Early in my career as a process engineer working with sensor data and trying to use it for controls (#FaultDetection / FDC), I was told “Don’t pull too many process runs at once, it will lock up the system” and “Why do you even need the data?” At every company, everywhere I’ve been, the story has been the same. I’ve even told it myself.
The fact remains that waiting for data to show up is an invisible spend at almost all semiconductor factories that totals in the millions ($USD) for a normal, medium sized factory every single year. The opportunity costs are even higher. With AI analytics providing real value in semiconductor manufacturing in the last few years, quickly getting to and using equipment data is proving to be crucial to maintaining competitive advantage.
The great news is that we don’t have pay this price anymore. The technology was the limiting factor ever since the early days of FDC. An enterprising semiconductor IT executive came across a technology, which is where I first heard of it, called kdb+ by Kx. Leave it to the finance world to fund development of super-high speed DB technology with high availability and data quality. Check out the ridiculous benchmarking numbers, where using their tech on a Raspberry PI server that is still 4x faster than MongoDB on enterprise level hardware. On AWS, were talking order of magnitude. Trust me, if there were a meaningful competitor out there I’d mention them as well, for now I’m just thrilled there’s at least one.
For reference, the 4-core configuration was said to match the equivalent config for InfluxDB and MongoDB
Think about not having to wait for your data anymore. If you’ve ever been frustrated by slow data and slow yield learning, if you want to save your company millions in waste, if you want to make quality of life better for your teams, all it takes is getting the right people in place and making the decision to make the change.