<?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 Alex Gkiouros on Medium]]></title>
        <description><![CDATA[Stories by Alex Gkiouros on Medium]]></description>
        <link>https://medium.com/@voovode?source=rss-6c3f3bbb6862------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*6H9xbYJk1Z6bz7wbdHcXUA.png</url>
            <title>Stories by Alex Gkiouros on Medium</title>
            <link>https://medium.com/@voovode?source=rss-6c3f3bbb6862------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 06:44:31 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@voovode/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[Forget the Agent Platform. The real story at Next ’26 was GKE]]></title>
            <link>https://medium.com/google-cloud/forget-the-agent-platform-the-real-story-at-next-26-was-gke-1c0d7fbfd142?source=rss-6c3f3bbb6862------2</link>
            <guid isPermaLink="false">https://medium.com/p/1c0d7fbfd142</guid>
            <category><![CDATA[data-lakehouse]]></category>
            <category><![CDATA[kubernetes]]></category>
            <category><![CDATA[agent-platform]]></category>
            <category><![CDATA[gke]]></category>
            <category><![CDATA[gemini-enterprise]]></category>
            <dc:creator><![CDATA[Alex Gkiouros]]></dc:creator>
            <pubDate>Sat, 25 Apr 2026 16:53:40 GMT</pubDate>
            <atom:updated>2026-04-27T09:02:13.224Z</atom:updated>
            <content:encoded><![CDATA[<p>Skipped Google Next at Vegas this year? These are the announcements I think will keep landing in customer conversations over the next quarter.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BOYgibVdjZM31EyIdmq4Ww.png" /><figcaption>Mandalay Bay — Next at Night</figcaption></figure><h3>Gemini Enterprise Agent Platform</h3><p>The headline of the keynote, and probably rightly so. The <a href="https://docs.cloud.google.com/gemini-enterprise-agent-platform/overview">Gemini Enterprise Agent Platform</a> is a full stack to build, run, and govern agents end to end: <a href="https://docs.cloud.google.com/gemini-enterprise-agent-platform/agent-studio/overview">Agent Studio</a> (low-code, natural language), Agent Designer (no-code, trigger-based, inside the Gemini Enterprise app), Agent-to-Agent Orchestration, and the <a href="https://docs.cloud.google.com/gemini-enterprise-agent-platform/govern">governance plane</a> around it (Registry, Identity, Gateway, Observability). There’s also an <a href="https://cloud.google.com/blog/products/ai-machine-learning/the-new-gemini-enterprise-one-platform-for-agent-development#:~:text=controls%20built%2Din.-,Inbox%20in%20Gemini%20Enterprise,-provides%20a%20central">Agent Inbox</a> so a human can actually see what their agents are doing, and long-running agents that operate autonomously inside secure cloud sandboxes.</p><p>If you’ve used Vertex AI Agent Builder, this is what it grows into when you add real enterprise governance. The Agent Gateway partner list (Cisco, Palo Alto, CrowdStrike, Okta, Ping, Zscaler, F5, and more) is the part that gets it past most CISOs.</p><h3>TPU 8t and TPU 8i</h3><p>Google split the eighth generation into two: a <a href="https://cloud.google.com/blog/products/compute/tpu-8t-and-tpu-8i-technical-deep-dive">training chip (8t) and an inference chip (8i)</a>. The 8i claims 80% better performance per dollar on inference. That is the FinOps number to model in your TCO sheet, especially if you’re running a serving fleet.</p><p>A5X instances also bring NVIDIA Vera Rubin NVL72 to GCP later this year. So both bets are covered: in-house silicon and the latest NVIDIA.</p><h3>Virgo Network</h3><p>The new <a href="https://cloud.google.com/blog/products/networking/introducing-virgo-megascale-data-center-fabric">megascale data center fabric</a>. High-radix switches, big bisection bandwidth, deterministic low latency. The point is keeping accelerators fed for distributed training and serving. Pair it with <a href="https://cloud.google.com/managed-lustre/docs/overview">Managed Lustre</a> at 10TB/s and Cloud Storage Rapid and the “I/O is bottlenecking my training” excuse becomes harder to defend.