What do you remember from the summer of 1975?
Racing fan? You probably remember that Bobby Unser won the Indy 500. (It was a bit short due to rain.) Interested in space exploration? You probably remember the Soyuz 19 and Apollo docking or the launch of Viking 1.
Maybe you remember long lines for gasoline in the U.S., the release of a famous movie about a huge shark, or mood rings and pet rocks being all the rage. I remember most of those events, too.
What I remember most is being diagnosed with type 1 (insulin-dependent) diabetes. I was just a kid. Did that diagnosis change my life? Yes, but at the time I couldn’t understand just how much. I also couldn’t understand why my parents were distraught and quiet on the ride home from the doctor’s office.
Type 1 diabetes treatment then had limited options with little progress
My older brother had been diagnosed five years earlier, so I had a clue about managing type 1 diabetes. His management method — which would become my management method as well — included measuring sugar and ketones with reagent strips; adhering to a tightly fixed, highly regulated diet; and giving myself regular injections with animal-derived insulin. At that time, we relied on the reagent strips because blood sugar monitors were usually found only at physicians’ offices, and then not very often. All this was standard treatment, one size fits most. It was the best we had.
Within just a few years of 1975, managing type 1 diabetes got a bit easier. In addition to animal-derived insulins, doctors had the option of human and human-analog insulin. We could migrate from sugar/ketone measuring reagent strips to direct, nearly instant blood sugar monitoring. I got my first home-use blood glucose monitor then. It was a shoebox-sized contraption requiring multiple steps for a fairly accurate measurement. But it worked, and you could measure your blood sugar right then and do something to help manage it.
Type 1 diabetes treatment now includes embedded, continuous monitoring, mobile phone apps and pumps
Taking a leap ahead to today, management options have advanced to the point where my 1975 self (and I’m guessing many physicians) would have no idea what you were talking about. Continuous blood sugar monitoring that you can spot-check any time, nearly anywhere with your mobile phone? Check. We’ve got that. A portable, wearable device that automatically administers insulin — on a physician’s prescribed schedule and when needed? Check. We’ve got that, too.
As amazing as current type 1 diabetes treatment is, it doesn’t get to the root of the problem. Why do people become diabetic?
Lots of type 1 diabetics are here, and more are coming
According to Beyond Type 1, 1.25 million Americans have type 1 diabetes, and another 40,000 are diagnosed every year. And that’s just the Unites States! Why is that? Why are so many new diabetics diagnosed each year, and for those of us already diagnosed, how might we better manage our condition?
We’ll come back to those questions in just a bit. First, let’s take a look at one possible path forward: precision medicine.
What is precision medicine?
Precision medicine — the matching of individual patients with targeted treatments that work fast — may hold some answers.
At Micron Insight 2018, Executive Vice President and Chief Business Officer Sumit Sadana described many health care challenges as a big data problem. He said that, when you talk about real, tough challenges like human health, human genomics and more, you are talking about exabytes of data. He also noted that — due to the massive scale of data that needs to be analyzed — solving these problems requires artificial intelligence and machine learning.
Micron QLC SSDs are a step in the right direction
At Micron, our vision is to transform how the world uses information to enrich life. Our focus on memory and storage technology innovation can change the fundamentals of how research is done. One example is quad-level cell (QLC) NAND and its benefits on highly read-centric, research-type workloads.
We wanted to embrace fast-growing, read-centric workloads like machine learning at a reasonable price, so we designed our Micron 5210 ION SSD to meet the growing demand for cost-effective, performance storage. With the Micron 5210, we shipped the industry’s first SSD built on QLC NAND — the next evolution in storage technology that delivers fast capacity for less. This product is optimized for read-intensive and performance-sensitive workloads like the data lakes of unstructured information that feed machine learning and artificial intelligence algorithms.
Until Micron introduced the 5210, these read-intensive workloads were shackled to the slow performance of hard drive technology developed 50 years ago.
Experts wrestle essential learnings from immense datasets
To understand how our QLC 5210 ION could improve research workloads with very large datasets, we worked with Colfax International’s Research Team on a machine learning project. Headquartered in Silicon Valley, Colfax Research helps leverage new platforms to build value in computational innovation.
