|
|
|
|
|
|
|
ZeroNines® Technology
|
It has long been recognized that the holy grail of high-availability information technology lies in geographically dispersed data centers, each capable of providing service in case the primary data center fails. Although simple in concept, the underlying technological challenges have made this goal unobtainable.
The industry has provided numerous solutions in the form of "fail-over", whereby if a processor ceases operations, another processor can immediately step in and take its place. However, this model requires that the processors share common storage - extremely difficult over long distances, and impossible between continents. Without common storage, how will the processor that has recently assumed control know about the transactions that occurred on the failed processor immediately prior to the fail-over?
|
The solution is found by taking a much simpler approach to the problem. One that does not involve fail-over, shared storage, complex networking technologies, or time-dependent inter-processor communications.
ZeroNines achieves absolute availability with geographically dispersed stand-alone servers, a shared-nothing model. Each server operates as though it is the only server, unaware of its distant twins. Each server processes every transaction simultaneously with the other servers, allowing for any server to assume or share primary responsibility as needed.
There are three technological challenges in implementing a shared-nothing model: 1) how to execute the transactions on every server transparent to the client and without requiring special software on every client, 2) how to ensure that the transactions successfully complete on every server, and 3) how to manage servers completing transactions at different speeds due to communication delays and processing times,
|
|
The ZeroNines® model of absolute
availability is immune to regional
failures such as power outages.
|
also known as latency. ZeroNines® technology completely addresses these challenges, and for the first time, enables the geographically dispersed shared-nothing model of high-availability.
In the ZeroNines® model of absolute availability, client requests are transparently routed through a ZeroNines® switch, which re-broadcasts the request to all servers. The first response received is returned to the client, who remains unaware that there is more than one server. The intelligent ZeroNines® switch distinguishes successful transaction completions from errors, shielding the client from any server failure. After the first server has responded to the client, the ZeroNines® switch continues to monitor the other servers to ensure that every server completes the transaction. This simple model allows multiple servers to operate completely independently while maintaining transactional synchronization. Absolute availability is achieved as the chances of all geographically dispersed stand-alone servers simultaneously failing becomes statistically irrelevant.
|
|
|
|
|
