Both machines ran the full-stack application we developed without issues. We also wanted the room to grow our data store (currently around 45TB). Its 10Gbit Ethernet was important for storing and retrieving data locally. It has four drive bays, which can be limiting, but it's expandable up to nine.įor production, we used a DiskStation DS1618+ ($1,999 at Amazon). The DS923+ offered us a nice combination of a good price ($599 at Amazon, without disks) and plenty of power for the job, with a dual-core AMD Ryzen R1600 CPU and 4GB of RAM installed (while supporting a maximum of 32GB). By using servers with the same software and CPU architecture, moving from test to production was trivial.įor our development and test system, we used a Synology DiskStation DS923+. We also wanted one based on an x86 chip so we could run pre-compiled binaries.īecause our main site would need to be public-facing, we decided to use two Synology units-one for development, test, and early rollout, and another for final production. Our server needed a box that could support all the applications and tools we needed (including Docker, MongoDB, Node.js, NFS, and WebDAV, for starters). Our Servers: Synology DiskStation DS923+ and DS1618+Īs reasonably priced NAS units have evolved, I've been keeping an eye on their compute power. As part of the research we plan to publish this year, we wanted to make that data and results available and usable to other researchers via a web interface. Over time, we've accumulated tens of thousands of scenes, and hundreds of thousands of images of those scenes, rendered by various combinations of lenses, sensors, and capture settings. Using our toolset, we can research and collaborate with other universities and corporate research groups to help evaluate which sensors and lenses are the most effective for training autonomous driving and car safety systems. It includes capabilities for generating real and synthetic scenes, along with a physically realistic, hyperspectral pipeline of accurately modeled lighting, lenses, and sensors through which we can render them. Our lab at Stanford ( ) builds a family of image and camera simulation tools called ISET (an Image Systems Engineering Toolkit). Credit: David Cardinal Our Project: Putting Stanford's ISET Tools on the Web
0 Comments
Leave a Reply. |