Machine Learning and Neuron Nerworks models - Before ONNX
ONNX, short for Open Neural Netwrok Exchange, is an open standard for machine learning (ML) models. The question must be asked before we move forward, why yet another ML model standard? One can already name a handful of famous or infamous examples of ...
SIGMOD 2019 Programming Contest - Parallel Radix Sort
This post intends to make a note of how Zhaoxing and I competed in the programming contest of ACM SIGMOD conference 2019, which is one of the top and premium venues for database-related publications. The contest is first hosted at SIGMOD 2009 and is made annual afterwards. We made the finalist...
Benchmark Cassandra & CockroachDB on multi-node cluster with Cassandra-stress tool/YCSB
Build Wordpress, Telegram, K9 Mails and ownCloud on Android 8.1.0
I am doing some research on the workloads of SQLite, the de facto database for Android, on mobile devices. And at the beginning, I wanted to write a wrapper for SQLiteDatabase.java and trace all relevant transactions. The amount of code refactoring is too huge for me to finish on my own. Then ...
Build AOSP 8.1.0 on Nexus 6p - An Incomplete Guide
Setup YCSB & MapKeeper to benchmark LevelDB
LevelDB gets its own benchmark binary called “db_bench” under the directory “out-shared” after make. But there are other tools could benchmark LevelDB like MapKeeper. YCSB is a Yahoo! cloud service benchmark without the support for LevelDB. It is a possibility to constitute YCSB into MapKeeper...
How Sorted String Table Works and LevelDB Performs one LookUp
The very basic idea behind LevelDB and BigTable of Google is Log-Structure Merge tree and Sorted String Table. Generally LSM tree use a log to trace all index hence it has better read performance while trade off write performance. From bLSM (reference spot), we can easily spot the difference o...