“Architecting off platform.”

“Architecting off platform.” Whuuutt 😱 whaddayamean?? Not ON Salesforce?!

Blasphemy!!

😅 I’m only being a tad facetious. Martin Gardner from Slalom kicking off tonight’s

#LDNArcs on this interesting topic.

With the help of otter and Claude, here are the notes from the cracking session:

**Why Go Off-Platform?**

• Governor limits (heap size, API calls, CPU timeout)

• Cost optimization (S3 vs Salesforce storage)

• Specialized capabilities (AI/ML, complex algorithms)

• Multi-system orchestration needs

**3 Core Integration Patterns:**

1️⃣ **Asynchronous Offload (Fire & Forget)**

Best for: Long-running processes

Example: Platform events → Heroku/Lambda → process → write back

✅ Resilient, decoupled

⚠️ Eventual consistency, needs user feedback

2️⃣ **Synchronous Augmentation**

Best for: Real-time data needs (credit checks)

✅ Immediate results

⚠️ Speed critical (~20 sec limit), tight coupling

3️⃣ **Data Virtualization**

Best for: External data in Salesforce (Connect/Data Cloud)

✅ Smooth UX, standards-based

⚠️ Limited triggers, reporting constraints

**Critical Gotchas:**

🔒 **Security**

– Never hardcode credentials.

Use Named Credentials, key vaults, Private Connect.

Users are your biggest vulnerability!

🛡️ **Sharing & Access Control**

– Salesforce sharing rules DON’T apply off-platform.

Must replicate manually.

Real challenge for document management and guest users.

⚡ **Idempotency**

– Design for duplicate messages.

Same message 10x = same result once.

🔄 **Eventual Consistency**

– Systems will be out of sync.

Build processes that handle this gracefully.

💥 **Error Handling**

– Implement exponential backoff, dead letter queues, and user-friendly messages.

Check out Nebula Logger for flow/Apex error logging.

**Case Study: Insurance Document Generation**

Problem: 20+ SOQL queries, CPU timeout

Solutions discussed: Batch Apex, query optimization, staging objects, Heroku, Data Cloud, GraphQL, AWS Lambda + S3

**Key Principles:**

✓ Go off-platform for the right reasons

✓ Choose the pattern that fits your use case

✓ Be pessimistic – plan for failure

✓ NEVER compromise security

✓ Consider legal implications (data residency, GDPR)

Thank you GoCardless for sponsoring this space in Barbican!

And thank you CGL Raksha Sanganee Sam Wadhwani MCMI ChMC Tom Bassett Martin Gardner

#GettingMyArchitectFix 🤩

#OnThePeiroll