Hadoop Ecosystem Evolves: 10 Cool Big Data Projects

Velostrata, a startup with expertise in data caching, streaming applications, and spoofing the Amazon Web Services environment, says its capability to run an enterprise's workload on Amazon, no conversion needed, is now generally available.

It can perform that process now, provided it's in the form of a VMware virtual machine, and has plans to expand to other formats.

The Velostrata approach can also draw the data the application needs out of a cache maintained on-premises, then write new data into the cache from the computation in the cloud. That means the enterprise data resides continually behind the enterprise firewall, even though parts of it are used momentarily in the cloud. The database system and other forms of permanent data storage stay on-premises at all times.

If it works as advertised, then it opens up a potential new avenue of hybrid cloud computing: a strong partnership between on-premises computing and the cheap, expansive power of cloud servers. (Velostrata has beta users that it's converted to paying customers.)

[Want to learn more about Velostrata? Read Velostrata Hybrid Offers Public Cloud Compute, On-Premises Data.]

But it's more complicated than simply running an application in the enterprise data center or in the cloud. As many operations managers know, it gets complicated fast when something in the cloud is supposed to work with something on-premises. Velostrata needs to do some fancy footwork to make the two parts work together.

One of its paying customers moved a complex ERP system into the Amazon Web Services cloud while maintaining the data on-premises. "There were a lot of modules involved running on eight virtual servers. It took 20 minutes to get it up and running, and another 20 minutes before it was running with performance," said Issy Ben-Shaul, cofounder and CEO.

One reason to think that Velostrata can actually tap-dance its way to multiple, successful implementations is the track record of the young firm's management team. Ben-Shaul is also the cofounder of Actona Technologies, a wide-area file caching system sold to Cisco in 2004 for $82 million, and Wanova, a virtual desktop deployment system sold to VMware for $100 million in 2012. His Velostrata cofounder is Ady Degany, now chief product officer, a former executive at StorSimple, which was acquired by Microsoft for an undisclosed amount in 2012.

One of the approaches that Velostrata implements for its hybrid operation is a caching system that works in both locations -- on-premises and in the cloud. It employs frequent look-ahead techniques on a running application to determine what data it will need next, fetches it from the cache on-premises, and delivers it to the caching system in the cloud, Ben-Shaul explained.

Ben-Shaul said Velostrata also establishes an Amazon Machine Image in the cloud before moving a workload. That AMI serves as a bootloader for the enterprise workload when it arrives. Velostrata inspects the workload in a VMware virtual machine, detects the differences in the environment in which it's been running versus the one it will encounter in the cloud, and injects the correct device drivers that it will need there before sending the workload to the bootloader. Once there, it can continue to run as an ESX Server virtual machine, but to AWS it looks like an AMI.

"We don't need to convert it to an AMI," Ben-Shaul said in an interview with InformationWeek in January. He called that bootloader a "pre-AMI" that can read the image of the application and start it running.

The unconverted workloads are VMware virtual machine files, known as VMDKs. They undergo the operating system device driver adjustment before they arrive. "We do injection on the fly," he said at that time, meaning the workload can stream as if it were a virtual device, ready to run upon arrival.

The combination of ready-to-use cached data and ready-to-run virtual machines means getting a workload up and running in the cloud can occur much faster than would happen if it had to undergo a conversion and be reformatted as an Amazon Machine Image. AMIs are Xen-based virtual machines with a few proprietary alterations by AWS.

"We are a little bit out of the box" of typical cloud operations, conceded Ben-Shaul at the end of his explanation.

Learn to integrate the cloud into legacy systems and new initiatives. Attend the Cloud Connect Track at Interop Las Vegas, May 2-6. Register now!

This approach is suitable to enterprises that want to keep their data on-premises but occasionally or frequently make use of the cloud's elastic compute power, he said.

"Because we don't move heavyweight storage to the cloud, we can have a workload running in a few minutes," he said in an interview prior to Velostrata's announcement of general availability March 10.

Management of the workload can be performed from vSphere's vCenter management console in a manner similar to the way it was managed when running in the enterprise data center. Velostrata has a vCenter plug-in that tracks and reports the activity of a Velostrata-engineered move to the cloud.

The data caching system is a read/write system that sends data to the cloud and writes to the data store changes that occur via the application there.

Ben-Shaul said he couldn't name the company that moved its ERP into AWS via Velostrata's product. But he did cite Guardian Industries Corp., a manufacturer of glass products for cars and buildings in 25 countries.

"We recently purchased Velostrata after testing the technology with a number of different use-cases and concluded that it performed extremely well," said Chuck Wiggins, director of infrastructure at Guardian.

"Velostrata's ability to stream stateful production workloads to the cloud in minutes without replicating the storage to the cloud fits squarely with our vision and enables us to realize our hybrid cloud strategy for production workloads," he said. It does so without requiring changes to its existing data center processes, he added.