عنوان مقاله:
روش های کنترل هم روندی در پایگاه داده های توزیع شده: یک بررسی و مقایسه
Concurrency Control Methods in Distributed Database: A Review and Comparison
سال انتشار: 2017
رشته: مهندسی کامپیوتر
گرایش: مهندسی نرم افزار
دانلود رایگان این مقاله:
دانلود مقاله روش های کنترل هم روندی
مشاهده سایر مقالات جدید:
A review on distributed database
At first, a distributed database model is explained in this part. A. A distributed database model: Fig.1 shows a common structure of a model. Every site in this model has four parts. A source which produces the transactions and maintains the levels of transaction information for site. A transaction manager who models the executive treatment of transactions. A concurrency control manager who implements the details of a special concurrency control algorithm and a resource manager who models CPU and I/O sources for site. In addition to these per-site components, the model also has a network manager who models the network communication behavior. 1) transaction manager Every transaction in workload, will have a master slave process, some of cohort groups and updaters. The master process occupies the site which transaction has been offered in. every batch of cohorts sends reading or writing requests to one or more files which have been saved in this site. A transaction, composed of a collection or group of cohorts, exists in every site that needs to access the data. Cohort groups connect to their updaters when they get permission to access required writing for repeated data and their updaters. A transaction can be executed sequential or parallel which is related to transaction class algorithm. 2) Resource manager Resource manager can be assumed as a model of operating system for a site which directs the physical resources of that site consisting CPU and discs. Resource manager provides CPU and I/O services for transaction manager and concurrency control manager and also provides message delivery services that use CPU sources. Transaction manager uses CPU and I/O resources to read and write a disc and also for sending messages. concurrency control manager uses the CPU resources too for processing the requests and sending the messages .
Comparison and analysis of existed methods
Here, four methods were used for concurrency control mechanism in distributed database. In two phase locking algorithm, we have problems like concurrency weakness, sequential rejection and deadlock and starvation happening [1,2,7,9,12] . In algorithm wound-wait, time priority of algorithms is observed [1,2,4,11,12] . In protocole according to timestamp, conflict sequence ability is assured and deadlock does not happen. but sequential harm or starvation is possible to happen and recoverability of execution plan is not assured [1,2,4,5,6,12]. But about optimistic algorithms or approval technic should say that: x In all previous techniques, system does some inspections before doing an operation on database. x In approval technic which is known as optimistic technique, no inspection is done before. Figure.10. c3 entrance and w3 remove from wv[2] x this technique, in comparison with locking technique, provides more concurrency. x In this technique, writing operation is not done in database directly, unless transaction execution has finished. x Instead in the middle of execution, all updates are done on the local versions of related data. x at the end of transaction execution, during an approval stage, it is inspected if any harm has been entered into serializability of execution plan or it was of no effect. x if serializability has not been observed, transaction is rejected in order to be restarted. x If no harm has been entered into serializability, stabilized transaction and databases are updated from local versions.
چکیده
در سالهای گذشته، پیشرفتهای قابل توجهی در توانایی عملکرد سیستمهای پایگاه داده‌های توزیع شده ایجاد شده است. در این سیستم اگر هماهنگی خوبی بین تراکنشهای مختلف وجود نداشته باشد، می‌تواند منجر به ناهماهنگی در پایگاه داده‌ها شود. امروزه به دلیل پیچیدگی بسیاری از سایت‌ها و روشهای ارتباط آنها، گسترش مدلهای مختلف بطور متوالی در پایگاه داده‌های توزیع شده کار دشواری است. هدف اصلی کنترل هم روندی در پایگاه داده‌های توزیع شده، اطمینان از عدم تداخل سایت‌های مختلف در دسترسی به پایگاه داده‌های عمومی است. الگوریتم‌های مختلف کنترل همروندی برای استفاده در سیستمهای توزیع شده پایگاه داده‌ها پیشنهاد شده‌اند. در این مقاله، برخی از روشهای موجود برای کنترل هم روندی در پایگاه داده‌های توزیع شده معرفی شده و مورد مقایسه قرار گرفته‌اند.