</p><h3>GKE: this is where it got interesting for me</h3><p>This was the part of the event that surprised me. The headline page on blog.google barely mentions GKE, but <a href="https://cloud.google.com/blog/products/containers-kubernetes/whats-new-in-gke-at-next26">the actual breakouts</a> were dense with announcements. Some context Google shared: 66% of organisations now run Kubernetes for generative AI workloads, multi-agent workflows are up 327% in a few months, and GKE powers AI for all of Google’s top 50 customers on the platform.</p><p><a href="https://cloud.google.com/kubernetes-engine/docs/concepts/hypercluster-overview"><strong>GKE Hypercluster</strong></a> (private GA). One conformant control plane managing up to 1 million chips across 256,000 nodes, spanning regions. The interesting bit is <strong>Titanium Intelligence Enclave</strong>, a hardware-rooted “no-admin-access” security engine where weights and prompts stay sealed even from platform admins. Clearly aimed at frontier builders, but the design choices are worth understanding even if you’re nowhere near that scale.</p><p><a href="https://cloud.google.com/kubernetes-engine/docs/concepts/machine-learning/agent-sandbox"><strong>GKE Agent Sandbox</strong></a><strong>. </strong>gVisor-isolated sandboxes for agents to run untrusted code and tool calls. Up to 300 sandboxes per second per cluster, sub-second time to first instruction. On Axion N4A it claims 30% better price-performance than the next leading hyperscaler. The only native agent sandbox among hyperscalers at the moment. I wrote about the N4A price-performance <a href="https://medium.com/doit-international/first-look-at-google-cloud-n4a-vms-benchmarked-against-n4-c4a-and-aws-m8g-aba8017ea927">here</a>.</p><p><a href="https://cloud.google.com/kubernetes-engine/docs/concepts/about-gke-inference-gateway"><strong>Inference Gateway</strong></a>. Capacity-aware routing and KV Cache management to cut TTFT (Time to First Token). Less hand-tuning to land on a decent inference baseline.</p><p><a href="https://cloud.google.com/blog/products/containers-kubernetes/whats-new-in-gke-at-next26"><strong>RL primitives</strong></a>. RL Scheduler (kills the straggler effect between sampling, reward, and training steps), RL Sandbox (kernel-level isolation, millisecond provisioning), RL Observability and Reliability dashboards. Anyone who has tried to run reinforcement learning on Kubernetes knows how miserable that path was. This looks like Google taking it seriously.</p><p><a href="https://cloud.google.com/kubernetes-engine/docs/how-to/expose-custom-metrics-autoscaling"><strong>Intent-based autoscaling on custom metrics</strong></a>. Scale on signals beyond CPU and memory. Honestly, should have been there years ago.</p><p><a href="https://cloud.google.com/blog/topics/google-cloud-next/google-cloud-next-2026-wrap-up/"><strong>Ambient networking</strong></a>. Integrated data plane for GKE and Cloud Run. Service discovery, zero-trust, traffic management, no sidecars. If your inference latency or platform bill has been hurt by Istio sidecars, this is the one.</p><p>GKE continues to be positioned as the platform for AI workloads. Whether you buy “Kubernetes as the OS of the AI era” as a marketing line or not, the technical substance is doing most of the work.</p><h3>Agentic Data Cloud</h3><p><a href="https://cloud.google.com/dataplex/docs/introduction"><strong>Knowledge Catalog</strong></a> (formerly Dataplex Universal Catalog, renamed in April). Gemini autonomously tags and links entities across your enterprise data so agents pick up your business context and naming conventions. Agent quality is mostly a context problem, and this is one of the better attacks on it I’ve seen.</p><p><a href="https://cloud.google.com/products/lakehouse"><strong>Cross-Cloud Lakehouse</strong></a> (BigLake is now Google Cloud Lakehouse). Iceberg-standardised, Cross-Cloud Interconnect into the data plane, bi-directional federation in preview. Query data sitting in AWS without moving it. Spanner Omni and Lakehouse federation for AlloyDB are also in preview.