What can be done with large, low cost, high performance memory?
OmniTier’s MemStac system solutions enable not only better implementation of standard caching systems, but also bring traditionally-“too big” scientific problems requiring very large memory into the mainstream. Feel free to browse the following items for a sampling of the transformative capabilities enabled by our solution!
DNA bioinformatics – UPDATED!
De novo genome assembly enumerates an entire DNA nucleotide sequence from a large number of much-smaller sub-sequences, without the benefit of a previously-known reference genome. High-throughput parallel sequencers can easily generate terabytes of raw data, which must be processed by an assembly algorithm. Solving this problem quickly is a problem of explosive interest in the bioinformatics field. Click on “learn more” find out how OmniTier’s CompStor product improves state-of-the-art sequencing!
A standard problem in the bioinformatics field is known as “k-mer counting.” The problem is simply stated: determine the number of times every small substring (“k-mer”) appears in a large DNA sequencing data set. The solutions to this problem are not nearly so simple. K-mer count algorithms are used for indexing and for error correction in DNA assembly. Click on “learn more” find out how OmniTier’s CompStor systems improve state-of-the-art k-mer counting!
eCommerce website acceleration using MemStac-enabled Intel Optane(TM) technology
OmniTier’s MemStac was recently featured at major industry events announcing the availability of Intel’s Optane technology. OmniTier applications expert Ameen Aslam demonstrates the many cost and performance advantages of a MemStac-enabled Optane system (hosted in the cloud at IBM’s Softlayer) in a webpage-server eCommerce problem.
Large sparse matrix computation
Sparse matrices occur naturally in many diverse applications, from VLSI circuit analysis to synthetic aperture radar to eCommerce recommendation engines. Most tools using commonly-available compute resources can process calculations involving matrices containing billions of non-zero elements before resorting to offline memory types (such as hard drives), slowing the calculation to impractical durations. See how an optimized MemStac-based solution demonstrates singular value decomposition for 10TB matrices and beyond!