Researchers move two days’ worth of data in just 100 minutes

100 minutes might seem like a long time. After all, it’s approximately the amount of time we spend watching a movie, so reading that it took researchers 100 minutes to download data might seem a bit, well, unimpressive; that is, until you understand the amount of data transferred. 

65 terabytes. 

That’s 65,560 gigabytes. Or 6,8157,440 megabytes. 

Or, to be super exact, 571,746,046,443,520 bits of data. 

Typically speaking, that amount of information requires, on average, two full days to transfer when using a 10-Gbps connection.

The team of researchers that achieved this feat come from the Argonne National Laboratory and DataDirect Networks (DDN). With help from Ciena, Brocade, and ICAIR, the group was able to sustain data transfer rates in excess of 85 Gbps (with peaks at over 90 Gbps) between storage systems in Ottawa, Canada and New Orleans, Louisiana, over a 100-Gbps wide-area network connection. 

They were able to sustain these higher rates by combining the embedded file system and virtual machine capabilities of the DDN storage controller, the high speed wide-area transfer capabilities of the Globus GridFTP server, and an advanced 100G wide-area network. 

“Embedding the GridFTP servers in virtual machines on DDN's storage controller eliminates the need for external data transfer nodes and network adapters,” explained Raj Kettimuthu, principal software development specialist at Argonne National Laboratory. “We sustained a data transfer rate of 85 Gbps for over 60 minutes — and occasionally for as long as 90 minutes — several times during the SC14 conference.” The SC14 conference is the leading international conference for high-performance computing, networking, storage, and analysis, and the event at which this technology was demonstrated. 

Those familiar with this data transfer technologies will point out that 90+ Gbps for memory-to-memory transfers is pretty straightforward and has been done before. Achieving the same rates for disk-to-disk transfers is much more difficult. Such challenges associated with this particular approach include choosing the right block size that works well for both disk I/O and network I/O, and selecting the appropriate combination of parallel storage I/O threads and parallel TCP streams for best possible end-to-end performance. 

There are two theories on how this particular setup works: network experts claim that storage is the bottleneck in the end-to-end transfer on high-speed networks, while storage experts state that the network is the bottleneck on transfers between sites with high-performance parallel file systems. 

The Argonne approach sought to find a middle ground and in doing so, achieved this record-breaking accomplishment.  

“This demonstration was aimed at bringing together the experts and latest developments in all aspects concerning disk-to-disk WAN data movement, including network, storage, and data movement tools,” said Kettimuthu.

The group believes that their approach can be fine-tuned to the point of being able to achieve 100+ Gbps wide-area transfer rates between storage systems using multiple WAN paths and additional storage resources in the end systems. 

Via NewsWise