A practical IBM i application modernisation roadmap for 2026: compare six strategies from screen transformation and API wrapping to selective rewrite and full migration, apply the strangler fig pattern, assess your RPG estate, and build the business case for phased modernisation.
Vector search and text embeddings bring semantic search to IBM i in 2026 — converting DB2 for i text fields into float arrays via OpenAI or watsonx.ai, storing vectors as VARBINARY or JSON in DB2, computing cosine similarity in SQL, and building RAG pipelines that ground LLM answers in real IBM i business data. This post covers the full embedding pipeline from generation to retrieval.
DB2 for i replication in 2026 is built on IBM i journals — the same journal receivers that underpin HA tools and Kafka CDC also power Q-Replication and MIMIX. This post explains journal-based logical replication, the Q-Replication Capture and Apply architecture, what MIMIX adds for HA and DR, and how to design a replication strategy that matches your latency targets and data distribution needs.
A hands-on guide to connecting IBM i with Apache Kafka in 2026 — running kafkajs and confluent-kafka-python producers in PASE, journal-based change data capture using QSYS2.DISPLAY_JOURNAL, Kafka Connect via IBM MQ bridge, consuming Kafka messages to write back into DB2 for i, and practical event-driven architecture patterns for IBM i integration.
A practical guide to IBM i display files in 2026 — DDS syntax, CRTDSPF, RPG EXFMT screen I/O, subfile programming with SFL and SFLCTL, indicator handling, and a clear-eyed look at when and how to modernise 5250 green-screen interfaces with Profound UI, Node.js REST APIs, or IBM i Access Client Solutions.
IBM i PTF management in 2026 spans four distinct fix types — individual PTFs, cumulative PTF packages, SF99xxx group PTFs, and Technology Refreshes — each with its own ordering, application, and scheduling discipline. This post explains how to check what is currently installed with DSPPTF and WRKPTF, which SF99xxx groups matter most (DB2, Java, open source, security, HTTP, and TCP/IP), how to order fixes via SNDPTFORD and IBM Fix Central, and how to apply them safely using APYPTF and APYGRPPF with an understanding of IPL-required PTFs and hold/release mechanics. The post also covers using IMGCLG image catalogues to load cumulative packages without physical media, and walks through the order of operations for planning a full IBM i OS version upgrade — from CHKPRDOPT and licence checks through GO LICPGM and the upgrade IPL itself.
IBM i high availability and disaster recovery in 2026 requires understanding the full stack: geographic mirroring at the storage layer, switchable Independent Auxiliary Storage Pools for role swaps, and journal-based logical replication for granular data protection. This post compares the major IBM i HA architectures and leading third-party tools — MIMIX, iTera, and PowerHA SystemMirror for i — explaining how to align your RTO and RPO targets with the right replication strategy. You will learn how IASPs enable a production disk pool to be varied off one system and varied on another, how geographic mirroring keeps a remote copy synchronised at the hardware level, and how journals underpin every logical replication product on the platform. Practical CL command examples cover VRYCFG, STRGEOMRR, DSPGEOMRR, CHGJRN, and the steps needed to execute and test a planned switchover.
Python runs natively on IBM i in PASE with direct DB2 access via ibm_db and pyodbc, IBM i program calls via itoolkit-for-i, and access to the full Python AI ecosystem. This post covers installing Python via yum, querying DB2 for i with ibm_db_dbi, loading IBM i data into pandas DataFrames, calling CL commands and RPG programs with itoolkit, extracting IBM i data to AWS S3 with boto3, batch AI classification with the OpenAI Python library, watsonx.ai integration, running Python from CL, and when to use Python versus Node.js versus RPG.
The IBM i OS API layer accessible from ILE RPG: QCMDEXC for running CL commands at runtime, the list API pattern using user spaces (QUSCRTUS, QUSLOBJ, QUSRTVUS), listing objects in a library at runtime, dynamic program calls using variable names and procedure pointers, system pointers and the capability-based OS model, and QUSROBJD for checking object existence without triggering an error message.
Advanced IBM i IFS topics for production environments: enabling journalling on IFS directories for recovery and replication, sharing IFS directories as Windows network drives through NetServer, NFS exports for Linux clients, saving and restoring IFS data with SAV/RST, large file support for files over 1GB, and CCSID encoding management — why CCSID 1208 (UTF-8) matters, how to set it, and the CCSID 65535 trap that breaks modern tooling.