Vladislav Maličević

We’re excited to announce pgstream v1.0.0, a major milestone and a long-planned architectural shift for pgstream. This release introduces stateless schema change replication, removing schema logs and simplifying how Postgres DDL changes are captured, replicated, and processed downstream. The result is a cleaner architecture, fewer moving parts, and more faithful schema replication across targets. If you work with Postgres CDC, schema evolution, or data pipelines, check out the full write-up 👇 https://lnkd.in/d-bEtE3a

27

1 Kommentar

Most production failures are not outages. They are visibility failures. Over the last two years, CloudInfosoft has observed this pattern repeatedly across fast-scaling, US-based mid-market systems. One recent example from a distribution platform: • ~18 microservices • ~12 upstream and downstream integrations • Average daily order volume increased ~2.4x in 14 months On paper, uptime metrics looked healthy. In reality: • Mean time to detect issues was ~42 minutes • Business-impacting failures were identified by customer support, not alerts • Root cause analysis routinely spanned 4 to 6 teams The issue was not reliability. The issue was that observability stopped at infrastructure-level metrics. What actually changed outcomes: • Instrumenting end-to-end business transactions from order creation through fulfillment • Defining ownership at API boundaries rather than individual services • Treating ERP as a system of record, not as a workflow engine Measured results over two quarters: • MTTD dropped from ~42 minutes to under 10 minutes • P95 order processing latency stabilized within a 12 percent band during peak load • Incident review meetings reduced from weekly to bi-weekly due to fewer cross-team escalations No system rewrite was required. What was required was acknowledging that the system had outgrown its original mental model. CloudInfosoft is interested in how other technology leaders are seeing this in their own environments. Two questions for peers: 1. Which metric consistently tells you first that something is wrong in production. 2. At what point did your ERP or core system stop reflecting how work actually happens Looking forward to comparing notes. #CTO #CIO #CEO #flutter #ERP #ERPIntegration #SystemsEngineering #Production #USA #cloudinfosoft.

8

What is the most common pitfall in NAV/BC cloud migration? Delays. They occur when business expectations and technical execution are not aligned. Why? Due to the fact that many projects fail to establish a clear migration and maintenance strategy from the outset. ❎ Without this step, go-live visions will change, budgets will decrease, and timelines will become longer. ✅ With it, businesses get stakeholders on the same page, plan realistically, and avoid unnecessary rework. To ensure a smooth migration to the Business Central cloud, it's important to define the strategy early as explained in the video. 📌Start exploring the cloud migration process in our Customer Portal → https://lnkd.in/e_27A6NU #BusinessCentral #CloudMigration #ERP #MicrosoftDynamics

16

Proven Scale Without Complexity Key Takeaways: * 2 million data assets managed without Kafka infrastructure * 1,000 concurrent users with sub-400ms response times * 10x faster than comparable open-source platforms * Simplicity enables scaling from small organizations to enterprise * Downstream notification system alerts consumers of schema changes automatically Key Teaching Moment: Real-world performance metrics that prove simplicity scales

14

In most teams today, the real challenge is not automation. It is orchestration.   We have thousands of automated tests but very few systems that know how to run them with intelligence. We have real devices and browsers in the cloud but no mechanism that matches the right test to the right environment at the right moment. We have CI CD pipelines pushing faster but execution layers that still operate on outdated assumptions.   The gap is not in our ability to create tests. The gap is in our ability to execute them with purpose.   This is where the future of testing is shifting. Away from manual coordination Away from heavy setup Away from brute force regression   And toward intelligent, context aware, device aware orchestration that aligns execution with actual product change.   QuantumRun embodies this shift.   It prioritizes based on code impact, orchestrates real devices on demand, isolates flaky behavior, and unlocks massive parallel execution without increasing complexity. In other words, it transforms test execution from an operational burden into a strategic accelerator. As teams push toward faster releases and higher quality standards, the next competitive edge will not come from writing more automated scripts. It will come from running the right tests, at the right time, on the right infrastructure, with complete clarity. Test orchestration is becoming the new center of gravity in quality engineering. QuantumRun is one step in that direction. #TestOrchestration #Automation #Pcloudy #QuantumRun

