Introduction
As we know that today’s most of our applications are web based and web applications can be accessed from all over the world, when more than one user access our application at the same time concurrency issues for example deadlocks, race condition occurs, it only occurs under certain timing conditions. It is very difficult to track down it since it may disappear when you try to debug it. It may occur very rarely and these are often not detected during in house testing although we can identify them through manual or automation procedures.
History
If we look at history there were notorious concurrency bugs, one of them was Northeast blackout in 14 August 2003, which cause widespread of power outage in Unites States of America because of race condition issue in software. You can read more about this event from Wikipedia page.
https://en.wikipedia.org/wiki/Northeast_blackout_of_2003
Contemporary society is dependent on software, software runs in Airplanes, handheld devices, nuclear reactor controls, air traffic control and so on, so software is everywhere.
Identification of Concurrency Issues
First of all we should identify which software module may have concurrency issues, once we identify we can produce it. Most of the applications have these issues because developers think sequentially, so they miss them.
- Employee Termination
If we send multiple requests at the same time to terminate an employee. System will terminate the single employee for multiple times. Although we should not terminate an employee if it is already terminated with the same reason.
// Active=1, Terminated =0;
Mov A, “1” // A=Active, “1”, employee status is checked
Add A, “0” // A=Terminate, “0” is added into a status
Mov Status, A // A=Terminated mean “0” is updated.
- Add New Hire
If our application takes a unique username for adding a new user and developer has placed the logic on saving check if username is already exists give message “Username already exists, try different one”
If Locking mechanism is not implemented and two different user sent the saving request of same username at the same time, system will same them, because two requests were received by procedure and it saved to DB.
First Name: Stephen
Last Name: Synder
UserName: US
First Name: John
Last Name: Doe
UserName: US
- Concurrent updating the info
If an employer tries update the employee info and at the same time a manager tries to delete that user, then concurrency issues will arise in below cases.
- If first request received by procedure is for deletion and then updating request received then an employer will face an exception.
- If first request received by procedure is for updating and then deletion request received then a procedure will be executed correctly.
- If both requests were received at the same time even without delay of a milli seconds then race condition will occur and time out will happen.
4.Concurrent TransactionsSimple procedureBalance=Balance+amount;
- }
- Customer #1
- {
- Void deposit (int amount)
- Let’s take an example of two customers one sends $13 and other sends $10 at the same time. Customer# 1 has account balance $100 and Customer# 2 has account balance $150.
- If multiple users at the same time save the transactions then wrong balance will be shown if locking mechanism is not implemented.
- Mov A, Balance // A=100, balance is checked.
Mov B, Balance //B=150, balance is checked.
Customer#1 balance will be check and it will be $100 and will move to register A and if context switch happen then- And then on context switch $160 will be added in register B,
- On context switch $113 will be moved to Customer #2 Balance instead of $160.
- Customer #1 balance will be checked, which will be $100,
- On next context switch $160 will be added in register B.
- On context switch $160 will be moved to Customer #2 Balance
- On context switch $160 will be moved to Customer #1 Balance instead of $113.
- And then on context switch $113 will be added in register A,
- On context switch $113 will be moved to Customer #1 Balance
- On next context switch $113 will be added in register A
- Customer #2 balance will be checked, which will be $150,
- Mov Balance, A // A=113, balance is updated.
- ADD A, amount // A=113, $13 is deposited into a balance
- Online Banking Simple Scenario:Suppose you have a $25 in your account and we want to pay an electricity and Broadband bill but electricity bill is $10 and Broadband internet bill is $15. As you have $25 in your account and bill exceeds the amount $20+$10=$30.
- If concurrency is not handled in your online bank account, what will happen if you submit two request at the same time?
Simple procedure
Void OnlineTransfer (int amount)
{
Balance=Balance-amount;
}
- Electricity Bill case
Mov A, Balance // A=25, balance is checked.
ADD A, amount // A=25-10=15, $15 is deposited into a balance
Mov Balance, A // A=15, balance is updated.
- Broadband Internet bill case
Mov B, Balance // A=25, balance is checked.
ADD B, amount // A=25-15=10, $10 is deposited into a balance
Mov Balance, B // A=10, balance is updated.
Explanation:
Case#1
Balance will be check and it will be $25 and will move to register B and if context switch happen then
- Balance will be checked, which will be $25,
- On next context switch 25-15=10, $10 will be added in register B
- And then on context switch 25-10=15, $15 will be added in register A,
- On context switch $10 will be updated into register B.
- On context switch $10 will be updated into register A.
Case#2
Balance will be check, witch is $25 and will move to register A and if context switch happen then
- Balance will be checked, which will be $25,
- On next context switch 25-10=15, $15 will be added in register A.
- And then on context switch 25-15=10, $10 will be added in register B,
- On context switch $15 will be updated into register A.
- On context switch $15 will be updated into register B.
In both cases wrong balance is updated, after paying two bills we are left in our account $15 in one case and $10 in one case although our earlier total account balance was $25.
