Dear,

To Compare, you may use log file to record the load events for each algorithm, and compare the result at the log file.

Regards.

Thnak you, Sir

But how to initiate the coding for my own load balancing algorithm?

It’s probably a good idea to write your own load balancer as an academic exercise (if you have the time). But you are probably better off going with something currently available for a live deployment. Reason being is that these products have been through several development and testing iterations and have been deployed and hardened. Developing something custom means that you have to take on the extra expense of the full software development cycle for a product that is, probably, already capable and available.

Now, if there is specific, necessary, functionality that is lacking in currently available solutions then the next best approach would be to take something that is open source and add that functionality. That way the amount of testing you need to to is limited to the functionality you add and maybe some regression to make sure you didn’t break anything… ;-)

That being said, I would argue that you should consider more appropriate responsibility of the servers involved. If you want to intelligently route requests, then it should probably be the load balancer keeping track of the amount of load that each upstream server is handling rather than having to ask the upstream server. It would be fairly inefficient, in a number of ways, to ask all the servers in your cluster how busy they are to make a decision as to which one should handle the next request. Firstly, you would tend to hold up incoming requests to find a server to route to. Which means that you could have issues with scaling. Secondly, and perhaps more importantly, you have a potential O(N) growth problem when scaling your upstream servers if the load balancer requires all of them to respond before deciding which will process “this” request.

i have written the base codes for the load balancer, but i m not getting the way to implement the same in cloud analyst tool. in my code load balancer is trying to find the best suitable neighbor for placing the vm and cloudlets so on. my problem is how to make required changes in the load blanacer algorithm to implement the same.

Dear,

You differentiate between VM allocation policy, and cloudlets allocation/scheduling.

But it is good to link the VM allocation policy with cloudlet calling to achieve the load balancing.  And I think that you should enter another parameter such as power consumption to rich your work. Only load balance is simple to monitor the cloudlets allocation and sort free VMs to be ready for allocation, then allocate to the first one and remove it from the free VMs.

Regards.

The simulation tool Cloud Analyst provides the following three in-built load balancing strategies:

The corresponding java codes for these load balancers are openly accessible in the Cloud Analyst module. By analysing the codes and understanding the details of the strategies and how they are connected to other entities of the module, you can decide whether the load balancing algorithm you want to design needs coding from scratch or whether you can just modify the existing codes and implement your strategy.

Another thing that you need to focus on is the service broker policy you want to design your algorithm for. Though you have the option to design a generalized strategy which can be used with any of the service broker, but it would help if you consider the details of existing service brokers while implementing your load balancer.

For further details about how you can add the code into existing module of Cloud Analyst, you can refer to: http://www.cloudbus.org/students/MEDC_Project_Report_Bhathiya_318282.pdf

For details of service broker policies, and how the existing algorithms can be modified for better performance, refer

Chapter Priority Based Service Broker Policy for Fog Computing Environment

& Conference Paper Security-based service broker policy for FOG computing environment

Hope it helps!

Best

~ Deeksha Arya