Professional Documents
Culture Documents
July Release
Live Server Performance Enhancement
● Low Memory Save
The save process is a high memory intensive process depending on the
data it works upon and if not run at regular intervals it can bloat the
tenant memory. Currently, when a save is triggered, merging of base
table (in-memory representation of on-disk file) and delta container (in-
memory new edits which were accumulated since the last save) happens
inside memory on multiple threads, simultaneously. This uses a lot of
process memory in a very short time span resulting on memory spike.
In the coming release the save behaviour is changed when the save is run
at high memory conditions
● When tenant memory is between 70 - 85% of allocated memory - the save
process will flush in memory tables (these are loaded into memory due to user
reads/writes) to disk at the start of the operation.
By this enhancement external systems can directly ingest the data into the API
endpoint after the NIFI provisioning.
Python 3.10 Template
In the current enhancement, Python 3.10 and the supporting packages are
introduced.
A Python environment is required to run and execute the python plugins. The o9
Platform offers
A two python templates as standard templates:
● 2020_10: Python 3.6 and the Supporting Packages
● 2021_10: Python 3.7 and the Supporting Packages
In the current enhancement, a new template 2023_10 with Python 3.10 and the
supporting
packages are introduced.
o9 Predict.ai
Improving accuracy of forecast is a key requirement for Demand Planning
applications that o9Predict.ai is addressing.
The existing ensemble model picks one winner per intersection. As part of this
enhancement, the plugin will use validation dataset actuals and forecasts from all
algorithms. Each algorithm's forecast output is used as drivers in a regression
model whose output is the actual. The weights are then stored for all future
forecast mode runs.
Thank You!!