Software Integration Services: Connect Your Systems Without a Rewrite

Software integration services connect the systems you already run, such as accounting, CRM, inventory, ecommerce and an older IBM i or AS400 platform, so data moves between them without people re-keying it. In most cases that is cheaper and safer than replacing a system that still does its job.

The hard part is not the first connection. It is choosing the right pattern, planning for the failures that appear months later and agreeing who looks after it. This guide covers each of those in plain English.

Key Takeaways

• There are five common ways to integrate business systems: direct API calls, file transfer, a message queue, an integration platform and a thin API layer in front of a legacy system.
• Most integration failures are quiet ones: duplicate records, syncs that stop without an alert, rate limits and a supplier changing a field.
• An older IBM i system can usually be connected rather than replaced, because IBM's own tooling can expose its programs and database as web APIs.
• A good scope lists every system, data flow, direction, frequency and error rule, and names one owner for each record type.
• Price the risk, not just the build: ask what happens when a connected supplier changes their interface, and who is paid to respond.

What Do Software Integration Services Actually Include?

System integration makes separate applications share data and trigger each other's actions. A typical job pushes a paid web order into accounting, updates stock and creates a CRM record, with nobody typing it twice.

The deliverables are more than code. A competent team maps the data, builds the connections, adds error handling and monitoring, documents how it works and sets up support. If a quote covers only the connection, running costs are left for you to find later.

Integration beats replacement when the existing system holds business rules that are expensive to rebuild and its only problem is isolation. Replace it when it is failing, unsupported or blocking growth.

Which Integration Pattern Fits: API, File Transfer, Queue or Platform?

The right pattern depends on how fast data must move, how much there is and what happens if a message is lost. The table summarises the five common choices.

• Direct API link - Best when: Two systems with modern APIs, near-real-time updates; Watch out for: Rate limits, and one system being down breaks the other
• File transfer - Best when: Nightly batches, bank or supplier feeds, older systems; Watch out for: Delays, half-written files, format changes
• Message queue or events - Best when: Bursts of traffic, several systems reacting to one event; Watch out for: Duplicate messages, harder debugging
• Integration platform - Best when: Popular cloud apps, simple rules, little custom logic; Watch out for: Fees that grow with volume, vendor limits
• API layer on a legacy system - Best when: One older system must serve several newer ones; Watch out for: Needs someone who understands the legacy logic

Direct API integration

API integration means one system calls another's published interface over the web. It is simplest for two modern systems, but it gets fragile as you add more: connect six applications to each other directly and you can end up maintaining fifteen separate links.

File transfer and batch jobs

One system writes a file, such as a CSV, and another collects it on a schedule. It sounds dated, but it suits nightly stock or ledger updates and is often the only option for older systems. Agree what happens when a file arrives late, twice or incomplete.

Message queues and event-driven design

The sending system drops a message into a queue and walks away, and other systems pick it up when ready. That copes with bursts. The catch is that queues can deliver a message more than once. Amazon's documentation says its standard SQS queues use at-least-once delivery and tells builders to make applications idempotent, meaning that processing the same message twice does no harm.

Integration platforms and middleware

An integration platform, often called iPaaS, is hosted middleware: software between your applications that supplies ready-made connectors, scheduling and monitoring. It suits popular cloud apps with simple rules. It suits unusual logic or high volume less, because the vendor sets pricing and limits.

How Do You Connect a Legacy System Like IBM i or AS400?

Legacy system integration usually means putting a thin API layer in front of the old system, so newer applications use a clean interface and never touch its internals. This is the core of most IBM i modernisation work, and our guide to AS400 modernisation without a risky rewrite covers the wider picture.

IBM's own tooling makes this practical. Its developer tutorial on creating REST APIs from SQL statements says IBM i's integrated web services can expose RPG, COBOL, C and command language programs as REST and SOAP services. It adds that IBM i 7.4, the release current in August 2019 when it was written, can also turn SQL statements into REST APIs. The tutorial lists the fixes each release needs, so a technical review of your version comes first.

Why not read the database directly?

