مدل کارآمد و تحمل پذیر برای مدیریت معامله زمان واقعی در محیط تلفن همراه با دست به دست شدن های متعدد تصادفی و قطع مکرر کوتاه

In this paper, we present an efficient model for real-time transaction management in a cellular environment with random frequent handoffs and disconnections. In this environment, the Mobile Support Subsystems (MSS) contain special Agents with caching techniques to contain the frequently accessed data from the Fixed Hosts (FH), which can be concurrently or exclusively accessed by transactions being executed in Mobile Hosts (MH) under the radio range of a particular MSS. Each Agent consists of a Cache Coordinator for communication with the Fixed Hosts and the Mobile Hosts and for disconnection and handoff management. The Fixed Hosts consist of a wired network with two types of server-Data Servers and Transaction Servers. The Transaction Servers contain a Transaction Manager for concurrency control and real-time transaction management. Using this schema, the transactions can continue in the Mobile Hosts regardless of random handoffs and frequent short disconnections enhancing real-time transactions and minimizing rollbacks.

Terminals, desktop computers, servers are the Fixed Hosts (FH) that are interconnected by means of a fixed network. Large databases run on servers that guarantee efficient processing and reliable storage of data.Mobile Hosts (MH) like palmtops, laptops, notebooks, or cellular phones are, of course, not always connected to the fixed network. They may be disconnected for different reasons. The transaction processing models should incorporate the shortcomings of mobile computing such as unreliable communications, frequent disconnections, limited battery power, accidental probability, low bandwidth communications and reduced storage capacity. Frequent aborts due to disconnection should be minimized in mobile transactions. Correctness of transactions executed on both fixed and mobile hosts must be ensured by the operations on shared data. Blocking of mobile transactions due to long disconnection periods should be minimized to reduce communication cost and to increase concurrency. After disconnection, mobile host should be able to process transactions and commit locally. Mobile computing provides the possibility of concurrent access of data by mobile hosts which may result in data inconsistency [3]. Concurrency control methods have been used to control concurrency [12, 17]. Due to limitations and restrictions of wireless communication channels, it is difficult to ensure consistency of data. In this paper, we present an Agent based framework for real-time transaction management and handling of random multiple handoffs and frequent short disconnections with minimum rollbacks while preserving the ACID properties. The remaining part of this paper is organized as follows. Section 2 summarizes the related research. Section 3 presents the proposed transaction framework for cellular environment. Finally Section 4 concludes this paper.

In this section, we explain how our proposed transaction framework works efficiently for mobile computing environment with high probability of occurrence of random multiple handoffs and frequent disconnections. During frequent handoffs, the transaction continues in the MH. Although the MH is aware of the occurrence of the handoff(s), the transaction middleware and its programs act as if no interruption has occurred and the transaction continues to execute in the MH until any further communication is required with the Fixed Hosts or MSS Agent. After handoff, when the MH gets connected to another MSS Agent, the new MSS Agent waits for a certain amount of time, checks for further handoff and fetches the dataset for the active transaction in the MH from the MSS Agent containing the dataset with the latest timestamp. Otherwise,if further handoff occurs within that waiting time, the MSS Agent simply ignores the transaction. The overhead of transferring the entire update task of the transaction to the MSS Agent by the MH after the occurrence of long as well as short disconnection is avoided in this proposed transaction framework in contrast to the framework discussed in [1, 18]. We have differentiated between short and long disconnections. During short disconnection, the MH continues executing the transaction and ignores the disconnection (unless an update commit is required during the short disconnection period) until it gets reconnected. During long disconnection, the MH’s middleware saves the Edited Copy dataset as a WML (Wireless Markup Language) file in the mobile unit secondary storage and transfers the update task to the MSS Agent only if update is required at that time. After reconnection, the MH’s middleware loads the stored WML file and continues the transaction unless the ACID properties are violated. Further in our approach, once an update request is made by a MH, it is forwarded to the Transaction Manager by the Cache Coordinator (in MSS Agent). Once the Transaction Manager validates the update request and updates the Data Server, it sends update confirmation to the MSS Agent and sends the updated dataset to all MSS Agents using the dataset at that time. This improves the overall concurrency of the system and preserves the ACID properties as well. Thus this model is suitable for real-time transactions, where real-time response is required with less overhead incurred to the system, as well as for normal transactions without timing constraints.