Osaka University, Scality Inc., and Fujitsu Limited have announced they are jointly executing a cold data storage-focused field trial of intercontinental data center coordination, together with partners in Japan and Europe.
Based on a consortium for collaboration between academia and industry established in April 2016, the field trial will be conducted from November 2016 through to December 2017, and carried out between the data centers in Osaka University's Cybermedia Center and the data centers owned by Paris-based AntemetA.
In this field trial, the partners will build and test the effectiveness of a storage platform coordinated across remote locations, which offers improvements in throughput, disaster resilience, and lower costs by eliminating redundancy within individual data centers. This is done by duplicating data that is not updated very often, known as "cold data," in data centers on each continent, and applying a data sharing technology from Scality called conflict-free replicated data type (CRDT), which is used in distributed computing environments.
Currently, in order to coordinate data between data centers, two methods are used: one in which coordination is done through synchronous communications over short distances where data transfer delays are small, and another is to make backups through asynchronous communication over long distances.
When using synchronous communication over short distances, there is a high probability that multiple data centers might be simultaneously affected by large-scale disasters. When asynchronous communication is used over long distances, there are large delays in transmitting data between data centers, so the only option was standby redundancy coordination for backups or disaster recovery.
At the same time, against the backdrop of changes such as the spread of technologies like AI and the IoT, data which is generally not updated, called cold data, has taken up more than half of all data, primarily in the form of image and video data, and there is an increasing demand for ways to accumulate new data and share it between multiple locations. In order to achieve data center coordination between multiple locations, however, there have been issues in achieving data synchronization between data centers, and in improving response speed when reading the data.
“Geo Replication” system
In this trial, the participants will develop the "Geo Replication" system to realize the layering and data redundancy, which had only been achieved between data centers in close proximity or within a stand-alone data center, at long distances such as between continents. By layering and storing not only continually updated "hot" data and infrequently updated "warm" data, but also "cold" data, which is seldom updated, across multiple types of storage devices, long-term storage functionality, lowered costs, and increased responsiveness and overall throughput can be achieved across the system as a whole.
The companies participating in this consortium will advance the development of the Geo Replication system, and the field trial will be carried out, aiming for commercialization, with partners in Europe, using data that is actually in use in Osaka University's data centers.