The idea behind the dynamic architecture is to allow the software to become scalable depending on the amount of load of the system. The pool of processes architecture will allow for multiple QNM's, but those QNM's will have a fixed number of checker processes in them. The idea behind the Dynamic Architecture is to allow each QNM to change the amount of checker processes it has depending on the work that is being done. This will in turn reduce the amount of QNM's needed and reduce the amount of time the jobs will take.
The diagram below shows the updated architecture with client server labels.
The diagram below shows the updated architecture with client server labels.
I suggest you modify the channels connecting NM to QNM to distinguish the pairs
ReplyDeleteI suggest you annotate the diagram to show the node boundaries.
relabel qnm qnmd for dynamic
It is basically coded apart from the NB. You did have the grace to produce the diagram before telling me that!
So the problem becomes one of showing the difference in performance between the two versions. You will need some scripts that processes files of data with known proportions of good and bad emails.
You will have to introduce a delay into the checking processing in some way that varies with whether good or bad but also has variance for different length good and bad. In the checker.
You must ensure that the input stream for each file runs at known rate ie you can have delays between emails but each time the file is used the same delays are used..
Not much progress due to a problem with sending messages over network to the NB process. I thought it was a serialisation problem but we shall see.
ReplyDeleteProgress has also been slowed due to external factors.
You have THREE weeks you need to write more.
You also have to design a poster. which needs to convey what you have done without actually being able to show it working.
Introduction and motivation
lit review and background technology
Two design approaches
Implementation and Testing of the designs
Results obtained
Evaluation
results obtained
and of the project process
conclusion and further work
It would appear that the problem between NM an QNM was the wrong implementation of getting data from the queue in NM. See AN EARLY CHAPTER OF ucape PART 1.
ReplyDeleteOtherwise has been doing writing and more to do!"!
The problem will be getting test results.
I suggested that for testing purposes Rory could consider his laptop as 1 x 2-core node
ReplyDelete2 x 2-core node a 1x 4 core node at least to show how he would have tested his system were he in the Uni using machines in D2.
I suggested he send me chapters ASAP. We could then arrange for another skype session once the comments had been sent back.
We discussed his report and I encouraged him to send me something sooner so I can comment on it.
ReplyDeleteHe does appear to be making progres.