Dealing with large STDF data sets
Former test engineer and yieldHUB Founder and CEO John O’Donnell reflects on the challenges of analyzing large STDF datasets and how modern, server-powered yield management tools now enable engineers to process gigabytes of data in seconds, saving time, improving insight and transforming engineering productivity.
Blog by Founder & CEO John O'Donnell
Earlier in my career, I worked as a test engineer at a large semiconductor company. One of the biggest challenges at the time was analyzing tens, sometimes hundreds, of megabytes of STDF data to understand product behavior and identify opportunities to improve yield.
Engineers essentially had two choices: spreadsheets or desktop data tools. Spreadsheets quickly became impractical due to their limitations at the time they could only handle about 65,000 rows, which meant meaningful STDF analysis was nearly impossible.
The desktop data tool we had available ran only on a specific operating system and required one of a small number of designated machines. You often had to wait your turn just to load data, and even then, the limited RAM on those computers restricted how much STDF you could analyze.
Fast-forward to today, and our customers routinely analyze several gigabytes of data on their laptops, and sometimes even on their mobile devices, through yieldHUB. It’s remarkable to see how quickly charts and tables appear, even when working with massive datasets. This is made possible because most of the heavy processing happens on the server side, allowing every user to tap into enterprise-level computing power.
The great thing about yieldHUB is that most of the processing is done on the server so each user can tap into the huge computing resource available on the server. We have one customer that uses a server with 128GB RAM, 5T storage and 24 CPU cores. This serious computing power combined with yieldHUB’s data management capability allows users to quickly search and group STDF data (and other data formats supported) and generate reports in seconds.
Even when analyzing hundreds of wafers, reports typically generate within just a few minutes. This is significantly faster than attempting to upload the same volume of STDF datalogs into a desktop tool, assuming the tool could even handle it.
Engineers save enormous amounts of time because they no longer need to hunt for data or manually copy files to their local machines. As soon as a datalog is processed by the yieldHUB server, it becomes instantly searchable and ready for analysis by any authorized user.
Although I’m no longer a test engineer myself, yieldHUB helps today’s engineers accomplish more in less time. With yieldHUB, they can focus on engineering and not on waiting for tools, juggling files or wrestling with system limitations.
