Search

link to homepage

Institute for Advanced Simulation (IAS)

Navigation and service


Federations

UNICORE Evolution

UNICORE is a software suite for building federated systems, providing secure and seamless access to heterogeneous resources such as compute clusters and file systems. UNICORE is deployed in HPC-oriented infrastructures such as HBP or PRACE.

Rich Client

UNICORE is open source software and is under continuous development and maintenance, led by FSD in collaboration with our national and international partner institutions.

The guiding principles followed in this development process are

  • open source, BSD licensed software: http://sourceforge.net/projects/unicore
  • platform independent implementations (Java, Python)
  • use of industry standard, open-source software and adoption of open standards (OASIS, W3C, OGF)
  • effective quality assurance (strong focus on unit testing, code coverage, and continuous integration)

Specifically, research and development in UNICORE focuses on several topics.

Web based and mobile access

In contrast to client applications that have to be installed on end-user machines, web based access has a lower barrier of entry and easier upgrades, as well as a more user friendly and simpler way to access distributed resources.

UNICORE Portal

The most important area of work is the development of lightweight RESTful APIs, which greatly simplify the development of web clients based on HTML5 and JavaScript as well as the integration into existing applications such as JupyterHub. These activities are tightly integrated with the work on federated identity management. UNICORE’s own web portal will be further maintained and evolved.

Alternative data processing models

Due to its tradition as a HPC-oriented middleware, UNICORE is built around the concept of a job that is executed on a cluster controlled by a batch system. For some applications this is less than ideal. Thus, we strive to extend UNICORE towards other data processing models such as data pipelines, interactive supercomputing or stream processing. This entails the seamless integration of diverse storage technologies such as file systems, SQL and NoSQL databases, object stores, cloud storage, focusing on scalable storage technologies, persistent identifiers and optimized data access.

Scalability and performance

In order to cope with the increasing demands of high-throughput, high-volume applications that are common in Big Data analytics applications, we expect further evolution and modifications to UNICORE. One example would be the provisioning of more lightweight interfaces (e.g. RESTful interfaces, pub/sub messaging, notifications etc), in order to further reduce the load on servers and clients and enable larger-scale applications.

Access to more types of resources

Currently, UNICORE excels at providing seamless and secure access to HPC clusters and filesystems. To widen the scope, we are investigating further types of resources such as SQL/NoSQL data bases or virtualised resources.

 

Federated identity management

A generic topic in today's federated systems is the difficulty of providing a unified, consistent user authentication, authorisation and identity management. The basic requirement is to overcome “silo” solutions (e.g. those purely based on X.509 or SAML) and to cover the full range of resources integrated by a software such as UNICORE.


Servicemeu

Homepage