Repository management service GitHub has added to the shoals of the open source ecosystem with a “bit of a peculiar OSS release”. The Microsoft subsidiary’s new generally available orchestration service is called skeefree and aims to “achieve automated schema migration flow at GitHub”.
It was developed as a consequence of the many manual steps required to make new features live, which is in no small part due to the organically grown system you often find in more complex products. The number of changes to the database schema needed to update code had over the years “amounted to a significant toil [sic] on the database infrastructure team”.
Skeefree helps with all of that by acting as an intermediary between components such as GitHub Actions and online schema migration tool gh-ost to automate the process. While this sounds all very good, the project doesn’t seem to be useful for more than inspiration to outsiders, given they don’t have access to the company’s internal libraries and services the project depends on. It also expects some form of ChatOps to be used and needs to be heavily adapted if used in a different environment.
IBM’s AI team starts taking events into account
IBM research staff members and their Watson colleagues have come up with a new dynamic, probabilistic graphical model to make sure the influence of external events is taken into account when using machines for decision making. The new approach is called event-driven continuous time bayesian networks (ECTBN) and addresses complex scenarios with inter-dependent variables which can be found in areas such as health and finance.
As the name suggests, ECTBNs combine the strengths of continuous time bayesian networks with those of graphical event models. This allows them to “capture the effect of historical arrivals of different event labels on the state variables of a system” thus providing more context for decision making. An example for an event could be a natural disaster, which for instance influences the price of a certain stock.
ECTBN Graphs include event labels, system variables, and model parameters to represent dynamic processes. While event labels can be part of cycles or self-loops, the approach assumes they can’t be influenced by state variables. A detailed example can be found in a blog post introducing the technique.
Healthcare focused cloud gets on the K8s train
Managed cloud provider Cloudicity has added a managed Kubernetes offering to its healthcare-focused portfolio. The service is based on Amazon’s EKS and is supposed to help IT departments with creating and monitoring container clusters in a HIPAA compliant manner. The latter is meant to be ensured by use of encryption, hardened images, and automatic remediation amongst other things.