Before I delve into the types, starting with time types, I want to say a little more about my working principle. It's strictly prototyping oriented...the bottom up fashion.
When I was inspired to build The Innovation Mesh for innovations in quantitative fields I built a rough scheme and backtested it by walking UnRisk through it.
Then I decided to disclose the ideas and a scheme. This is what I'll do in the near future. But it's still a prototype and it will be refined, when not reinvented with growing practical experience. I plan to provide releases.
Remember, the types shall enable the innovator to find a short description of what her innovation does and how it is positioned. Something that wraps the mind of investors, actors, clients, partners, marketers…Types are a way of cataloguing innovations related to the expectation of actors (those who use, enrich, integrate, market,...it)
The Innovation Mesh is a diagnosis tool. Although it is for quant fields it's not quantified, but it suggests some conventions, rules, workflows and even licensing schemes.
Time Types
First I thought it shall be session related, but then I decided for runtime. How fast will the innovation work related to the reality and how well is it synchronized with it.
High Performance Computing - mostly compact solutions or tools needed to beat real time significantly to, say, early anticipate anomalies in a real behavior, or to be applied in scenarios...
Real-Time Computing - these are usually built to control real behavior (synchronously). This systems are often embedded into products, processes, trading environments…they're online.
Off-Time Computing - theses are usually working offline to real systems supporting complex workflows and tasks, by analysis, predictive modeling, simulation, optimization…
Time types influence other types more than we may think now.
In a quick view Engineering is characterized by definition-and-verification and management by plan-and-control. This suggests that engineering tasks are usually offline, whilst management usually needs systems that are linked to real systems.
Production engineering is about building the right factory for the products to be made, whilst production management is about controlling its parts through the factory.
In the details engineering and management tools may need fast computing in their internal engines, but the overall time types are different.
Fast computing is often used millions of time to value one object. It needs to be blazingly fast and accurate…think of the valuation of a portfolio of derivatives across scenarios.
Another view into systems is their purpose: analysis, prediction or control. Control systems usually have real-time requirements, whilst analysis systems are usually off line. You gather information in order to extract (computational) models…if the data acquisition is semi online, you have real-time requirements…
These things are important if you apply an intelligent combination of, say, modeling and machine learning techniques….but more in an internal view. I'll come back to this in future posts.
However, with an unclear assessment of these types you may run into the risky horror of operational risk.
What's Real-Time?
In earlier time of computer science it was assumed that real-time computing needs real-time, operation systems, real-time networks and real-time programming languages (in short everything had time stamps - it had a scheduling policy with priorities, synchronized request and suspends...).
In the meantime computing became so fast that I simplify to "meet all required deadlines". But, real-time computing is not the same as high performance computing. Meet the deadlines needs to consider loads.
In a real-time system, such as the FTSE 100 Index it became horrible if the date came delayed for its applications.
Consequently real-time systems are often stand alone.
Real-time periods to deadlines may be very long (climate) or medium (a blast furnace process) or very short (an ABS system).
Simulation is the imitation of a real-world behavior. We use it to gain insight. But also, when the observation of the real behavior took too long, the real system cannot be engaged…In any case it's beneficial if the simulation time is below the real time. From their character they're still slow (offline).
Reality Types? That's what comes up next.