</p><p>Worth flagging: Lightning Engine for Apache Spark at claimed 2x price-performance, BigQuery fluid scaling at up to 34% lower cost for autoscaling workloads, and an in-memory tier for Bigtable.</p><h3>Agentic Defense (Google &amp; Wiz)</h3><p>Now that the <a href="https://cloud.google.com/wiz">Wiz acquisition has closed</a>, the integration ships. Threat Hunting Agent (proactive hunting, writes detection rules autonomously), Detection Engineering Agent, Dark Web Intelligence (98% accuracy claim on millions of daily events), Third-Party Context Agent. Wiz coverage now extends to Databricks and more AI studios.</p><p>The framing Kurian used in the press session was AI red, blue, and green teams (attack, defend, fix) all running as agents. That mental model lands better than the product-by-product list.</p><h3>Workspace Intelligence</h3><p><a href="https://workspace.google.com/blog/product-announcements/introducing-workspace-intelligence">Ask Gemini in Chat</a> now synthesises across Docs, Drive, Meet, and Gmail and takes actions in place. Schedule a meeting, draft a brief, all from the chat surface. For end-user adoption inside enterprises, this is probably the most consequential announcement of the week.</p><h3>What I’d actually do with this on Monday</h3><p>If you’ve been waiting for a reason to consolidate fragmented GKE clusters, Hypercluster gives you one. Even if you don’t sign up for private GA, the design is the new reference.</p><p>If you’re running agent workloads outside GKE Agent Sandbox, run the price-performance numbers. The gVisor-on-Axion combination is hard to ignore at 30% better than the next hyperscaler.</p><p>If you’re doing RL training on GKE today, the scheduler and sandbox primitives change the cost picture meaningfully.</p><p>If you have BigQuery autoscaling workloads, audit them this quarter. 34% on average is not bad at all.</p><p>The one I’m most cautious about: a single GKE control plane managing a million chips across regions sounds wonderful until you think through blast radius and change management. Private GA is the right place for it. Let’s validate before we bet a pipeline on it.</p><h3>Bonus: DESIGN.md</h3><p>Not strictly a Next ’26 announcement, but Google Labs published this in the same window and it deserves a mention if you build anything agentic that touches a UI — who doesn’t these days?</p><p><a href="https://github.com/google-labs-code/design.md">DESIGN.md</a> is a format spec for describing a visual identity to coding agents. YAML front matter for the machine-readable tokens (colors, typography, spacing, components), markdown prose underneath for the rationale. The agent gets exact values plus the <em>why</em>, all in a single file your designers and developers already know how to read and review.</p><p>It ships with a CLI for linting (broken token references, WCAG contrast checks, section ordering), diffing two versions to catch regressions, and exporting to Tailwind theme config or W3C DTCG tokens. Currently at alpha, Apache 2.0, already sitting at 8k+ stars.</p><p>If you’ve watched coding agents flail at “match our brand” instructions, this is your missing context layer.</p><p>Found this useful? Connect with me on <a href="https://www.linkedin.com/in/alexgkiouros/">LinkedIn</a> for more insights, architecture, AI, and practical tech strategies. I regularly share ideas, lessons learned, and real-world solutions built from the front lines.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1c0d7fbfd142" width="1" height="1" alt=""><hr><p><a href="https://medium.com/google-cloud/forget-the-agent-platform-the-real-story-at-next-26-was-gke-1c0d7fbfd142">Forget the Agent Platform. The real story at Next ’26 was GKE</a> was originally published in <a href="https://medium.com/google-cloud">Google Cloud - Community</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Your GCP load balancer has been lying about capacity and IN_FLIGHT mode fixes that]]></title>
