Page not found

Fluid Power Energy Storage and Conversion

Abstract: 

Fluid Power Energy Storage and Conversion

This research group is studying energy storage and conversion in fluid power systems. This work involves significant computational resources in the areas of CFD of fluid flow through various geometries and FEA of mechanical components in hydraulic components.

Return to this PI's main page.

Group name: 
vandeve0

Performance Evaluation of Storage Systems

Abstract: 

Performance Evaluation of Storage Systems

This group has focused their research on the design and evaluation of large-scale storage systems. They have investigated  the impact of NV-RAM on future computing and communication environments, and how to design and evaluate large scale Solid State Drives (SSD) and Shingled Magnetic Recording (SMR) drives. They are especially interested in I/O workload analyses for many emerging applications like Big Data or real-time video and audio streaming, and high-performance computing (like Lustre file system).

Return to this PI's main page.

 

Group name: 
dudh

A Parallel Programming Framework for Graph Data Analytics and Scalable Transaction Management Models for Key-Value Based Storage Systems

Abstract: 

A Parallel Programming Framework for Graph Data Analytics and Scalable Transaction Management Models for Key-Value Based Storage Systems

This project uses Itasca for two research projects:

  • Beehive: A Parallel Programming Framework for Graph Data Analytics: The goal of this project is to develop a parallel programming model utilizing key-value based storage systems for graph data analysis problems. The focus is on the investigation of parallel programming models for problems involving graphs or pointer-based data structures, where it is hard to extract coarse-grain parallelism for using the MapReduce model. The researchers have developed a parallel programming framework called Beehive. The Beehive model utilizes this key-value based distributed global storage together with the transactional data management primitives for parallel programming for supporting optimistic execution of data parallel tasks on different parts of the input graph. This framework is still under development and several aspects of its programming model are being investigated for scalability and fault tolerance. The ultimate goal of this project is to demonstrate the the Beehive system can be used for parallel computing of large-scale graph problems.
  • Scalable Transactions on Cloud Storage Systems: The focus of this project is to develop scalable techniques for transaction management in key-value based storage systems with geo-scale replication of data. Such systems are commonly used in cloud computing environments for large-scale data management. This investigation is based on the snapshot isolation (SI) model for transaction execution. This project is investigating transaction management in replicated data management system in large cluster environments as well as across different cloud data centers which are geographically distributed. The thrust of the plan is in two directions. One is to support transaction management techniques with causal consistency guarantees with partial replication of data. Under such a scheme, not all data is replicated at all sites in the system. The second thrust will be on supporting transactions with multiple consistency levels simultaneously in replicated database systems.

Return to this PI's main page.

Group name: 
tripathi

Solid State Transformers, Power Electronics, and Integrated Magnetics and Storage

Abstract: 

Solid State Transformers, Power Electronics, and Integrated Magnetics and Storage

Integrated Magnetic Design can lead to substantial reductions in weight of power electronic converters and increase power density. The integrated Cuk topology promises such benefits in addition to be able to synthesize zero-ripple terminal currents with appropriate choice of air-gap in the magnetic core. However, such an integration of windings usually entails complex design procedures, and several unwanted capacitive effects. Finite Element Modeling is required to extract the leakage inductances and interwinding capacitances so as to aid in the design process. Use of Ansoft Maxwell enables these researchers to do these. Apart from this, they are also performing further research on modeling of two and multiple winding transformers to determine the transformer AC resistance and leakage inductance.

Return to this PI's main page.

 

Group name: 
mohan

What are the different storage tiers?

Data can migrate between the following Tiers on Stratus: Active Analysis Volume Storage Secure Archive dbGaP Cache (s3cache) Persistent Secure Storage (s3secure) Sanitized Data (i.e., non-protected and non-governed) Tier II (tier2) Archive Tape Storage* (*) Availability TBD

MSI Upgrades Storage Solutions

One petabyte of tape storage for under $10,000? Leveraging new advances in tape media, the Minnesota Supercomputing Institute has expanded its storage portfolio to offer easy access to deep archival storage at low cost. MSI currently offers a high-performance storage solution from Panasas (3.2 PB...

Data Storage Acceptable Use

Data storage is a finite and valuable resource. Storage services provided to MSI users are solely intended to support data and computationally intensive research. Storing personal files that are not related to data-intensive or high-performance computing workflows should not be stored on MSI...

Hoard

The Hoard memory allocator is a fast, scalable, and memory-efficient memory allocator for Linux, Solaris, Mac OS X, and Windows. Hoard is a drop-in replacement for malloc that can dramatically improve application performance, especially for multithreaded programs running on multiprocessors and multicore CPUs.

Pages