Infor LN & BaanFree calculatorPDF, Excel, Word exportInfor LNAll Infor LN and Baan tools

Infor LN Bshell Application Server Memory and CPU Sizing Calculator

This calculator sizes memory and CPU for an on-prem Infor LN 10.x application server from the number of bshell processes you expect at peak and the memory each one really uses. It is meant for refresh, virtualisation and consolidation planning where you can measure the current system.

Your numbers

users
60 %

Share of named users logged on at the busiest hour. Count it, do not guess: sample the bshell process list at peak for a week.

bshells

1.0 means one bshell per user. Raise it for users with extra bshells from integrations, AFS/BOD connections or several open clients.

MB

Measure the resident memory of bshell processes with ps (Unix/Linux) or Task Manager (Windows) during normal work.

bshells

Job daemon and job queue bshells running at peak.

GB

Operating system, LN shared memory, database client, monitoring and backup agents.

bshells

Assumption. Most bshells are idle most of the time; lower this for heavy reporting or MRP-type users.

30 %

Your results

RAM required
24.8727
GB of memory for the application server including OS overhead and headroom.
CPU cores required
21
Physical or dedicated virtual cores, rounded up.
Peak bshells
190
Interactive bshells at the busiest hour plus background job bshells.
Memory used by bshells alone
11.1328
GB, before overhead and headroom.
Memory per named user
101.8784
MB of server RAM per named user, useful for comparing against your current server.

This is a measurement-based rule of thumb, not an Infor sizing formula. Every default is an editable assumption; replace them with values measured on your own LN server, and use Infor's sizing service for contractual hardware commitments.

Save results:

RAM (GB) and cores vs named users

Sweeps Named users from 50 to 1000, holding your other inputs fixed.

Send us your bshell measurements

Share your peak bshell count, average resident memory and user numbers and we will review the sizing with you.

Optional. The tool and every export work without this. No spam.

Was this tool useful?

How the estimate works

In LN each logged-on user session runs in a bshell process on the application server, and background jobs run in their own bshells. The calculator multiplies named users by peak concurrency and by bshells per user, adds job bshells, then converts that peak count into memory (count times measured MB per bshell, plus OS and services overhead, plus headroom) and into cores (count divided by how many bshells one core carries, plus headroom, rounded up). Nothing here is an Infor formula: it is arithmetic over numbers you measure. Infor publishes performance and tuning guidance for LN on docs.infor.com; check it for your release and use any figure it gives in place of these defaults.

  • Peak bshells = users x concurrency x bshells per user + job bshells
  • RAM GB = (peak x MB per bshell / 1024 + overhead GB) x (1 + headroom)
  • Cores = ceil(peak / bshells per core x (1 + headroom))

How to measure bshell memory and concurrency

On Unix or Linux, list the bshell processes at the busiest hour with ps and their resident size (RSS), then average them; repeat over several days, because bshells grow while users open sessions and run large reports. On Windows use Task Manager or a performance counter on the bshell processes. Count how many bshells exist at peak: that count divided by named users gives your real concurrency times bshells per user. Count job daemon bshells separately. Shared memory and the database client libraries are counted once, in overhead, so do not double count them in the per-bshell figure.

Worked example

With 250 named users, 60 percent peak concurrency and 1.2 bshells per user you get 180 interactive bshells, plus 10 job bshells, so 190 at peak. At 60 MB each that is about 11.1 GB, plus 8 GB overhead, times 1.3 headroom, so about 24.8 GB of RAM. At 12 bshells per core, 190 bshells need 15.8 cores, times 1.3 is 20.6, rounded up to 21 cores. If you move to 400 users the chart shows RAM and cores rising in a straight line, which is why measured concurrency matters more than the named-user count.

Session timeouts and idle bshells

Idle bshells still hold memory. If users leave LN open all day, peak concurrency approaches 100 percent of logged-in users even when few are working. An idle session timeout reduces that, but set it with the business: a timeout that ends sessions during long data entry costs more than the RAM it saves. Before you add memory, check how many bshells at peak belong to users who have not done anything for an hour; the answer often changes the sizing more than any hardware option.

Frequently Asked Questions

How many active users can an Infor LN application server handle?

There is no fixed number. It depends on how many bshells each user creates, how much memory each bshell holds, and how heavy the sessions are. Measure bshell count and resident memory at your busiest hour, then use this calculator to translate that into RAM and cores for the user count you plan to support.

What is a bshell in the LN session architecture?

The bshell is the LN runtime process on the application server that executes sessions and 4GL or 3GL logic for a user or a job. Clients such as LN UI connect to a bshell, which talks to the database through the database driver. More open bshells means more server memory, so bshell count is the main sizing driver.

Does the session timeout setting reduce server memory?

It can, because idle sessions keep their bshell alive and hold memory. Ending idle sessions after a defined period lowers the peak bshell count. Agree the timeout with users first, and measure the bshell count before and after so you size from the real effect rather than an expectation.

Is 60 MB per bshell an Infor number?

No. It is an editable starting assumption. Real values vary with LN release, sessions used, report sizes and customisations. Replace it with the average resident size you measure on your own server at peak. If Infor guidance for your release gives a figure, use that instead.

If you want your LN users to ask questions about orders, stock and jobs in plain English on your own servers, we can add a private AI layer next to this LN system.

ArchRay builder
Design the on-prem AI stack for Infor LN and Baan: model, GPUs, RAG pipeline, guardrails
See memory, speed and cost as you swap parts, then export the diagram. Free, in your browser.
Open the builder

Keep exploring

Free tools

ERP fixes and how-tos

Guides

Integrations