            <link>https://medium.com/google-cloud/your-gcp-load-balancer-has-been-lying-about-capacity-and-in-flight-mode-fixes-that-58929dac3f00?source=rss-6c3f3bbb6862------2</link>
            <guid isPermaLink="false">https://medium.com/p/58929dac3f00</guid>
            <category><![CDATA[websocket]]></category>
            <category><![CDATA[server-sent-events]]></category>
            <category><![CDATA[google-cloud-platform]]></category>
            <category><![CDATA[load-balancer]]></category>
            <category><![CDATA[infrastructure]]></category>
            <dc:creator><![CDATA[Alex Gkiouros]]></dc:creator>
            <pubDate>Fri, 20 Mar 2026 06:00:08 GMT</pubDate>
            <atom:updated>2026-03-20T06:00:08.971Z</atom:updated>
            <content:encoded><![CDATA[<p>If you’re running WebSocket connections, streaming APIs, or server-sent events behind a Google Cloud Application Load Balancer, you’ve probably been using RATE as your balancing mode; It counts new HTTP requests per second and spreads traffic accordingly.</p><p>Here’s the problem: RATE has no idea how long those requests take.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HSgnEFtfgDx1QKFThoCW2Q.jpeg" /></figure><p>A request that completes in 50ms and a request that holds a connection open for 30 seconds both count as “1” to the RATE balancer. The moment the request arrives, it’s counted, and then forgotten. This means a backend drowning in slow, concurrent requests can look identical in capacity terms to one that’s idle.</p><p>Google Cloud has shipped a fix for this in Preview: the <strong>IN_FLIGHT balancing mode</strong>, paired with a new <strong>traffic duration</strong> setting on backend services.</p><h3>What IN_FLIGHT does</h3><p>Instead of counting the <em>rate</em> of new requests, IN_FLIGHT counts the number of <em>concurrent, in-progress</em> HTTP requests against each backend zone. A request is counted when it arrives and uncounted when the response completes. This gives the load balancer a real-time view of actual backend load rather than a snapshot of arrival rate.</p><p>You configure it with one of three target capacity parameters:</p><ul><li>max-in-flight-requests target concurrent requests for the entire zone</li><li>max-in-flight-requests-per-instance target per VM, scaled across the zone</li><li>max-in-flight-requests-per-endpoint target per NEG endpoint</li></ul><p>When a zone hits its target, the load balancer directs new requests to another zone (or region, for global load balancers). When <em>all</em> zones hit target, it overfills proportionally, same spillover behaviour as RATE.</p><h3>The Traffic Duration setting</h3><p>IN_FLIGHT is gated behind a new per-backend attribute called trafficDuration, which accepts two values:</p><ul><li><strong>SHORT</strong> (default): optimised for requests completing in under one second. This is the implicit setting if you don&#39;t specify anything, and it unlocks the existing RATE and UTILIZATION modes.</li><li><strong>LONG</strong>: for requests where backends need more than one second to generate responses. This is what unlocks IN_FLIGHT.</li></ul><p>Setting trafficDuration to LONG also changes which balancing modes are available. You lose RATE and CONNECTION, but gain IN_FLIGHT alongside UTILIZATION and CUSTOM_METRICS.</p><h3>When do you need this</h3><p><strong>WebSocket backends.</strong> A WebSocket connection that stays open for minutes registers as a single request to RATE, which counted it on arrival and moved on. With IN_FLIGHT, that connection continues to occupy capacity until it closes. This prevents the load balancer from piling hundreds of long-lived connections onto a single zone while others sit idle.</p><p><strong>Streaming and SSE.</strong> Server-sent event streams, long-polling, or chunked transfer responses all suffer from the same blind spot. The request is in-progress for the duration of the stream, but RATE only sees the initial arrival.</p><p><strong>File uploads/downloads through the ALB.</strong> Large file transfers that take multiple seconds to complete will be accurately represented as ongoing load.