Horizontal vs Vertical Scaling: When to Use Which for Growing Apps

·5 views

As applications grow to serve millions of concurrent users, a single server often struggles with load, crashes, and data inconsistencies, making scaling essential. Vertical scaling solves this by upgrading a single machine's CPU, memory, and storage, but hits a hard limit and creates a single point of failure. Horizontal scaling adds more machines and uses a load balancer to distribute traffic, theoretically allowing unlimited growth and better fault tolerance. However, horizontal scaling requires applications to be stateless, as session-based authentication can break when requests are routed to different servers. A common fix is to store sessions in a shared external database like Redis, which also supports automatic session expiration via its TTL feature.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

How to Run a Structured Pilot Before Committing to Google Beam

Google has announced a broader rollout of Beam, its enterprise communication platform delivered via HP Dimension hardware, with integrations for Google Meet and Zoom and access through select Industrious workspaces. However, experts caution that vendor-reported benefits may not translate directly to every engineering or distributed team's real-world needs. Organizations considering adoption are advised to identify a specific recurring meeting problem first, then design a controlled pilot using comparable sessions with clearly defined success criteria. The pilot should test practical scenarios such as code review, diagram sharing, and accessibility requirements, rather than relying on polished product demonstrations. Teams should also request full documentation on costs, data handling, fallback options, and both adoption and rejection criteria before making any purchase decision.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Why Prompt Engineering May Be the Wrong Skill to Master in the AI Era

A software engineer argues that the widespread focus on prompt engineering is misplaced, despite its popularity as a sought-after skill in the AI industry. While better prompts can produce cleaner, more polished outputs, they also make incorrect answers appear more convincing and authoritative, making errors harder to detect. The author contends that the real failure mode is not obvious AI mistakes but plausible, well-structured outputs that are subtly wrong and only discovered in production. He further argues that prompt engineering is essentially a form of recall-based knowledge, the same type of skill that AI itself has already begun to commoditize. According to the author, the skill that truly retains value is not the ability to ask better questions, but the critical judgment to evaluate and verify the answers produced.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Why RLS Approval Still Causes Postgres 'Permission Denied' and How to Fix It

A developer building a coffee shop stored-value system encountered a persistent 'permission denied for table users' error in PostgreSQL, despite having a valid Row-Level Security (RLS) policy that explicitly allowed admin access. The root cause turned out to be a two-gate permissions model: GRANT controls table and column access, while RLS only governs which rows are accessible — and the developer had been debugging the wrong gate. Because the admin RPC function was set to SECURITY INVOKER, it ran under the authenticated role, which had never been granted write access to sensitive columns like balance and tier fields. The fix was to create a SECURITY DEFINER wrapper function that runs as the database owner, with an is_admin() guard placed as the very first statement to prevent privilege escalation. This pattern allows only verified admins to write privileged columns, while keeping those columns fully locked down from ordinary authenticated users.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How Shared Transactions in Code Quietly Implement the Unit of Work Pattern

A technical article on DEV Community illustrates how the Unit of Work pattern can exist in code without any formal labeling or dedicated class structure. The example involves a salary and batch processing screen that executes two database operations — an insert and an update — sharing the same connection and transaction objects. The second operation is conditionally gated on the success of the first, ensuring both either complete together or roll back entirely. This all-or-nothing structure prevents partial writes that could leave a database in an inconsistent state, a particularly serious risk in payroll contexts. The piece argues that recognizing such patterns by their intent and behavior, rather than by explicit naming conventions, is key to understanding the design choices already present in everyday code.

0 comments Read more at DEV Community