Integration platforms and ESB development
Resilient microservices connecting 1C, BPM systems and external APIs. We address load, data loss and synchronization problems that standard connectors cannot handle.
When standard integrations stop working
Signs that your architecture needs engineering attention.
Blocking during peak hours
Scheduled 1C exchange jobs for marketplaces or electronic document services block user sessions. Warehouse operations stop and accountants cannot post documents.
Inventory out of sync
Missing retries and idempotency lead to duplicate orders or lost price updates when APIs fail. Marketplace penalties keep growing.
Slow data exports
OData and 1C web services struggle with more than 10,000 SKUs. Exports take hours instead of minutes, and analytics rely on yesterday’s data.
Technology stack
Microservice architecture
Independent Go and Symfony 7 services let you scale orders, inventory and counterparty processing separately without rewriting the monolith.
Message queues
RabbitMQ or Kafka for asynchronous exchange, with delivery controls, exponential backoff and dead-letter queues for failed messages.
Direct SQL replica reads
Direct access to read-only PostgreSQL replicas reduces production ERP load and speeds up large exports. In suitable workloads, the reduction can reach 90%.
Integration platform development process
Assessment and design
We analyze architecture, load and pain points, design data flows, select the stack and define service boundaries.
MVP development
We implement and test the key data flows: reading 1C replicas, queue processing and external API calls.
Load testing
We simulate three times the current peak load and test retries, dead-letter queues and monitoring under stress.
Deployment and handover
We deploy on-premises or in the cloud, configure Prometheus and Grafana monitoring, provide documentation and train your team.
Typical integration architecture
Example e-commerce architecture handling more than 10,000 orders per day.
What you receive
An engineered system with code, documentation and monitoring.
Retry settings for the Ozon API
Exponential backoff with jitter prevents a surge of requests when an API recovers. Dead-letter messages are retained for seven days before automatic archiving.
# messenger.yaml — transport configuration
framework:
messenger:
transports:
ozon_sync:
dsn: '%env(RABBITMQ_DSN)%'
options:
exchange: { name: ozon_integration, type: direct }
queues:
ozon_sync_queue:
binding_keys: [ozon_sync]
arguments:
x-dead-letter-exchange: 'dlx_exchange'
x-message-ttl: 604800000 # 7-day TTL
routing:
'App\Message\SyncOzonStock': ozon_sync
For businesses with annual revenue of RUB 500 million or more · NDA available
Frequently asked questions
Does connecting the integration platform require changes to 1C?
No. It runs as an external service and connects through OData or direct SQL replica reads. The standard 1C configuration remains unchanged, helping preserve compatibility with updates.
What infrastructure is required?
A starting setup uses 2 vCPUs and 4 GB RAM for the service, plus separate RabbitMQ and PostgreSQL instances. For workloads above 50,000 transactions per day, we consider a three-node cluster with horizontally scalable consumers.
Can it integrate with our existing BPM system?
Yes. The integration layer connects to Elma365, SimpleOne, BpmSoft and other BPM platforms through REST or gRPC APIs. We adapt data exchange to your system’s requirements.
Who maintains the system after implementation?
We provide documentation, configured monitoring and team training. If needed, an SLA support agreement defines incident response times.
