Today the Texas Advanced Computing Center announced that the Wrangler data analysis and management supercomputing system is now in early operations for the open science community. Supported by a grant from the NSF, Wrangler uses innovative DSSD technology for data-intensive computing.
NSF is excited to support Wrangler, a new data-focused, community-accessible, cyberinfrastructure resource at TACC,” said Bob Chadduck, the program officer overseeing the NSF award. “Wrangler will significantly accelerate the ability of the research community to do data-centered science, to tackle today’s most complex, data-intensive challenges and to advance science, engineering and education across all fields.”
Dell Inc. and EMC are the two strategic partners providing the technology that make up the core of Wrangler. Wrangler uses EMC’s DSSD rack-scale flash technology to ensure speed and performance, enabling real-time analytics at scale.
TACC has been a valued customer for DSSD in breakthrough storage infrastructure innovation,” said Matt McDonough, senior director of product management and marketing for EMC. “The TACC Wrangler data analysis and management system validated the need for a new performance-oriented infrastructure to eliminate I/O bottlenecks for existing and emerging data-intensive, real-time and high-performance analytics applications. These applications need the fastest possible storage closest to the server, and DSSD offers this while mitigating latency and delivering blazing speed and performance.”
TACC provides an ecosystem of hardware resources and services to satisfy a wide range of research needs in high performance computing, according to Executive Director Dan Stanzione.
Wrangler will be a superb complement to our existing resources at TACC, including the Stampede supercomputer,” Stanzione said. “With its unique capabilities for streaming data, Wrangler doesn’t replace our existing resources, but rather enables entirely new use cases and brings in new users. We believe Wrangler will be a groundbreaking data intensive computing system for the scientific community, and we’re excited to see the discoveries Wrangler will enable.”
Many communities are ready and committed to adopt the system. Two projects already underway involve the evolutionary biology community, which uses an open source bioinformatics tool called OrthoMCL to group genes into families based on sequence similarities.
Hans Hofmann, director for the Center for Computational Biology and Bioinformatics at The University of Texas at Austin (UT Austin), is researching how monogamous behavior of vertebrate animals, such as mice, birds and fishes, independently evolved multiple times. His team’s research measures the activity of thousands of genes in the brain to find commonalities.
The problem is not one of computation; it’s that you always need to refer back to a database that’s populated with sequencing data specific to your scientific question. Wrangler is really built for these kinds of specialized applications,” Hofmann said.
Hofmann’s research shows that there is a common signature of gene expression activity in the brain of monogamous individuals across 450 million years of vertebrate evolutions.
“We could not have found this research result without OrthoMCL or Wrangler,” Hofmann said. His team’s research will be published in a few months.
Dhivya Arasappan, a bioinformatics consultant, works with various labs at UT Austin. One of Arasappan’s current projects involves Rhazya stricta, a plant found to produce medicinally important alkaloids. One of the important steps in the project is to group genes from Rhazya and related plants into families using OrthoMCL.
According to Arasappan, OrthoMCL has a unique workflow that fits well into Wrangler’s flexible environment. “The very first step the application does when given a set of sequences is to create and load all of the data into the database,” said Arasappan. “Wrangler can compare details in large data repositories, be they in a database or in files on disk, to enable research not previously practical to undertake. This helps researchers using the same system to avoid using the same database and accidentally overwriting information — this is where Wrangler does the magic.”
The result is that Arasappan’s analysis on Wrangler has had a much faster turnaround than running it on Stampede. For example, the problem of comparing multiple species fails on Stampede at a certain point because it’s not able to scale. “But when we run it on Wrangler we were able to go to six to eight species with no problems because it runs faster and scales better on Wrangler,” Arasappan said. They are hoping to find out which genes are responsible for producing medicinal compounds.
“We’ve built a really great relationship with the staff at TACC and this is just another example,” Hofmann said.
In addition to hosting part of the system for replicated storage, Indiana University will participate in operations and training, and will help users optimize their network performance between their home institutions and Wrangler. The Computation Institute, a joint initiative of the University of Chicago and Argonne National Laboratory, will integrate their Globus Online service within the Wrangler project to make transferring data to and from Wrangler simple and fast.
Wrangler’s performance and storage capabilities for Big Data applications will be enhanced through tight integration to TACC’s Stampede supercomputer and to NSF Extreme Science and Engineering Discovery Environment (XSEDE) resources around the country. Immediately upon deployment, Wrangler will be a part of the broader XSEDE ecosystem. Integration with Globus Online, the official data transfer mechanism for XSEDE, will provide for rapid, reliable and secure data exchange with other elements of the national cyberinfrastructure.
Wrangler provides many different services to help researchers solve their data computing needs, including:
- Massive, replicated, secure high performance data storage (10PB each at Indiana and TACC)
- A large scale flash storage tier for analytics, with bandwidth of 1TB/s and 250M IOPS (6x faster than Stampede)
- Embedded processing of more than 3,000 processors cores for data analysis
- Flexible support for a wide range of data workflows, including those using Hadoop and databases.
- Integration with Globus Online services for rapid and reliable data transfer and sharing.
A fully scalable design that can grow with the amount of users and as data applications grow.