</p><p><strong>What about AI/ML inference?</strong> You might think IN_FLIGHT is a natural fit for LLM serving, where requests can take seconds to complete. It would be <em>better</em> than RATE, but it&#39;s still a blunt instrument for this workload. If you&#39;re serving models on GKE, look at the <a href="https://docs.cloud.google.com/kubernetes-engine/docs/concepts/about-gke-inference-gateway">GKE Inference Gateway</a> instead. It routes based on KV cache utilisation, queue depth, and request cost, which are the signals that actually matter for inference. For non-GKE setups, the CUSTOM_METRICS balancing mode lets you define application-specific signals. IN_FLIGHT is a better default than RATE for slow requests, but purpose-built tooling exists for inference specifically.</p><h3>How to configure it</h3><p>Create a backend service and add a backend with LONG traffic duration and IN_FLIGHT balancing mode:</p><pre>gcloud compute backend-services create my-backend-service \<br>    --protocol=HTTP \<br>    --health-checks=my-health-check \<br>    --global</pre><pre>gcloud compute backend-services add-backend my-backend-service \<br>    --instance-group=my-instance-group \<br>    --instance-group-zone=europe-west2-a \<br>    --balancing-mode=IN_FLIGHT \<br>    --traffic-duration=LONG \<br>    --max-in-flight-requests-per-instance=100 \<br>    --global</pre><p>IN_FLIGHT works with all Application Load Balancers <em>except</em> the classic ALB. That means global external, regional external, cross-region internal, and regional internal ALBs are all supported, as well as Cloud Service Mesh.</p><p>Supported backend types include instance groups (zonal and regional managed), zonal NEGs with GCE_VM_IP_PORT endpoints, and hybrid connectivity NEGs. Serverless NEGs, internet NEGs, and Private Service Connect NEGs don&#39;t support balancing mode parameters at all, so they&#39;re unaffected.</p><h3>TL;DR</h3><p>If your requests complete in milliseconds, RATE is fine. But the moment your workloads involve persistent connections, streaming, or anything that holds backend resources for more than a second, IN_FLIGHT gives the load balancer the signal it&#39;s been missing: <em>how busy are my backends right now</em>, not just <em>how fast are new requests arriving</em>.</p><p>It’s a small configuration change that closes a long-standing gap in how GCP load balancers understand capacity for modern, long-lived workloads.</p><p>Want to ask a question or share a thought with me?<br>Let’s connect on <a href="https://www.linkedin.com/in/alexgkiouros/">LinkedIn</a>!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=58929dac3f00" width="1" height="1" alt=""><hr><p><a href="https://medium.com/google-cloud/your-gcp-load-balancer-has-been-lying-about-capacity-and-in-flight-mode-fixes-that-58929dac3f00">Your GCP load balancer has been lying about capacity and IN_FLIGHT mode fixes that</a> was originally published in <a href="https://medium.com/google-cloud">Google Cloud - Community</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[First look at Google Cloud N4A VMs: benchmarked against N4, C4A and AWS M8g.]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/doit-international/first-look-at-google-cloud-n4a-vms-benchmarked-against-n4-c4a-and-aws-m8g-aba8017ea927?source=rss-6c3f3bbb6862------2"><img src="https://cdn-images-1.medium.com/max/1792/1*I561Lz9sosdGpsgP4c5KRg.png" width="1792"></a></p><p class="medium-feed-snippet">I&#x2019;ve tested the newest N4A instance family offering from Google Cloud so you don&#x2019;t have to.</p><p class="medium-feed-link"><a href="https://medium.com/doit-international/first-look-at-google-cloud-n4a-vms-benchmarked-against-n4-c4a-and-aws-m8g-aba8017ea927?source=rss-6c3f3bbb6862------2">Continue reading on DoiT »</a></p></div>]]></description>
            <link>https://medium.com/doit-international/first-look-at-google-cloud-n4a-vms-benchmarked-against-n4-c4a-and-aws-m8g-aba8017ea927?source=rss-6c3f3bbb6862------2</link>
            <guid isPermaLink="false">https://medium.com/p/aba8017ea927</guid>
