Page not found. Your pixels are in another canvas.
A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.
This is a page not in th emain menu
Service invocation is tricky. The protocol itself seems to be super simple: we call a function, and it returns. But the problem is, when the function is on another machine, how we know the state of that machine / the whole system. The developers of the caller do not want to care, so we protocol designers should care for them. But trying to make everyone happy is so tough here.
There is always a need for service discovery and load balancing, but no universal solution. Sometimes naive ways may satisfy most needs.
I find that some terms are confusing.
TL; DR: Reflection should be used for encoding or serilization.
When solving a problem, it is common that we have two different strategies that fit in different cases. For example, one algorithm may have a better time complexity but uses more memory than the other. Or, one is fast when there are only a few of large objects, but the other works better when there are many small objects. However, I found that there usually exists a solution which is a naive mixture of the two strategies, and its performance will be the sqrt of the two.
This is a course project that reads the number indicated by a vehicle instrument. We use Kirsch and Sobel operators to sharpen the image, Otsu algorithm for thresholding, and RANSEC to detect the scale marks and the hand. We calculate the result based on the angle.
Threshold is widely used in RoboCup Middle Size League (MSL) for ball recognization. We proposed an algorithm to find the best threshold without manually tuning arguments. This is submitted to Science Challenge in RoboCup China Open 2016.
This is a course project for CS 251B. We modified gem5 simulator to support a subset of RISC-V vector extensions draft in system-call emulation mode. We also wrote some testing assembly & C code which can be run on a bare metal machine for evaluation.
NFD Configuration Manager (NDN-FM) provides a graphical interface to increase the ease with which users can manage local instances of the NDN Forwarding Daemon (NFD) and other packet forwarders implementing the NFD Management Protocol: NDN-FM can monitor the status of the forwarder; create, update, and delete faces and routes; manage certificates; and run basic NDN debugging tools.
Pull Request User Guide KEP(TBA)
Kubernetes has a well-rounded ecosystem for pods in the cluster. But it does not provide a standard interface for pods to interact with external workloads, such as VMs. This project proposed ExternalWorkload custom resource to represent workloads, provides pod to workload load balancing, and supports mixed backends for Services based on EndpointSlices. I have implemented a PoC demo, which is merged into GCE Ingress Controller. The KEP is being reviewed.
I gave several tutorials to other students about basic algorithms. You can find the slides (Chinese) on the GitHub repo.