If your crypto backend falls over at peak traffic, or nobody on the team wants to own the servers, I take that over. I'm Oussama Allamou (0xMowgli). I run production infrastructure for trading systems and for backends that handle millions of requests: Docker on Linux, bare metal, multi-region PostgreSQL, dedicated chain nodes, and monitoring with Prometheus and Grafana. I deliver infrastructure you can deploy to repeatedly, see inside, and trust on your busiest day.
/ 01 · Scope
Docker-based and repeatable, with a way back when a release goes wrong. Nginx at the edge.
Services and PostgreSQL across regions, with latency-based switching between them.
Dedicated Solana nodes with gRPC and shred streams, and the RPC access your EVM services depend on.
PostgreSQL, MongoDB and Redis: replication, backups you have actually restored, and caching in front of the slow parts.
Prometheus metrics, Grafana dashboards, and alerts tied to what users feel instead of noise.
Queues and caching sized so a traffic spike is absorbed and not passed straight to the database.
/ 02 · Process
I read the deployment, the dashboards if there are any, and what broke last.
Metrics and alerts go in first. They show what to fix next, and they prove later that the fix worked.
In order of risk. The database and the signing path usually come before anything else.
One command, the same every time, with a rollback.
Either way the setup is written down, so it does not live in one person's head.
/ 03 · Proof
Low-latency Solana trading engine
I run its whole production setup: dedicated nodes, gRPC streams, Redis and the services on top.
Marketplace connecting Moroccan cooperatives directly with buyers
A non-crypto product where the DevOps is also mine, next to the web dashboard, backend and mobile app.
Private work in this area includes backends handling millions of requests and a Solana gRPC and shred-stream provider on dedicated nodes. Details are shared on a call.
/ 04 · FAQ
The base is the same: deployments, databases, monitoring. Crypto adds chain infrastructure. Nodes that must stay at the tip of the chain, RPC providers that rate-limit or serve stale data, private keys that sign transactions and must never leak, and traffic that spikes with the market. Someone who has not run nodes or handled signing keys will learn it on your production system.
Both have a place. Cloud suits stateless APIs, managed databases and anything that scales up and down. Bare metal suits chain nodes and latency-sensitive trading systems, where you want the whole machine at a fixed monthly cost. A mix is common.
Keys never go in the repository or in plain environment files on shared machines. Signing sits in the smallest service possible, reachable by the fewest people, with access logged. The exact setup depends on the threat model. A hot wallet for a bot is a different problem from a treasury. I design it with you and say what risk remains.
Only with a reason. Two are good ones: users or counterparties in different parts of the world where latency matters, and a requirement to survive a provider outage. It adds cost and complexity, mostly around the database. If neither applies, one well-monitored region with tested backups is the better choice.
Visibility comes first and comes quickly: metrics and alerts are days of work. After that the order follows the risk. Some fixes are a config change. Others, like moving a database to a replicated setup, need planning and a quiet window. After the audit you get a ranked list showing what each step buys.
/ Related
/ Contact
Tell me what you run and what broke last. I'll tell you what I would fix first.