            <category><![CDATA[google-cloud-platform]]></category>
            <category><![CDATA[arm]]></category>
            <category><![CDATA[axion]]></category>
            <category><![CDATA[cost-optimization]]></category>
            <category><![CDATA[benchmark]]></category>
            <dc:creator><![CDATA[Alex Gkiouros]]></dc:creator>
            <pubDate>Wed, 28 Jan 2026 20:50:25 GMT</pubDate>
            <atom:updated>2026-01-29T15:06:07.070Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Understanding Google Cloud's New Spend-Based CUD Model]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/doit-international/understanding-google-clouds-new-spend-based-cud-model-311dea4a4a79?source=rss-6c3f3bbb6862------2"><img src="https://cdn-images-1.medium.com/max/1000/1*C9D5afR9cGFpQT210jUiIw.png" width="1000"></a></p><p class="medium-feed-snippet">January 2026 brings simplified billing, expanded discounts, same savings</p><p class="medium-feed-link"><a href="https://medium.com/doit-international/understanding-google-clouds-new-spend-based-cud-model-311dea4a4a79?source=rss-6c3f3bbb6862------2">Continue reading on DoiT »</a></p></div>]]></description>
            <link>https://medium.com/doit-international/understanding-google-clouds-new-spend-based-cud-model-311dea4a4a79?source=rss-6c3f3bbb6862------2</link>
            <guid isPermaLink="false">https://medium.com/p/311dea4a4a79</guid>
            <category><![CDATA[committed-use-discounts]]></category>
            <category><![CDATA[google-cloud-platform]]></category>
            <category><![CDATA[finops]]></category>
            <category><![CDATA[google-cloud-billing]]></category>
            <dc:creator><![CDATA[Alex Gkiouros]]></dc:creator>
            <pubDate>Thu, 22 Jan 2026 17:36:24 GMT</pubDate>
            <atom:updated>2026-01-23T11:22:07.656Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Google Cloud Assist Investigations: AI-Powered Root Cause Analysis]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/google-cloud/google-cloud-assist-investigations-ai-powered-root-cause-analysis-f173ceabaef4?source=rss-6c3f3bbb6862------2"><img src="https://cdn-images-1.medium.com/max/1024/1*EniHaelfQf9aN_dhoYqNiQ.jpeg" width="1024"></a></p><p class="medium-feed-snippet">When your Cloud Run service starts throwing 5xx errors at 3 AM, or your GKE cluster suddenly experiences latency spikes, traditional&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/google-cloud/google-cloud-assist-investigations-ai-powered-root-cause-analysis-f173ceabaef4?source=rss-6c3f3bbb6862------2">Continue reading on Google Cloud - Community »</a></p></div>]]></description>
            <link>https://medium.com/google-cloud/google-cloud-assist-investigations-ai-powered-root-cause-analysis-f173ceabaef4?source=rss-6c3f3bbb6862------2</link>
            <guid isPermaLink="false">https://medium.com/p/f173ceabaef4</guid>
            <category><![CDATA[google-gemini]]></category>
            <category><![CDATA[gemini-cloud-assist]]></category>
            <category><![CDATA[google-cloud-run]]></category>
            <category><![CDATA[google-cloud-platform]]></category>
            <category><![CDATA[gcp-security-operations]]></category>
            <dc:creator><![CDATA[Alex Gkiouros]]></dc:creator>
            <pubDate>Thu, 30 Oct 2025 02:09:49 GMT</pubDate>
            <atom:updated>2025-10-30T02:09:49.403Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Monitoring Google Cloud Armor Traffic: A Comprehensive Dashboard]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/doit-international/monitoring-google-cloud-armor-traffic-a-comprehensive-dashboard-a36fb42d3620?source=rss-6c3f3bbb6862------2"><img src="https://cdn-images-1.medium.com/max/2600/1*3HCKoMEJ2f6v19jHAck9YA.jpeg" width="6500"></a></p><p class="medium-feed-snippet">We&#x2019;ll be creating a Cloud Monitoring dashboard to better understand and monitor incoming traffic and attacks evaluated by Cloud Armor.</p><p class="medium-feed-link"><a href="https://medium.com/doit-international/monitoring-google-cloud-armor-traffic-a-comprehensive-dashboard-a36fb42d3620?source=rss-6c3f3bbb6862------2">Continue reading on DoiT »</a></p></div>]]></description>
