dbiir/pard

A Parallel Database Running like a Leopard

pard's Issues

Test

Run tests against pard.
Detect bugs.

Refactor with Netty

Use Netty to refactor use of ByteBuffer and Thread executions.

Two phase commit

Drop

Drop schema
Drop table

Executor

Execute local tasks.

Jetty throws exception when quit process

java.lang.NoClassDefFoundError: org/apache/log4j/spi/ThrowableInformation

垂直分区查询计划下的选择下推 + 从选择中提取Join条件

Show plan and execution

Show schemas.
Show tables.

Unified and efficient communication and data transfer framework

Use Netty as the weapon to develop a unified and efficient framework for all kinds of messages and data transfer.

Remove unnecessary JS code

Remove unnecessary javascript code.
OR compress javascript code to less lines.

Replace `null` with `Optional`

Optional is more expressive than null.

Load plan

Load statement.
Load data from local file system.

Load execution

Basic Join

Join is separated into three stages:

1. Stage0 -- JoinPrepare
   In this stage, the coordinator sends tasks to query local data and then the data are shuffled to corresponding servers, and inserted into local database. Once the coordinator collects all success status, then next stage is triggered.
2. Stage1 -- JoinExecute
   In this stage, each worker executes join locally. And results are collected by the coordinator.
3. Stage2 -- JoinFinish
   In this stage, the coordinator finished following operators, and return results to user. Finally, send DropTask to each worker to drop temp tables.

Add support for order by and limit

Exchange framework

The exchange framework is already done.
Need more tests.

node keeper

Use rpc to detect server status change.

Simple implementation design:
Use a scheduled execution thread to send heartbeat rpc, and collects server status.

Multi-phases join and Join DAG

Code Review

We need do a full review of current codebase to eliminate potential bug.
This is important and urgent.
New features are not supposed to add before code review is done.

Final report

课程报告撰写

Query optimization

query optimisations.

`@`: select data from specified site

Fix bug for queries like select * from emp@pard0;

Format client output

Delete plan and execution

Delete statement

Vertical partition

Support vertical partitions.

Enhance the web view

Make the web page more powerful from that can see the job id, status, sql statement and other things.

Test with a new version

Test current codebase after refactoring with Netty and etc.

This new version supports CREATE, INSERT, LOAD, SHOW, DROP. SELECT(simple)
This new version does not support DELETE, JOIN

Delete parser

Buffer Manager

Show parser

Load parser