In their groundbreaking（开创性） 1993 paper Herlihy and Moss propose to implement transactions for memory operations in hardware since software alone cannot deal with the problem efficiently.
Digital Equipment Corporation, at that time, was already battling with scalability problems on their high-end hardware, which featured a few dozen processors.
One thing about future development of memory technology is almost certain: latency will continue to creep up.
延迟将继续蔓延。
We already discussed, in section 2.2.4, that the upcoming DDR3 memory technology will have higher latency than the current DDR2 technology.
FB-DRAM, if it should get deployed, also has potentially higher latency, especially when FB-DRAM modules are daisychained.
The multi-media extensions in today’s mainstream processors implement vector operations only in a limited fashion.
Vector instructions are characterized by large numbers of operations which are performed as the result of one instruction (Single Instruction Multiple Data, SIMD).
Compared with scalar operations, this can be said about the multi-media instructions, but it is a far cry from what vector computers like the Cray-1 or vector units for machines like the IBM 3090 did.
消息事务是在生产者producer到broker或broker到consumer过程中同一个session中发生的，保证几条消息在发送过程中的原子性。
在支持事务的session中，producer发送message时在message中带有transactionID。
broker收到message后判断是否有transactionID，如果有就把message保存在transaction store中，等待commit或者rollback消息。
消息生产者使用持久（persistent）传递模式发送消息的时候，Producer.send() 方法会被阻塞，直到 broker 发送一个确认消息给生产者(ProducerAck)，这个确认消息暗示broker已经成功接收到消息并把消息保存到二级存储中。这个过程通常称为同步发送。