            <link>https://medium.com/doit-international/monitoring-google-cloud-armor-traffic-a-comprehensive-dashboard-a36fb42d3620?source=rss-6c3f3bbb6862------2</link>
            <guid isPermaLink="false">https://medium.com/p/a36fb42d3620</guid>
            <category><![CDATA[ddos-attack]]></category>
            <category><![CDATA[google-cloud-platform]]></category>
            <category><![CDATA[google-cloud-armor]]></category>
            <category><![CDATA[adaptive-protection]]></category>
            <category><![CDATA[dashboard]]></category>
            <dc:creator><![CDATA[Alex Gkiouros]]></dc:creator>
            <pubDate>Mon, 19 Aug 2024 09:03:38 GMT</pubDate>
            <atom:updated>2024-10-18T13:53:20.394Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Identifying GCP’s Hidden Network Inter-Zone Egress Costs]]></title>
            <link>https://medium.com/doit-international/identifying-gcps-hidden-network-inter-zone-egress-costs-e1acc11ac03c?source=rss-6c3f3bbb6862------2</link>
            <guid isPermaLink="false">https://medium.com/p/e1acc11ac03c</guid>
            <category><![CDATA[google]]></category>
            <category><![CDATA[bigquery]]></category>
            <category><![CDATA[cloud-cost-management]]></category>
            <category><![CDATA[doit-international]]></category>
            <category><![CDATA[google-cloud-platform]]></category>
            <dc:creator><![CDATA[Alex Gkiouros]]></dc:creator>
            <pubDate>Tue, 12 Oct 2021 11:01:11 GMT</pubDate>
            <atom:updated>2024-05-28T08:16:49.131Z</atom:updated>
            <content:encoded><![CDATA[<h4>Learn how to identify your Inter-Zone Egress costs in a few easy steps, using commonly available methods.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*SJ-xG_cZGCLtpdORz8ndjA.jpeg" /></figure><p>Ever wondered where those Inter-Zone Egress costs are coming from? Found yourself looking at GCP’s <a href="https://cloud.google.com/vpc/network-pricing">network pricing</a> page many times to break it down? Me too. So I thought I might as well try to help clear things up.</p><h4>First, Flow Logs</h4><p>Let’s start with <a href="https://cloud.google.com/vpc/docs/using-flow-logs#enabling-vpc-flow-logs">enabling the VPC Flow Logs</a> for your project. This is the quickest and easiest way to see bytes flowing between the Virtual Machines (VMs) on your network without having any ‘dirty’ screen/tmux sessions with tools analyzing network traffic on the VMs.</p><p><em>Let’s wait for a few minutes so that the traffic can flow while we’re logging it now.</em></p><p>Using a filter on Cloud Logging like the below, we shall show some logs, assuming VPC Flow Logs are enabled:</p><pre>logName=”projects/<strong>PROJECT_NAME</strong>/logs/compute.googleapis.com%2Fvpc_flows&quot;</pre><p>It should look somewhat like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rDmRJWjyv51UkqVHP409zw.png" /><figcaption>VPC Flow Logs example</figcaption></figure><h4>The quick Spreadsheet way</h4><p>Instead of waiting for a full day in order to analyze this via BigQuery or another log sink, we can simply export a spreadsheet/CSV file. For some networks, this might be a fine way to get an initial idea of what’s going on.</p><blockquote><strong>If you want to query the logs using SQL, I recommend </strong><a href="https://cloud.google.com/logging/docs/export/configure_export_v2#creating_sink"><strong>creating a sink</strong></a><strong> and pushing the logs to a BigQuery dataset.</strong></blockquote><p>Now, since you’re able to see VPC flow logs in Cloud Logging, let’s <a href="https://cloud.google.com/logging/docs/view/logs-viewer-interface#download_logs">export a CSV file to a Drive</a> of the last 10,000 entries. This will essentially helps us see how many bytes are flowing and where.