NETWORKS

Makani Networks offered a glimpse of its vision for the next generation storage infrastructure with the MakDB and MakSTOR products.

Three major trends delineate Makani Networks's vision of its next generation storage products. The first trend is that of rapidly declining prices of DRAMs making large amounts of main memory available at an affordable price point. Second is the growing ecosystem of open source solutions such as Hadoop that has commoditized the market for unstructured data processing, transformation, and analysis. The final trend is the shift from online transactions processing as we see it today to vast amounts of real-time data collection (including storing, processing, and mining) from smartphones, laptops/tablets, appliances, sensors, and wide range of other devices.

Makani Networks' MakDB and MakSTOR products are designed for performance, multiple choices with query data model, plus high availability and fault-tolerance. Whether the data model is SQL-driven or Map/Reduce, the most important fundamental shift from other approaches is that the underlying metadata schema and the data chunks are made main memory resident. This enhances read and write scalability and improves I/O performance drastically by a factor of 100x-1000x, with additional gains made available through other performance optimizations. The system also replicates each chunk for (partition) fault tolerance and performance; while a multi-master system ensures distributed redundancy and high availability. The system offers a flexible data consistency model.

With rapidly depreciating prices of DRAMs, a new generation of extremely fast in-memory hybrid database and storage products are possible. A further 10x-20x reduction in the prices of DRAMs and Moore's Law doubling computing capability every 18 months makes high-performance cloud computing and in-memory cloud storage quite affordable. Makani Networks exploits such trends and advances to architect and design its next generation storage and database infrastructure.