Affordable Large RAM Subsystem July 2021

HPC

Big Memory - A New Computing Paradigm

"Ideally one would desire an indefinitely large memory capacity...It does not seem possible physically to achieve such a capacity. We are therefore forced to recognize the possiblity of constructing a hierarchy of memories." - John von Neumann

In recent years, data is growing at a larger and faster rate. While volatile memory is fast, it is also expensive and limited in storage. Traditional storage model cannot keep up with that growth as large data must be placed in the disk and retrieved when needed. A Big Memory Server is one with RAM subsystem ranging from 6-24 TB, larger than that available on contemporary systems. They enable many data-intensive applications to run in-memory, such as low-latency trading, ML inference, large exhaustion-by-proof enumeration etc... Modern approach such as Intel Optane is closing the gap in the memory hierarchy and solutions exist that are implemented on enterprise level. In our work, we propose an alternative solution using NAND-flash SSD. Our goal is to create HPC system for analytics of scale that would otherwise not be affordable for small datacenters. See our presentation in the upcoming IEEE HPEC conference in September.

Paper

@MIT-CRIBB


GPU-accelerated Supercomputer June 2021

HPC

Our private 10 CPU + 4 GPU nodes computing cluster

A project by the BUHPC club, a 10-node Raspberry Pi + 4-node Nvidia Jetson Nano Cluster, networked over Ethernet with master node login. Access granted to club members upon request, we use it to experiment new ideas and distributed programs. Supports OpenMPI 3.0, AMD AOCL, CUDA 11, GCC

Source Code

A picture I took while setting up the hardwares

cluster

Competitive Programming Library May 2021

Algorithm C++

My setup and custom algorithm for competetive programming.

The repo includes various tricks and techniques I've collected over the years from difference sources as well as my own modifcation. It also has my vim configuration. I will continue to update the library, currently there are several unpushed pieces of algorithms and they need to be tested first. This also serves as a reference for team practices as well as for my students. They are intended to be straight out of the box for easy copy and pasting during contest.

Source Code


INTLSnip September 2020

Algorithm Artifical Intelligence

A snipping tool that resizes the image intelligently using content aware scaling

I got the inspirtation while getting training data. I wanted a way to avoid squishing image and still preserve the features. The tool is implemented using the algorithm described in this paper

Source Code

Before carved

The idea behind is really simple, we convolve against the image with a sobel filter to create mapping of weights, then use dp to find best path.

At 50% reduction, the size is cut in half but the features remains.

Resized image

XRAY July 2020

Deep Learning Artifical Intelligence

Analyzing CT scans of poential covid patients

I've always been quite curious about the application of deep learning in the medical field. This is my attempt at building a classifier for identifying medical images of Covid patients

I do not claim any diagonstic performance of the model. I was able to get 100% accuracy on the testing data. However this was likely due to a small dataset. This is merely a proof of concept inspired by this paper

Source Code


EasySchedule December 2019

Java

A planner for scheduling my classes because of the tedious gen ed credits

The GUI is built in JavaFX, feel free to make any modification as needed. The data is written into CSV for parsing, you will probably need to add more to it yourself if you're not majoring in CS. Just fill the parameters. Funny enough, right after I built the planner my friend told me they introduced a system that calculates it for you.

Source Code

Resized image