A small search firm
Zvents released under a free license a unique development - a database management system that is easily parallelized on hundreds of machines. In the role of servers, standard cheap hardware can be used; if necessary, it is replaced on the fly without data loss. The new program
Hypertable is theoretically designed to work on a cluster of 1000 nodes, although the current alpha version 0.9 was tested only on ten. But the tests went great, and Yahoo has already shown interest in the development. The fact is that Hypertable currently uses the
Hadoop file system, the lead developer of which works just at Yahoo.
The leaders of the company Zvents say that they were forced to open their program in the form of open source, because they have a small software company and absolutely no money for infrastructure. Such commercial-scale DBMS should be tested on huge clusters.
The Hypertable program is modeled on the famous Bigtable database used by Google. This distributed system, when it was
presented to the scientific community in 2006, created a furor as one of the best inventions in the computer field. True, the patented
Google File System is used as the file system.
Software interfaces to Hypertable are slightly different from those that provide access to the Bigtable. Although Hypertable cannot be compared in functionality with powerful SQL databases, it is much more advanced than the primitive
memcached , which is often used in the “cache” for SQL. By the way, leading programmers from MySQL say that theoretically Hypertable can also be adapted for such a role.
')
via
LinuxWorld