Skip to content
Development of a YARN application for transactional key-value operations.
Java Shell Gnuplot
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
benchmarks
config
report_benchmark_scripts
src/ch/epfl/tkvs
test/ch/epfl/tkvs
.gitignore
BigData2015_ATransactionManagerOverYARN_Report.pdf
README.md
README_BENCHMARKS.md
benchmark.sh
compile.sh
configure.sh
printlogs.sh
run.sh
start.sh

README.md

Transactional key-value store for YARN

Development of a YARN application for transactional key-value operations.

How to use it

If you are running the , ./configure.sh local to configure the slaves file. Otherwise, if you are running it on the cluster of the course, please run ./configure.sh N to run the system on N nodes.

  1. HDFS and YARN must be running
  2. Start the YARN Application with ./start.sh. The Client REPL will start.
  3. Try :benchmark, it's gonna be fun. Please be nice to our immature system

Example benchmark test: :benchmark t 100 r 10 k 500 ratio 20 4. Stop the YARN Application with :exit. 5. View all logs with ./printlogs.sh.

You can look at the example client source code in test/ch/epfl/tkvs/userclient/UserClient.java

Selecting the concurrency control

At the moment, the key-value store support 3 concurrency algorithms: 2PL, MVCC2PL and MVTO. To select one of them, write simple_2pl, mvcc2pl or mvto in ./config/algorithm. By default (in absence of the file for instance), MVTO will be used.

You can’t perform that action at this time.