In-memory technology is the vanguard of future computing. FedCentric has pioneered in-memory database (IMDB) since 2005, deployed the world’s largest IMDB (8TB) in a production environment in 2011, and share a patent for processor and memory affinity. FedCentric works closely with innovators in the field of advanced data management techniques. The common framework enables performance and functionality breakthroughs in one discipline to benefit others. Layered on these frameworks are discipline specific algorithms, code, and tools for bioscience, cyber, and fraud. We have employed large memory solutions for Frederick National Lab for Cancer Research, United States Postal Service, and our Argus Cyber Security Platform.
Working with our partners SGI, HPE, and Intel, FedCentric provides the best of bread open source and commercial HPC software stack. FedCentric storage engineers have implemented high-speed storage systems at government labs using parallel files systems including Lustre and GPFS.
FedCentric constantly evaluates the best of breed HPDA software solutions to complement the resources in large memory servers. Open source software stacks like Apache Spark thrive in big memory and can outperform Hadoop by 100X. These speedups are game changers, enabling researchers and analysts to perform more complex queries and reduce time to insight. GPU accelerated databases like MapD can provide 100-1000X while through common SQL, JDBC, ODBC, or Thrift interface.
In-Memory Data Bases
In-memory databases reduce latency and service time, two critical components for improving data-intensive processing. By eliminating spinning disk rotational latency, applications spend their time processing data instead of waiting for I/O operations to complete. By storing everything in memory, data access is at memory speeds, instead of multiple layers or mapping buffer caches to files, pages, and rows on disk. FedCentric has been at the vanguard of large scale in-memory databases including Oracle TimesTen (TM) IMDB and MemSQL. Our expertise in scaling up large in-memory data-intensive workloads led us to dramatically increase the performance of the USPS Revenue Protection system from around 5000 scans processed/second to over 1.1 million scans processed/second, all on commodity-based hardware.
FedCentric worked with MapD to scale up very large analytics workloads on their in-memory column-store SQL massively parallel database. Using NVIDIA GPUs, MapD is orders of magnitude faster than even the fastest CPU-based solutions. MapD executes millions of analytics queries on billions of rows of data and visualizes the results in seconds instead of hours.
FedCentric is one of the first companies to leverage the benefits of Oracle 12c Multitenant in creating high performance scale-up database consolidation solutions. Instead of spreading Oracle databases across thousands of servers and tens of thousands of CPUs, FedCentric implements large-memory HPE servers running Oracle 12c Multitenant. The resulting large memory/low core count solution results in a much more efficient use of hardware resources and operations personnel.