xin zhang

Implement lenet version of Convolutional Neural Networks for handwritten character recognition algorithms.
1) Implemented lenet version of CNN algorithm.
2) Tested the algorithm on MINST dataset and achieve 98% recongition rate.

Implemented a latch-free Skip List in the Peloton database management system: implement a multi-thread skip list in Peloton database with concurrency control in a modern, order-preserving indexing data structure. This projects add another latch free index for Peloton database management system in addition to BW+ tree. This skip list index support insert, delete, scan, backward scan, garbage collection functions.
1) Implemented latch free index for Peloton database using skip list
2)… Show more

Implemented a latch-free Skip List in the Peloton database management system: implement a multi-thread skip list in Peloton database with concurrency control in a modern, order-preserving indexing data structure. This projects add another latch free index for Peloton database management system in addition to BW+ tree. This skip list index support insert, delete, scan, backward scan, garbage collection functions.
1) Implemented latch free index for Peloton database using skip list
2) Implemented garbage collection for skip list which can recycle the logical deleted skip list node.
3) Added test cases for Peloton database. Show less

Implement CloudFS using C as user-level code that uses libfuse to service file
system calls of applications attempting to access the SSD.
1) Implemented the CloudFS system which is a hybrid file system with two different storage components: a local SSD and a cloud storage service similar to Amazon S3.
2) Implemented de-duplicate algorithms to reduce cloud usage.
3) Implemented snapshot to help user to recover certain snapshot. Also use cache algorithm to further reduce the cloud… Show more

Implement CloudFS using C as user-level code that uses libfuse to service file
system calls of applications attempting to access the SSD.
1) Implemented the CloudFS system which is a hybrid file system with two different storage components: a local SSD and a cloud storage service similar to Amazon S3.
2) Implemented de-duplicate algorithms to reduce cloud usage.
3) Implemented snapshot to help user to recover certain snapshot. Also use cache algorithm to further reduce the cloud usage. Show less

· Course project of Cloud Computing
· Developed a web service on AWS using VertX as server, and use MySQL and HBase as database to handle different queries.
· Wrote programs in Hadoop MapReduce to extract, transform, and analyse a 1TB twitter dataset
· Used sharding and in-memory caching to accelerate query processing

Designed and developed a Flash Translation Layer for SSD Simulator. This Flash Translation Layer can do address translation, garbage collection, wear leveling optimization.

In this project, we implemented an on-line hockey video game using django, javascripts. This on-line hockey game support enter game lobby, select team, room management and hockey. This game support real time updated of both the statistic of lobby information and hockey game and game recovery when one of the user is off-line. I am mainly focus on implement the backend of lobby room part, which include user login, use logout, enter room, leave room, select team, start game, show the real time… Show more

In this project, we implemented an on-line hockey video game using django, javascripts. This on-line hockey game support enter game lobby, select team, room management and hockey. This game support real time updated of both the statistic of lobby information and hockey game and game recovery when one of the user is off-line. I am mainly focus on implement the backend of lobby room part, which include user login, use logout, enter room, leave room, select team, start game, show the real time information of lobby. Show less

• Deployed a VM-based YARN cluster. Installed YARN and modified scheduling policy server
to schedule the jobs that were submitted to the cluster. Implemented the FIFO-random, FIFO Heterogeneity awareness, Shortest Job Heterogeneity awareness scheduling policy.
• Proposed job utility awareness scheduling policy which consider preference of GPU job and MPI job and group job utility.

In this project, we add the constraint function for memory based Peloton database. In this project, we add constraint check from scratch. The goal of this project is make peloton has the same constraint function as postgres. The constraint function include: primary key, unique, default, not null, check and foreign key constrain check. This constrain check function not only support single column check but also multiple constrain checks. We did following things:
1) Implemented single column… Show more

In this project, we add the constraint function for memory based Peloton database. In this project, we add constraint check from scratch. The goal of this project is make peloton has the same constraint function as postgres. The constraint function include: primary key, unique, default, not null, check and foreign key constrain check. This constrain check function not only support single column check but also multiple constrain checks. We did following things:
1) Implemented single column constraint check: including not null, primary, unique, check and foreign key.
2) Implemented multi columns constraint check: including not null, primary, unique, check and foreign key.
3) Added test cases for constraint check for peloton.
4) Added function for cascade foreign key delete function
5) Added Alter table function for peloton. Show less

In this project, we implement D-Cube algorithm to detect dense block in dataset with high dimension. Different from java version of original D-Cube algorithm which make use of cloud, we run the experiment on single machine with the assumption that not all inner result can be fit into memory. D-cube algorithm can detect suspicious cluster in dataset and very useful for network attack detection, etc.
1) Implement the D-Cube algorithm in python and postgresql.
2) Do the experiments on five… Show more

In this project, we implement D-Cube algorithm to detect dense block in dataset with high dimension. Different from java version of original D-Cube algorithm which make use of cloud, we run the experiment on single machine with the assumption that not all inner result can be fit into memory. D-cube algorithm can detect suspicious cluster in dataset and very useful for network attack detection, etc.
1) Implement the D-Cube algorithm in python and postgresql.
2) Do the experiments on five popular real datasets: Amazon review, Yelp review, english wikipedia revision history, DAPRA dataset, AirForce dataset and show the efficiency of implemented algorithm. Show less