Five Engineering Practices Developers Can Use to Control Cloud Costs

·1 views

Cloud spending is shaped by application architecture, resource allocation, deployment decisions, and data storage, not just by infrastructure pricing. Because developers make many of these choices during routine work, the article argues that cost optimization is an engineering responsibility rather than solely a job for finance or infrastructure teams. It outlines five practices for AWS, Azure, and Google Cloud users that aim to reduce spending without hurting performance or reliability. Key among them is right-sizing resources by comparing allocated capacity with actual CPU, memory, storage, and traffic needs, while still accounting for peak demand. The piece also highlights autoscaling to match capacity with fluctuating workloads and tiered storage to avoid paying premium rates for rarely accessed data.

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)

Why Stripe payouts often differ from bank deposits and ledger records

A synthetic September example shows a €3,496.99 gap between the ledger's net Stripe activity and the Stripe deposits in the bank. The gap is not one error but several differences with opposite signs. The largest share, €3,347.78, is a payout that reached the bank on 1 October after the month-end cutoff. The remaining differences include duplicate sales, unbooked disputes and fees, and amount mismatches. The approach breaks the reconciliation into operation, payout, and bank levels, with a bridge that must add up to the cent.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Building a Dependency-Free Job Queue in Go Using Channels and Goroutines

The article argues that many applications can handle background job processing without Redis or other external queues, since Go's concurrency primitives are sufficient for in-process work. It notes that external queues remain the better choice when persistence across restarts, distributed workers, or guaranteed delivery are required. The author presents a channel-based design where jobs are closures passed through a buffered channel to a pool of worker goroutines. The Enqueue method returns false when the queue is full, leaving the caller to decide how to respond, and workers handle context cancellation separately from channel closure. A sample HTTP server demonstrates the queue sending emails in the background with graceful shutdown.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to Work with Multiple Git Remotes in One Repository

A git remote is the location where a repository is stored, and hosting platforms such as GitHub are commonly used for this purpose. The guide explains how to create a new, empty repository and link an existing local repository to it using the git remote add command with a chosen name and URL. Once connected, changes can be pushed to the chosen remote. A default remote can be set through the push.default configuration, or a specific remote and branch can be named explicitly in each push command.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Artificial Analysis Scores DeepSeek V4.1 Flash 1017 Higher, at Lower Per-Task Cost

Artificial Analysis rates the reasoning version of DeepSeek's V4.1 Flash 1017 at 45 on its Intelligence Index, up from 39 for the previous V4.1 Flash. The model's per-token list price is unchanged, yet the cost to run the index fell from $0.49 to $0.26 per task, with output token usage dropping from 250 million to 210 million. The analysis notes that the answers remain relatively long, and that the reported cost assumes peak-hour pricing and maximum reasoning effort, which many users may not match. The findings were posted to Hacker News on October 19, after DeepSeek had released the update on October 17 with a promise of shorter outputs.

0 comments Read more at DEV Community