUE8UP-w.png" width="1235"></a></p><p class="medium-feed-snippet">You may have worked with EKS in the past and one of the first infuriating aspects of setting up your cluster for other users, is the&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/binome/getting-started-eks-access-entries-6ae5eec82544?source=rss-de2914d425d2------2">Continue reading on Binome »</a></p></div>]]></description>
            <link>https://medium.com/binome/getting-started-eks-access-entries-6ae5eec82544?source=rss-de2914d425d2------2</link>
            <guid isPermaLink="false">https://medium.com/p/6ae5eec82544</guid>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[ek]]></category>
            <category><![CDATA[rbac]]></category>
            <dc:creator><![CDATA[Daniel Tapia]]></dc:creator>
            <pubDate>Thu, 07 Mar 2024 11:45:01 GMT</pubDate>
            <atom:updated>2024-03-07T11:45:01.708Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[System Design:  Migration of Queues with no Downtime!]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://danielftapiar.medium.com/system-design-migration-of-queues-with-no-downtime-d0ed2922e94a?source=rss-de2914d425d2------2"><img src="https://cdn-images-1.medium.com/max/1658/1*9Z0p43yZh_exziUat-bDdA.png" width="1658"></a></p><p class="medium-feed-snippet">Designing a Migration Plan of Broker Queues with Minimal Downtime.</p><p class="medium-feed-link"><a href="https://danielftapiar.medium.com/system-design-migration-of-queues-with-no-downtime-d0ed2922e94a?source=rss-de2914d425d2------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://danielftapiar.medium.com/system-design-migration-of-queues-with-no-downtime-d0ed2922e94a?source=rss-de2914d425d2------2</link>
            <guid isPermaLink="false">https://medium.com/p/d0ed2922e94a</guid>
            <category><![CDATA[system-design-concepts]]></category>
            <category><![CDATA[migration]]></category>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[infrastructure-services]]></category>
            <dc:creator><![CDATA[Daniel Tapia]]></dc:creator>
            <pubDate>Thu, 25 Jan 2024 18:02:45 GMT</pubDate>
            <atom:updated>2024-01-25T18:02:45.334Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[ReInvent 2023: Trying out Pod Identity Agent]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/binome/reinvent-2023-trying-out-pod-identity-agent-de823f6b1178?source=rss-de2914d425d2------2"><img src="https://cdn-images-1.medium.com/max/600/1*A0N1D-jF-kwsmUTIaMWUtg.png" width="600"></a></p><p class="medium-feed-snippet">Out of re:Invent 2023 one of the features that caught my eye is the new EKS add on: Pod Identity Agent that promises to simplify IAM&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/binome/reinvent-2023-trying-out-pod-identity-agent-de823f6b1178?source=rss-de2914d425d2------2">Continue reading on Binome »</a></p></div>]]></description>
            <link>https://medium.com/binome/reinvent-2023-trying-out-pod-identity-agent-de823f6b1178?source=rss-de2914d425d2------2</link>
            <guid isPermaLink="false">https://medium.com/p/de823f6b1178</guid>
            <category><![CDATA[iam-roles]]></category>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[aws]]></category>
            <dc:creator><![CDATA[Daniel Tapia]]></dc:creator>
            <pubDate>Tue, 28 Nov 2023 21:08:03 GMT</pubDate>
            <atom:updated>2023-11-28T21:09:43.447Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Recovering Files  Removed With rm]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/binome/recovering-files-removed-with-rm-6eb4540bba0f?source=rss-de2914d425d2------2"><img src="https://cdn-images-1.medium.com/max/1000/1*qUNrAXMwWImRhKDKo3Uh2g.jpeg" width="1000"></a></p><p class="medium-feed-snippet">How to recover from a nightmare</p><p class="medium-feed-link"><a href="https://medium.com/binome/recovering-files-removed-with-rm-6eb4540bba0f?source=rss-de2914d425d2------2">Continue reading on Binome »</a></p></div>]]></description>
            <link>https://medium.com/binome/recovering-files-removed-with-rm-6eb4540bba0f?source=rss-de2914d425d2------2</link>
            <guid isPermaLink="false">https://medium.com/p/6eb4540bba0f</guid>
            <category><![CDATA[recovery-data-ssd]]></category>
            <category><![CDATA[forensics]]></category>
            <category><![CDATA[disk]]></category>
            <category><![CDATA[filesystem]]></category>
            <dc:creator><![CDATA[Daniel Tapia]]></dc:creator>
            <pubDate>Tue, 25 Jul 2023 23:07:43 GMT</pubDate>
            <atom:updated>2023-07-25T23:07:43.638Z</atom:updated>
        </item>
    </channel>
</rss>