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*tGFTmWy3-gCT0IiiJWozCA.gif" /><figcaption>Downloading VPC Flow Logs</figcaption></figure><p>On that Google Spreadsheet export and after creating the pivot table with source/destination and bytes_sent, you might want to <a href="https://gist.github.com/loganvolkers/933af8513ed8c2268f59c85a31761a43">format bytes</a> into something more readable using the format below:</p><pre>[&lt;1000000]0.00,&quot; KB&quot;;[&lt;1000000000]0.00,,&quot; MB&quot;;0.00,,,&quot; GB&quot;</pre><p>Select the bytes column, head over to Format &gt; Number &gt; More Formats &gt; Custom number format, paste the above and click Apply.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2FiD1jHbpoPVG9sJXPBhag.png" /><figcaption>Custom number format</figcaption></figure><p>There you go. It’s as easy as that. You now have an initial picture of how your network activity looks. The sum column should have the bytes used tied with source~destination, easily identifiable.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xLtBqxcM3JeI7ojNnfmmWQ.jpeg" /><figcaption>Pivot table on Google Spreadsheets</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/220/1*PJKBp6KK2eyWmOwuFvibKw.gif" /></figure><h4><strong>The BigQuery way</strong></h4><p>Not able to see a pattern yet? Try a different time period, or wait a full day and try some queries against BigQuery. If you created a sink earlier to BigQuery, you now have a BQ dataset with a sample of logs and bytes recorded per instance.</p><p>An example query that can identify the sum of bytes sent between VMs could look like this:</p><pre>SELECT CONCAT(jsonPayload.src_instance.vm_name, &quot; to &quot;, jsonPayload.dest_instance.vm_name) AS vm_traffic,<br> sum(CAST (jsonPayload.bytes_sent AS INT64)) AS bytes_sent<br>FROM `<strong>PROJECT_NAME</strong>.<strong>DATASET_NAME</strong>.compute_googleapis_com_vpc_flows`<br>WHERE TIMESTAMP &gt;= &quot;2021-07-14 00:00:00 UTC&quot;<br>GROUP BY vm_traffic<br>ORDER BY bytes_sent</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3UodHrJIpu1F5Wwy6Ra4wA.png" /><figcaption>BQ bytes_sent example</figcaption></figure><p>How about now? Can you see what’s at the top of the list? :)</p><h4>How DoiT’s Reporting helps</h4><p>Google Spreadsheets and BigQuery are great tools for querying this data, but if you are a user of DoiT International’s cloud management platform, you have an even easier path to the it, complete with clear visualizations.</p><ol><li>Log in to DoiT’s <a href="https://console.doit.com/">Navigator</a>.</li><li>Under Analytics, click Reports and then Create new Report.</li><li>Under Metrics, keep ‘Cost’ (this can also be set to Usage if you want to see usage instead).</li><li>Group by ‘Project/Account ID’.</li><li>Filter by SKU, find and tick ‘Network Inter Zone Egress’:</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*q20pv81aF6NaO0lwMaKL_g.png" /><figcaption>DoiT Navigator SKU Filtering</figcaption></figure><p>6. Click Save and Run the report:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KnC3Pbz1oi8Rtzk3J1vQcA.png" /><figcaption>DoiT Navigator Network Inter Zone Egress Report</figcaption></figure><p>There you have it. You can also find other operations like email scheduling and more on the left menu.</p><p>ps. Don’t forget to turn off your VPC Flow logs.</p><p><em>To stay connected, follow us on the </em><a href="https://blog.doit-intl.com/"><em>DoiT Engineering Blog</em></a><em>, </em><a href="https://www.linkedin.com/company/doitintl/"><em>DoiT Linkedin Channel</em></a><em> and </em><a href="https://twitter.com/doitint"><em>DoiT Twitter Channel</em></a><em>. To explore career opportunities, visit </em><a href="https://careers.doit-intl.com/"><em>https://careers.doit-intl.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e1acc11ac03c" width="1" height="1" alt=""><hr><p><a href="https://medium.com/doit-international/identifying-gcps-hidden-network-inter-zone-egress-costs-e1acc11ac03c">Identifying GCP’s Hidden Network Inter-Zone Egress Costs</a> was originally published in <a href="https://medium.com/doit-international">DoiT</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>