How to Build a Scalable Backend Stack for Startups: Languages, Services, and Tools

Backend Technology Stack Overview

The backend stack can be divided into four layers: programming languages, components (e.g., MQ, databases), processes and standards (development, release, monitoring, code conventions), and systems that enforce these processes.

Combining these layers yields the overall backend architecture shown in the diagrams.

Component Selection

1. Project Management / Bug Tracking

Redmine – Ruby‑based, many plugins, but some are outdated.

Phabricator – PHP, originally from Facebook, integrates code review and task management.

Jira – Java, supports user stories, tasks, burndown charts, suitable for large teams.

悟空 CRM – Customer‑centric CRM with basic project‑tracking features; suitable for small B2B teams.

2. DNS

Alibaba Wanwang – integrated DNS service after Alibaba’s 2014 acquisition.

Tencent DNSPod – provides DNS resolution and protection; widely used in China.

For overseas services, Amazon Route 53 is the typical choice.

3. Load Balancer (LB)

Supports L4 (TCP/UDP) and L7 (HTTP/HTTPS) protocols.

Centralized certificate management for HTTPS.

Health checks.

Use cloud provider LB (Alibaba SLB, Tencent CLB, AWS ELB) or self‑hosted LVS + Nginx.

4. CDN

Domestic leaders: Wangsu, Tencent Cloud, Alibaba Cloud.

International leaders: Amazon CloudFront, Akamai.

Multiple CDNs are recommended for redundancy.

5. RPC Frameworks

Cross‑language options: Thrift, gRPC, Hessian, Hprose. Service‑governance options: Dubbo, DubboX, Motan, rpcx.

6. Name/Service Discovery

etcd – distributed key‑value store used by Kubernetes.

Consul – service discovery and configuration tool.

Apache ZooKeeper – coordination service originally for Hadoop.

7. Relational Databases

MySQL – mature, widely used; MariaDB as a community‑maintained fork.

NewSQL examples: CockroachDB, TiDB (widely adopted in China).

8. NoSQL Databases

Key‑Value: Redis, Memcached.

Column‑Family: HBase, Cassandra.

Document: MongoDB, CouchDB.

Graph: Neo4j, InfoGrid.

9. Message Middleware

Used for asynchronous processing, system decoupling, and traffic shaping (e.g., flash‑sale order queuing). Common choices include Kafka, RocketMQ, RabbitMQ, and custom MySQL/Redis queues.

10. Code Management

Git – de‑facto version control.

GitLab – open‑source Git hosting with CI integration.

Gerrit – provides advanced code review workflow.

11. Continuous Integration (CI)

Jenkins – Java‑based, extensive plugin ecosystem.

TeamCity – user‑friendly, commercial after a certain scale.

GitLab CI – built‑in CI for GitLab projects.

Travis, Strider, Go Cruise Control – other options.

12. Log System

ELK stack (Elasticsearch, Logstash, Kibana) plus Filebeat for lightweight log collection. Nginx can act as a reverse proxy for Kibana security.

13. Monitoring System

Prometheus (written in Go) with Grafana for visualization. Prometheus scrapes metrics via HTTP, supports exporters for databases, message queues, etc.

14. Configuration Management

Centralized stores (etcd, ZooKeeper) with UI/API and version history.

Automation tools (Puppet, Ansible) to push config files.

15. Release / Deployment System

Typical flow: code → artifact → deployable service → production. Open‑source tools like Walle, Piplin, or a Jenkins + GitLab + Walle combo can cover artifact management, release workflow, permissions, gray‑release, and rollback.

16. Jump Server

Jumpserver provides role‑based access, audit logs, and session recording for privileged operations.

17. Machine Management

Ansible – agentless, SSH‑based, suitable for small‑to‑medium fleets.

Puppet/Chef – Ruby‑based, require agents.

SaltStack – ZeroMQ‑based, high concurrency.

Startup Recommendations

1. Choose the Right Language – prioritize team familiarity, modern features (memory management, concurrency), community support, and hiring ease.

2. Choose Components and Cloud Provider – prefer mature open‑source projects with strong community backing; use cloud services for speed, then migrate to self‑hosted solutions as scale grows.

3. Define Processes and Standards – establish coding conventions, release procedures, operational guidelines, database access policies, alert handling, and reporting mechanisms.

4. Build or Adopt Supporting Systems – select open‑source tools that match the chosen language and component stack to solidify processes.

5. Consider Long‑Term Trade‑offs – evaluate technology lock‑in, team skill growth, performance requirements, and future scaling needs (e.g., 200‑500 engineers).

Reference Architecture

The final architecture for a cloud‑based startup backend combines the selected services into a cohesive diagram (see image).

Reference Materials

http://database.51cto.com/art/201109/291781.htm

https://zh.wikipedia.org/wiki/Kafka

https://prometheus.io/docs/introduction/overview/

http://deadline.top/2016/11/23/配置中心那点事/

http://blog.fit2cloud.com/2016/01/26/deployment-system.html

http://ju.outofmemory.cn/entry/351897