Photonics and light-based computing: series brief
The Computer Weekly Developer Network (CWDN) now embarks upon a series of guest analysis pieces covering the world of photonics.
The flow of network packets via photonics light-based networks provides a rich dataset for training machine learning and predictive analytics, now is a perfect time to learn more about this still-nascent technology.
In this series of articles, we explore current trends in the use of AI-enabled tools that use advanced modelling to help network administrators manage networks more effectively.
But what do we need to cover in the world of photonics and optical computing?
This time, CWDN will look at everything stemming from what is happening at NTT (a renowned big player in this space) and the other players that work in the photonics market.
Key vendors operating in this arena include Intel (silicon photonics and co-packaged optical I/O chiplets); Ayar Labs; Lightmatter (building photonic AI acceleration and photonic interconnect technology); Luminous Computing (a startup focused on photonic processors for AI workloads); Infinera (specialist in high-capacity optical networking and photonic integrated engines for terabit-scale transport); plus Cisco and also Ciena.
From electrons to photons.
We are also interested in examining the foundational physics underpinning photonic integrated circuits (PICs), after all, now could be the time to question exactly how Moore’s Law fatigue is helping to push the shift and evolution from electrons to photons.
We are also open to commentary on the emergence of the co-packaged optics and the way they work inside hyperscale data centre architecture. We are also looking for some honesty around where silicon photonics actually delivers in the modern incumbent enterprise software IT stack… and indeed where it either struggles or represents too much refactoring of the installed base of software applications and data services to make it useful.
CWDN is interested in looking into the programmability of the optical layer. Why? Because if software developers are now able to offload the heavy lifting of matrix multiplication to light-based accelerators, what does that mean for the average software engineer’s API calls?
Disaggregated network models
With an eye on how standards bodies are helping to manage and oversee interoperability in the area of photonics, this is also an opportunity to look at how newly disaggregated network models work and to what extent they can run with much lower-latency and lower power consumption.
The big question is, is photonics the dawn of a major new level of computational power that we can productively use at the coalface of enterprise software, or is it just faster plumbing?
Image; Google Gemini