Because that bypasses the business rules held in the programs. A price calculated by an RPG program and a price read from a table can differ. An API that calls the existing program keeps one source of truth. The same layered thinking applies to industry business systems, where several departments depend on the same records.

What Goes Wrong After Go-Live, and How Do You Monitor It?

Integrations rarely fail loudly. They drift out of step, and nobody notices until a customer or an accountant does. These are the four failures we plan for.

• Duplicate records. The same order arrives twice because a request was retried. Stripe's webhook documentation says endpoints might occasionally receive the same event more than once and recommends logging processed event IDs.
• Silent sync failures. A connection stops, no error is raised and the gap grows. Daily count checks and a no-activity alert catch it.
• Rate limits. Many services cap requests per minute. The standard response is HTTP status 429, which MDN notes may include a Retry-After header saying how long to wait. A well-built connection slows down and retries.
• Schema changes. A supplier renames or removes a field and your mapping breaks. Version the mappings and watch the supplier's change notices.

What should monitoring include?

At minimum: an alert for every failed message, a list of items that could not be processed, a daily reconciliation of record counts and a dashboard your operations lead can read. Stripe also advises returning a success response quickly and doing heavy work afterwards, and says it retries failed deliveries for up to three days in live mode. Other systems differ, so check each one's retry rules.

Who Owns the Data, and Who Fixes It at 2am?

Every record type needs one system of record, the one that wins when two systems disagree. Without that decision, two-way syncs overwrite each other.

Then agree who responds when something breaks. Ask your vendor in writing who is alerted, how fast they respond and who fixes the data. If a supplier's API is down, your integration partner can only diagnose it, so the contract should say what they do while waiting. Budget for that support as a running cost, not an afterthought.

How Do You Scope an Integration and Price the Risk?

Write a one-page inventory before asking for quotes, so every supplier prices the same job.

• Systems: every application, its version and its administrator.
• Data flows: orders, customers, stock, invoices or other records.
• Direction: one-way or two-way, and which system wins a conflict.
• Frequency and volume: real time, every few minutes or nightly, and records per day now and in two years.
• Error handling: what happens to a failed record and who is told.
• Access: test environments, sample data and API documentation.

Where does the cost risk sit?

Quotes hide three unknowns: undocumented behaviour in the legacy system, a third-party interface that changes and data quality problems found during mapping. An honest quote either includes a short discovery phase to remove them or states its assumptions and prices a contingency. As an illustration only, a supplier might offer a fixed-fee discovery first, then quote the build once the data flows are confirmed. A single number with no stated assumptions is a warning sign.

For platform versus custom build, ask for both with three-year totals in AUD, including support.

Related Resources

• AS400 modernisation without a risky rewrite
• Industry business systems from 9Byte
• IBM tutorial on creating REST APIs from SQL statements
• IBM i integrated web services documentation
• Stripe webhook documentation
• MDN guide to the 429 Too Many Requests status
• Amazon SQS at-least-once delivery

Common Questions About Software Integration Services

1. What are software integration services?

They are the design, build and support of connections between business applications so data flows automatically.

2. What is the difference between API integration and middleware?

API integration connects two systems directly through their published interfaces. Middleware sits between several systems and handles routing, data conversion and monitoring.

3. Can you integrate an AS400 or IBM i system with modern software?

Usually yes. IBM documents ways to expose IBM i programs and Db2 data as web services, so newer applications can call them without replacing the platform. A technical review of your release comes first.

4. How long does a system integration project take?

It depends on the number of systems, the quality of their documentation and the data clean-up needed, so there is no standard figure. A short discovery phase gives a firmer estimate.

5. How do I know if my integration has stopped working?

You will not, unless it is monitored. Ask for alerts on failed messages, a daily record-count check between systems and a named person who receives the alerts.

Connect What Works Before You Replace It

Most businesses do not need a new platform to stop re-keying data. They need their existing systems to agree with each other, with someone accountable when they stop.

If you are weighing a connection project, write down your systems, data flows, direction, frequency and error rules, then send 9Byte a brief and we will respond with options and the assumptions behind each price.