Nicolas Nicolaou was able to test his AM-NVE solution for distributed systems with technologies used for Virtual Reality and Augmented Reality applications with Fed4Fire+ the largest worldwide federation of experimental Internet facilities supporting the Next Generation Internet.
Tell us about yourself
I am from the Eastern most member of the EU, the beautiful island of Cyprus and I hold a PhD in Computer Science and Engineering since 2011, from the University of Connecticut (UCONN) in the USA with an expertise in the areas of Distributed Storage Systems, Blockchains, and Security of Embedded Systems and Critical Infrastructures.
My fascination around Distributed Systems started when I joined the team of Prof. Schwartzman at UCONN, and when started investigating very interesting problems in how distinct, loosely connected devices may cooperate and collaborate for accomplishing efficiently complex tasks.
Since 2011 I worked in various academic positions before taking the decision to depart to an industrial career and establish my own SME in 2019, Algolysis Ltd, a company devising algorithmic solutions for real-life problems with a particular focus on devising novel distributed techniques that can power the next generation of data storage, analytics, and visualisation systems.
Describe your project
Our project was entitled Atomic Memory for Network Virtual Environments (AMNVE) and received funding from the Innovative Experiments Category “SME & NGI Cascaded Experiments” (F4Fp-SME_NGI), with the experiment title F4Fp – SME – AMNVE.
The project AMNVE is an attempt to combine harmonically, on a single innovative product, technologies proposed for the distributed systems with technologies used for VR/AR applications. This could lead to an ultimate flag product for Algolysis as the company fosters expertise both in distributed systems as well as in computer graphics and AR/VR.
What is innovative about your project?
Network Virtual Environments (NVEs) are used to allow users to emerge in common virtual worlds, from different places in the globe. The biggest challenge in NVEs is to ensure that the objects in the world are shared efficiently between the participants, providing the illusion of the real-time interaction to the world residents. A major question that arises is the following: How can we ensure that the users retrieve a consistent view of the world despite their concurrent interaction with objects in that world?
Algorithms making the NVEs of today possible often rely on single server/centralised architectures that inherently impose participation limits, consist of a single point of failure and performance bottleneck, and rely on a fault-prone network infrastructure. Distributed Shared Memory (DSM) implementations offer availability of strongly consistent shared data spaces on top of distinct, globally distributed devices, despite of failures and asynchrony. Initial attempts to implement DSM, required high communication demands. In contrast to initial attempts, recent DSM solutions managed to shift the viewpoint around the practicality of strongly consistent DSM, offering fast atomic operations.
What did the NGI/FED4FIRE+ opportunity enable you to do?
With the support from Fed4FIRE+, we set to investigate the possibility of allowing NVE applications to utilise fast DSM solutions as a Memory-as-a-Service that will: (i) allow consistent manipulation of shared objects, (ii) preserve liveness even in cases of node failures, and (iii) offer an intuitive and transparent interface to NVE developers to store and retrieve consistent, networked, shared data despite concurrency, node failures, and perturbations in the underlying communication channel.
Within the project we utilised the Grid5000 testbed which provided the perfect playground to deploy and test our distributed system. The outcomes of the project suggest that the technology offers promising capabilities, but it is not fully ready to be deployed widely for time sensitive applications (like VR environments). While we have demonstrated that DSM as a technology is a technically feasible alternative to existing solutions, the overall operation latencies observed in these experiments suggest that for the purposes of Networked Virtual Environments further optimisations may be necessary.
Although the technology itself looks promising especially when deployed on powerful machines, latencies in our current implementation should be decreased before it can be declared as a viable alternative to existing methods. It must be pointed out that the benefits of providing fault-tolerance and atomic read/write operations are still of extreme importance for next generation collaborative applications and may be useful in domains where high throughput real-time updates are not necessary.
What are your next steps?
For Algolysis, the project results are of great importance as they demonstrated the maturity level of the technology before full commercialisation, which will require additional market analysis and validation, before financial feasibility and marketability is clarified. However, it is important to point out that the particular algorithms address problems frequently faced in distributed systems by all major Virtual Environment producers, such as the multi-billion video games industry. Therefore, this project is an attractive and potentially high-yield endeavour. Additionally, it has been of crucial importance that we were able to devise reusable experimental software (that can be further improved) for the efficient deployment of the 3D interactive NVE.
More information on Nicolas Nicolaou
- LinkedIn: https://www.linkedin.com/in/nicolaoun/?originalSubdomain=cy
- Google Scholar: https://scholar.google.com/citations?user=OBrnuTUAAAAJ&hl=en
- Nicolas is Co-Founder and Senior Research Engineer at Algolysis Ltd. algolysis.com