Technology
Over the course of the last nine years Pixia has developed commercial-off-the-shelf software that represents a quantum leap in accessing ever-growing volumes of data.
The legacy approach to storing large datasets is to fragment them into thousands or millions of smaller chunks and store each chunk as a separate file. This is evident in full motion video where video streams are stored as independent two minute clips. 24 hours of video from a single source now results in 720 individual files on disk! It is also evident in large still image products, where imagery is split into tiles and each tile is saved as a separate file. CIB-1 for the CENTCOM AOR is approximately 1 terabyte but it is actually stored and delivered to users in over 3 million files in over 7 thousand directories!
Current operating systems are based on fundamental principles established in the 1970s and the 1980s. The idea of dealing with thousands of terabytes of storage containing varied chunks of data is a recent event. A good example of standard input/output (I/O) from a storage device failing to perform is in the system cache. With the cache constantly filling, CPU bound processes that utilize the data spend more time swapping pages to and from secondary memory instead of actually processing the data.
Pixia has developed a technology where files are placed in a single container which is in turn stored as a single file on top of the file system. The immediate and obvious benefit of this is improved data management via a massive reduction in the number of files stored on disk. In the examples mentioned previously, users would store one file (versus 720 files) for 24 hours of video and would store one file (versus over 3 million) for CIB-1.
In addition to the data management benefits of logically consolidating files on disk, there is a huge performance improvement when accessing the individual files embedded within the container. Merely opening and closing files is a very time consuming process that can now be virtually eliminated by opening and closing one single container file instead of thousands of actual data files. A good example of this is wide area persistent surveillance where individual frames are stored as files on disk. To play back the frames at high enough rates to create motion video is a challenge, especially when each frame has to be read from a separate file. With Pixia's technology these frames are all placed in a single container and playback is extremely fast. Even on laptops with internal hard drives, Pixia is able to play back the wide area persistent surveillance at over 30 frames per second.
Pixia's technology is agnostic to the actual data being containerized. This is not a compression technology and most users of Pixia prefer leaving all of their data uncompressed, but compression can be used where necessary. Additionally, the software enables the user to index the data being containerized beyond the record level all the way to the byte level. This enables metadata to be stored at virtually any level of fidelity desired by the user. Instead of having metadata assigned to a single 2 minute video clip the user can now assign metadata to a specific region of a single frame.
Pixia has SOA servers that serve data to clients using standard protocols. The data can be converted on-the-fly to many different formats based on the client's requirements.