6

Your CI/CD pipeline passes every test. Your uptime is 99.9%. Your Datadog dashboards are green across the board. And none of that matters for EU AI Act compliance. I've been talking to SaaS CTOs who wrap LLMs into their products. Most of them assume compliance is a legal problem to handle before August 2026. It's not. It's an architecture problem. And by the time your engineering team realizes the codebase is structurally incompatible with Article 12 traceability requirements, you're not filling out forms. You're rewriting your backend under deadline pressure. Here are 7 architecture traps we keep seeing: 1. The "It's Just a Feature" classification error. Your email summarizer is low-risk until a customer uses it for recruitment screening. Now it's high-risk under Annex III. If your multi-tenant architecture treats all tenants the same, you're stuck applying governance to everyone or failing to comply for the ones that matter. 2. Documentation as a dead PDF. You write a 40-page technical doc. Two weeks later, OpenAI deprecates that model version. Your PDF is now outdated. The Act requires living documentation, not snapshots. 3. Outsourcing bias testing to your model provider. OpenAI's safety filters won't catch your prompt structure down ranks non-native English speakers. You are the system provider. The bias in your prompts and RAG logic is your responsibility. 4. Thinking your CI/CD pipeline is a Quality Management System. Article 17 requires risk management and incident reporting. A failing unit test is a code issue. A model hallucinating legal advice is a risk issue. Your DevOps tools don't know the difference. 5. Ignoring the value chain. If OpenAI changes terms or has an outage that breaks your high-risk system, the regulator comes to you. Not them. Article 25 is clear on this. 6. Planning to start compliance in 2026. Notified Bodies (independent auditors) are already scarce. By mid 2026, the queue will be months long. If you aren't audit-ready, you might be legally unable to operate while you wait. 7. Scattered logs across five different systems. A regulator asks why your AI rejected a candidate on November 14th. If you need to manually stitch CloudWatch, your database, your vector DB, and your provider's dashboard to answer that question, you fail Article 12. The fix for all of these isn't a compliance tool you bolt on later. It's architectural decisions you make now: tenant-aware risk profiles, docs-as-code in your CI/CD, fairness regression testing, centralized AI decision logging. The companies that treat compliance as an engineering discipline (not a legal checkbox) will move faster, not slower. Full breakdown with engineering patterns for each trap 👇 https://lnkd.in/dyaFXp5j What's the biggest compliance blind spot in your AI architecture?

2

🗣️ "I remember thinking PostgreSQL 7 was as good as it gets. Well, let me put it mildly: I was wrong, and things are a lot better than they used to be." That reflection comes from our Founder & CEO, Hans-Juergen Schoenig, and it frames a crucial question for every PostgreSQL Professional: What comes after PostgreSQL 18's groundbreaking Asynchronous I/O? 🤔 Hans looks beyond PostgreSQL 18 to explore the future of disk I/O, latency, and throughput in PostgreSQL. If you manage high-volume data, then this expert analysis is for you. ▶️ https://ow.ly/x3l750X4c2S #PostgreSQL #PostgreSQL18 #Database #FutureofPostgres #CYBERTEC

19

🚀 Modern enterprises are moving at top speed. If you’re running Kubernetes at scale, you already know that reliability isn’t optional. In our latest blog, we explore how the new integration between Gremlin and Dynatrace automatically discovers Kubernetes services and lets you onboard fault-injection tests in minutes. 👉 The highlights: • Auto-discovery of K8s services via Dynatrace topology mapping. • Health checks tied to observability metrics (error-rate, latency) for meaningful failure testing. • Scale reliability testing across your org without slowing deployment cadence. Read more here: https://hubs.la/Q03Sw95h0

5