Zettaset Patents Data Access Approach

Big data security specialist Zettaset said it has been awarded a U.S. patent for a technique designed to boost data access and performance in distributed computing frameworks like Hadoop and NoSQL.

Zettaset said this week the U.S. Patent and Trademark Office issued a patent for its DiamondLane technology on June 23. The U.S. patent covers “distributed storage medium management for heterogeneous storage media in high availability clusters” (U.S. Patent No. 9,063,939-B2).

Zettaset, Mountain View, Calif., said DiamondLane could be used to prioritize access to data in distributed clusters based on user preferences. Data is then assigned to higher- or lower-access rate storage across nodes and racks in a cluster. The company claims the data storage optimization technology would allow system administrators to fine-tune access to specific data files or applications, then optimize performance for “high-priority” users or mission critical applications.

The system prioritizes data based on how often it is accessed, assigning a “hot” data tag to the most frequently used data. High priority data is then assigned to higher-access-rate storage media like solid-state devices (SSDs) or RAM.

Storage allocation and distribution of tiered data can be configured as:

• Hot: RAM backed by SSD
• Warm: SSD
• Normal: Hybrid SSD/High-RPM hard-disk drive (HDD)
• Cold: HDD

Zettaset added that its DiamondLane technology is designed to speed up access to “hot” data while reducing network latency. Conversely, “cold” data can be assigned to be stored by less costly media like regular HDD.

The company said its patented approach improves on an existing “dynamic data tiering” approach that has been optimized for Hadoop and NoSQL environments.

DiamondLane identifies information in one or more files by access patterns and uses proprietary protocols and algorithms to distribute and redistribute those files in a Hadoop cluster, for example, based on those patterns. Data is dynamically rebalanced and re-distributed to storage types with appropriate access rates as patterns change. Access patterns and distribution/balancing algorithms consider multiple parameters, including access frequency, data volume, storage types and storage capacity, the company said.

Access patterns along with distribution and balancing algorithms take into account several parameters, including access frequency, data volume, storage types and storage capacity.

The proprietary technology is being promoted as a way for IT organizations having trouble scaling up their big data platforms for production deployment while maintaining security and performance. Moreover, the company touts DiamondLine as specifically aimed at boosting data access and performance in distributed computing environments like Hadoop and NoSQL.

Last year, Zettaset was granted a separate U.S. patent focusing on Hadoop security. That patent covers the underlying technology in the high availability version of Hadoop that prevents a “split-brain” situation where multiple master nodes think they are controlling a Hadoop cluster.

Recent items:

Avoiding Split Brained-ness in HA Hadoop Clusters

Hadoop and the Encyrption Mandate