In a paper published this week in the journal Nature Photonics, an international team of researchers led by a group at the University of Southern California in Los Angeles describe a method for transferring data at 2.56 terabits per second.

That's 85,000 times faster than today's 30 megabyte-per-second broadband Internet — roughly the equivalent of transferring 70 full-length DVDs in about one second.

The method in question uses "twisted light" beams to carry data through a new data stream channel, like a radio with its very own radio station.

The team — comprised of investigators from China, Pakistan, and Israel — says that the benefit of their data transmission technology is that it eliminates the need for bandwidth altogether.

If you'd like to dive into the photonic nitty-gritty, click here to read their paper.

But before you get too excited, this technology is not going to be available any time soon. Apparently the Earth's atmosphere interferes with twisted light data transmission over long distances. So while this may not prove to be a viable replacement for the current Internet infrastructure, there could be a future for this technology in the data center in the form of interconnects or networking fabric —assuming there are processing cores fast enough to handle that rate of data transfer.

A group from the University of Pittsburgh demonstrated a different technique to transmit data earlier this year at rates that also left current bandwidth limitations in the dust. The team, led by Hrvoje Petek, created a "frequency comb" that created over 100 terahertz of bandwidth using a group of atomic motions in a semiconductor silicon crystal.

Amazon Web Services experienced yet another outage last week, this time due to the failure of a cooling fan. Yup, you read that right — a cooling fan brought down the great Amazon cloud. The cloud outage affected a number of websites, including Heroku, Pinterest, Quora, and HootSuite.

The source of the outage occurred at a northern Virginia data center when the facility lost utility power. The center then switched to a backup generator power, but then nine minutes later, a defective cooling fan caused one of the backup generators to overheat and fail.

The incident began on June 14th at 8:44 pm PST and lasted until 10:19 pm PST.

According to the AWS Service Health Dashboard blog:

So it seems that cloud computing continues to receive doses of reality that counter its posturing as the ultimate replacement for the onsite data center as well as the promise of infinite resiliency and reliability.

Last week's outage was the third major outage in the last 14 months for the "US-East-1 availability zone" — Amazon's oldest availability zone based in a data center in Ashburn, Virginia.

Last April, the US-East-1 zone had a major outage as well as another less serious incident in March.

And back in 2010, the US East region experienced a series of four outages in a single week.

The first trillion objects are always the hardest.

Amazon Web Services has announced that their S3 cloud computing storage service is now storing one trillion objects (1,000,000,000,000 or 10¹²) — that's 142 objects for every person on the the planet or 3.3 objects for every star in our Galaxy.

Basically, it would take you 31,710 years to count them all, says Amazon's blog, so just to reiterate, the idea is: Their cloud stores a lot of data.

Amazon attributes this growth to customers' use of the S3 object expiration feature, which allows users to specify an expiration date on their data. So far, S3 users have used this feature to delete over 125 billion objects since its release last year.

The two things about this announcement that are notable: Amazon's cloud as a storage solution is here to stay; the days of the cloud as an IT novelty or fad to be viewed with suspicion are clearly over.

The other noteworthy take away is that the object expiration feature.

This new feature might come in handy for small labs or individual investigators if they have an informatics pipeline that connects a sequencing platform directly to the cloud, a model currently being pursued by Illumina in the form of their BaseSpace cloud service.

Researchers can thank the Pittsburgh Supercomputing Center, or PSC, for seriously ramping up the Galaxy platform. The folks at PSC have just completed a new super high-speed link from Galaxy to the National Science Foundation's Extreme Science and Engineering Discovery Environment, or XSEDE using their Three Rivers Optical Exchange, a 10-gigabit per second fiber-optic high-performance Internet networking hub.

There are currently over 10,000 Galaxy users running 4,000 to 5,000 analyses per day, and that number is growing, so the need for a HPC resource on steroids is clear.

"The network connection to XSEDE through PSC is a huge breakthrough," says Anton Nekrutenko, Galaxy co-developer and associate professor at Penn State. "It provides us with the ability to run up to 150,000 jobs per month, and we expect to quadruple that as this link gets fully up and running. It allows biologists to take advantage of HPC resources in ways they otherwise could not, not only the computing, but the storage resources at XSEDE sites. It democratizes research by making XSEDE useful for a scientific community that traditionally has not been a heavy user of high-performance computing."

The new bandwidth hookup is made possible with a four-year $1.5-million NSF's Academic Research Infrastructure program grant.

The latest storage vendor to rollout a flash/disk array that uses dedupe and compression to deliver impressive cost per gigabyte numbers is Tegile. Their new Zebi product is a multi-protocol box that comes equipped with iSCSI, SMB, NFS, Fibre Channel, dedupe, and compression.

The Zebi box has 10 to 90 terabytes of raw storage capacity and with the inline dedupe and compression it can provide up three to five times the raw capacity. The Zebi HA2100EP model has 96 gigabytes of RAM, 1.2 terabytes of flash and 16 terabytes of disk and the A J2100 expansion tray has 800 gigabytes of flash and 28 terabytes of disk capacity.

The company’s main pitch for their system is that roughly 75 percent less capacity is needed for seven times the IOPS.

So for, Tegile’s customer list for their Zebi box includes Washington and Lee University, which has used the box to achieve seven times the IOPS of its previous storage and a 70 percent reduction in virtual desktop infrastructure capacity needs.

Flash as a large-scale storage medium came to the forefront in 2009, when the San Diego Supercomputer Center built the first flash-memory based supercomputer, called Dash. Since then, Appro, Amax, and Nimbus (to name a few) have all started offering flash-based solutions targeting the HPC market.

So why flash? It provides consistently faster data transfer times and improved latency than traditional mechanical hard drives. And because there is no motor to power, flash drives use less energy, thus saving money.

In addition to SDSC, academic researchers from Carnegie Mellon University and Intel Labs Pittsburgh have been working with an experimental cluster architecture called FAWN, or Fast Array of Wimpy Nodes. Each node is comprised of an embedded single-core 500 MHz AMD Geode LX processor board and a 4 GB compact flash card.