ambrosex / adbs_lab

姓名：熊荣康

学号：SA21229005

简单的存储和缓冲区管理器

├── ./autorun.sh
├── ./build
│   └── ./build/linux
│       └── ./build/linux/x86_64
│           └── ./build/linux/x86_64/release
│               ├── ./build/linux/x86_64/release/adbs_lab
│               ├── ./build/linux/x86_64/release/adbs_lab_multi_thread
│               └── ./build/linux/x86_64/release/test.dbf
├── ./data
│   └── ./data/data-5w-50w-zipf.txt
├── ./doc
│   ├── ./doc/adbs-lab.pdf
│   └── ./doc/adbs-project-zh.md
├── ./include
│   ├── ./include/argparse
│   │   └── ./include/argparse/argparse.h
│   ├── ./include/BufferPoolManager.h
│   ├── ./include/ClockReplacer.h
│   ├── ./include/cmdline.h
│   ├── ./include/Common.h
│   ├── ./include/DataStorageManager.h
│   ├── ./include/LRUReplacer.h
│   ├── ./include/Page.h
│   └── ./include/Replacer.h
├── ./LICENSE
├── ./README.md
├── ./src
│   ├── ./src/BufferPoolManager.cpp
│   ├── ./src/ClockReplacer.cpp
│   ├── ./src/DataStorageManager.cpp
│   └── ./src/LRUReplacer.cpp
├── ./test
│   ├── ./test/main.cpp (single thread)
│   └── ./test/main_multi_thread.cpp    (multi thread)
└── ./xmake.lua

使用

mkdir build;
cd build;
xmake -P .;

实现结果(包括设计结果、实现结果、运行结果等）

run

single thread

cd build/linux/x86_64/release/;
./adbs_lab -l -f /home/rongkang/WorkSpace/HomeWork/adbs_lab/data/data-5w-50w-zipf.txt;

cd build/linux/x86_64/release/;
./adbs_lab -c -f /home/rongkang/WorkSpace/HomeWork/adbs_lab/data/data-5w-50w-zipf.txt;

multiple threads

10 threads lru

cd build/linux/x86_64/release/;
./adbs_lab_multi_thread -l -t 10 -f /home/rongkang/WorkSpace/HomeWork/adbs_lab/data/data-5w-50w-zipf.txt;

20 threads lru

cd build/linux/x86_64/release/;
./adbs_lab_multi_thread -l -t 20 -f /home/rongkang/WorkSpace/HomeWork/adbs_lab/data/data-5w-50w-zipf.txt;

30 threads lru

cd build/linux/x86_64/release/;
./adbs_lab_multi_thread -l -t 30 -f /home/rongkang/WorkSpace/HomeWork/adbs_lab/data/data-5w-50w-zipf.txt;

40 threads lru

cd build/linux/x86_64/release/;
./adbs_lab_multi_thread -l -t 40 -f /home/rongkang/WorkSpace/HomeWork/adbs_lab/data/data-5w-50w-zipf.txt;

10 threads clock

cd build/linux/x86_64/release/;
./adbs_lab_multi_thread -c -t 10 -f /home/rongkang/WorkSpace/HomeWork/adbs_lab/data/data-5w-50w-zipf.txt;

20 threads clock

cd build/linux/x86_64/release/;
./adbs_lab_multi_thread -c -t 20 -f /home/rongkang/WorkSpace/HomeWork/adbs_lab/data/data-5w-50w-zipf.txt;

30 threads clock

cd build/linux/x86_64/release/;
./adbs_lab_multi_thread -c -t 30 -f /home/rongkang/WorkSpace/HomeWork/adbs_lab/data/data-5w-50w-zipf.txt;

results

thread	replacer	hit number	hit rate	io number	time
1	lru	169565	33.91%	502821	0.95s
1	clock	169565	33.91%	502821	2.11s
10	lru	2007256	40.15%	4555873	51.81s
20	lru	4421657	44.22%	8489680	121.78s
30	lru	6704605	44.70%	12624313	205.05s
40	lru	8392585	41.96%	17667722	254.76s
10	clock	2127384	42.55%	4368532	144.28s
20	clock	4339470	43.39%	8611720	405.87s
30	clock	6508731	43.39%	12916270	630.45s

• 在单线程情况下，lru和clock算法的hit rate是一样的，但是lru算法用时更短
• 在多线程情况下，线程数少的时候clock的hit rate更高，但是用时多了一个倍
• 在多线程情况下，线程数多的时候lru的hit rate更高，同时用时更短

压缩文件

tar -zcvf ../SA21229005-熊荣康-lab.tar.gz ./*