In SAP upgrades, parallelism refers to executing multiple tasks simultaneously instead of sequentially, which significantly reduces the runtime of various SUM upgrade phases and improves overall upgrade performance by efficiently utilizing system resources through parallel processes.

Consider we are performing an upgrade using SoftwareUpdateManager (SUM) tool in nZDM approach, the static traditional method requires entering numerical values for several parameters related to parallelism during the INITSUBST phase of SUM. You must set these parameters associated with the different processes and tools to optimize the upgrade process.

Those Include:

ABAP Processes
SQL Processes
R3trans Processes
R3Load Processes
Table Split Processes

1. ABAP Processes:

This parameter defines how many processes are used to run upgrade tasks in parallel (batch or dialog). It affects phases like JOB, RUN, DBCLONE, ACT_UPG, PARDIST* and XPRA.
Recommended: up to 10 wp’s (20+ for high-end systems with sufficient Hardware)

2.   SQL Processes:

This parameter controls the number of parallel SQL processes at the database level to speed up operations like SELECT, CREATE, and ALTER on tables. It is used in phases such as PARMVNT*, PARCONV*, SCEXEC and SQLSELEXE.
Recommended: Equal to total physical CPUs

3.   R3Trans Processes:

This parameter handles the parallel import processes of tp and R3trans transport tools on the phases where import of support packages take place. This occurs mainly in the phases DDIC_UPG, TABIM_SHADOW*, and TABIM*.
Recommended: Equal to total physical CPUs

4.   R3Load Processes:

This parameter helps to load the data and to import repository objects during the different phase of SUM like EU_IMPORT*.
Recommended: 3–5 × number of CPUs

5. Table Split processes:

This parameter defines the maximum number of parallel processes to be used when splitting multiple tables at the same time.

In a Nutshell – with this manual static parallelism approach, we must predefine all parallelism parameters and these values stay constant throughout the run unless we change manually in SUM utilities section.

Disadvantages of this static approach:

If we set the values too high this impacts the system performance in turn impacting the business users during uptime.
If we set too low there is significant longer SUM run during uptime and increased Downtime phase.
This approach does not react to CPU spikes, I/O bottlenecks and Lock

Have a look- how this static approach will be changed in the future upgrades using Agentic AI.

With the recent release of S/4HANA 2025 – upgrades are being fundamentally transformed by Agentic AI, shifting from manual, rule-based processes to autonomous, self-healing systems. This is achieved through the new agentic AI framework, where task-specific agents work together under the orchestration of Agentic AI copilot Joule.

For example, consider a SUM phase utilizing the R3 load work process with 24 parallel load jobs in static mode. An Agentic AI layer continuously monitors the runtime metrics such as active R3load jobs and CPU utilization. The AI then observes that CPU headroom exists and recommends increasing parallel jobs- the R3Load parallelism parameters are increased slightly (for example from 24 to 28 jobs), improving throughput and vice versa. This creates a continuous closed-loop cycle of observe–analyze–tune–validate, allowing more adaptive and efficient R3load execution compared to static parameter settings, while still respecting SUM’s strict phase stability constraints.

A static upgrade → fixed parameters, reactive
An agentic upgrade → adaptive, predictive, optimized

The above example indicates the usage of Agentic AI only in parallelism processes, we should also remember that suitable AI agents based on the requirement would completely transform the entire upgrade process (END-END) as outlined below:

In next 1-2 years this would be really an emerging pattern in upgrades – combining SAP tools with AI assistants to automate decisions, self-tuning runs, proactive monitoring, and AI-driven exact downtime prediction, error handling during the upgrade lifecycle. In one line it’s a smart co-pilot that continuously tunes your upgrade while it’s running.

“AI is no longer just a trend—it’s a game changer. For upgrade consultants, the future lies in combining deep domain expertise with AI-driven intelligence to deliver better outcomes.”