Using imprecise concurrency control and speculative lending of prepared data-item in distributed real-time nested transactions

The majority of the studies conducted in scheduling real-time imprecise transactions mostly concentrate on flat transaction models. In this paper, we apply this technique to distributed real-time nested transaction models. We consider that nested transaction is a collection of both essential and non-essential subtransactions. We propose a real-time imprecise concurrency control protocol (ICCP) that resolves the conflicts in favor of the essential subtransactions. For the lending of data-item, we have used the PROMPT commit protocol, which causes intra-aborts cascade in nested environment. To alleviate this problem, we propose a S-PROMPT commit protocol: the borrowing subtransaction carries out a speculative execution by accessing both before and after-image (prepared-data) of the lending subtransaction. Simulations we have carried out show that by allowing better concurrency between transaction trees and between subtransactions on the same transaction tree and the speculative lending of data-item, the firm real-time performance of nested transactions are greatly increased.