Colfax Research analyzed how the Micron 5210 SSD outperforms legacy (traditional) machine learning storage (7200 RPM HDDs) for lakes of artificial intelligence and machine learning data. It then published its findings in a white paper entitled “How New QLC SATA SSDs Deliver 8x Faster Machine Learning.” After looking at throughput and completion time of a TFRecord using a 7.68TB 5210 ION SSD and a 7200 RPM HDD, Colfax Research said:
“In our study, with the Micron 5210 ION SSD, a read-intensive transformation of an image dataset with the purpose of a TFRecord file creation was accelerated by about 8x compared to a similar-sized HDD. In a 100,000 image dataset at 23MB per image, our test HDD took 15.17 hours to resize the images and pack them into a single TFRecord file while the Micron 5210 ION SSD took only 1.88 hours to do the same task — 13 additional hours for the HDD to complete the same work.”
Colfax Research’s test results are shown below:
In this example, the high-resolution images (up to 100,000 of them in Colfax Research’s testing) had to be resized and appended to a TFRecord before analysis could begin. How much of a difference could Micron’s 5210 ION SSD make to speed the preprocessing and wrestle essential insights from the data more quickly?
In discussing why the Micron 5210 SSD was beneficial, Colfax Research shared its take on the value of the QLC SSD:
“For years, 7200 RPM hard disk drives (HDDs) have been the standard media on which machine learning (ML) training data sets have been stored. These traditional HDDs have been preferred due to their low cost and easy to adopt SATA interfaces. However, HDDs suffer from relatively slow throughput. Solid State Drives (SSDs) have been too expensive to justify their potential gain. Despite SSD cost / GB decreases, their overall cost has remained just out of reach for all but the most demanding (and expensive) ML platforms. This barrier is starting to fall with the introduction of the world’s first QLC SSD, the Micron® 5210 ION enterprise SATA SSD.”
The company went on to say:
“Faster completion times can have dramatic effects on overall project value, delivery and expense. When analyzing the potential benefits of the Micron 5210 ION, a cost per GB analysis at the time of acquisition may be short-sighted. One should consider the total cost of ownership, the value of keeping expensive CPU assets busy, and the potential reductions in yearly operating expenses related to power and cooling.”
Micron memory accelerates health care AI
I’m encouraged by ongoing advances in algorithmic optimizations in health care artificial intelligence. Micron memory is playing key roles to enhance the power of computer hardware, along with our fast, vast storage. But artificial intelligence’s deep neural networks put stress on the data center. While the data is being processed, it should reside in fast memory for efficient compute. One of our solutions, Micron 64GB 3200 MT/s RDIMM, at twice the density for the same configuration, is very valuable in training neural networks due to their iterative nature.
I get to do my (very small) part working at Micron
I’m proud to tell people that I work for Micron. At trade shows, technology conferences or parent/teacher conferences, I puff up a bit when I tell people that I work for Micron’s Storage Business Unit. I admit — it is a bit of a brag.
And not just because Micron makes cool stuff (we do!) but because we’re helping enable precision medicine. Micron is helping to fundamentally change how researchers and scientists move medical diagnostics and treatments forward, the very medicine that may (eventually) bring a cure for type 1 diabetes.
Early diagnosis and better management are solving the precision medicine puzzle with immense genomic data
Of the 1.25 million Americans with type 1 diabetes, I am one. And there are 40,000 more of us every year in the U.S. Looking back over my diabetes journey, I’ve seen some amazing changes. I’ve seen treatments move from animal-derived medication to human medication to automated delivery devices you wear. I’ve seen monitoring move from reagent strips to in-home glucose monitoring, to easily portable monitors and now to implanted, real-time sensors you can read with your phone.
What else might I see? With the rapid pace of human genome research and understanding, I can’t wait to find out. Earlier, more precise diagnosis? Possibly. More personalized treatment? Could be.
And maybe … just maybe … prevention.
We’ll have to see. Those same researchers, scientists and physicians have mountains of data to collect and analyze (an estimated 2,314 exabytes in 2020).
What am I looking forward to the most? A new treatment that restarts my pancreas so I don’t need to take insulin anymore? No. Complete integration of real-time glucose monitoring with automated, direct insulin deliver? No.
What I look forward to the most is my kids — or maybe their kids — talking about type 1 diabetes in the past tense, as a historical condition that’s not part of their lives. I look forward to type 1 diabetes being just a history lesson.
