In the context of a MapReduce deployment in Hadoop, where is the final output of each Mapper worker stored? 

(select the option that best fits the question).

In BigTable, the master is a single process that performs top-level coordination for the entire cluster.

Although this may potentially incur in a bottleneck, BigTable frees the master from substancial coordination effort because of an important design decision. 

Which one is the most relevant?

(select the option that best fits the question).

Considering emulation by interpretation with decode-and-dispatch (D&D), there are a number of specific features to D&D that penalize its performance when compared to other interpretation-based approaches. 

Identify one of them.

(select the option that best fits the question).

Regarding the conditions for ISA virtualizability, as defined by Popek and Goldberg, there are a number of assumptions regarding the system. 

Identify examples of such assumptions among the following statements. 

(select all the correct options).

What sequence of byte-code could be observed in code that carries out the raising of an exceptional condition during the execution of a Java program, e.g. , an exception of type MyException?

(select the option that best fits the question).

When deploying an application on the standard environment of Google App Engine (GAE) PaaS, in order to achieve scalability, applications do not have complete freedom.

What are they forbidden from doing by default?

(select the option that best fits the question).

The Nova Scheduler is the component in OpenStack that is responsible for launching instances (virtual servers) and deciding in what physical machines (hosts, nodes) they should be launched.

Two of the main resources that the Nova Scheduler has into account when making scheduling decisions are: CPU (e.g., number of cores) and Memory (e.g. GB of RAM).

CPU and Memory are allocated following an optimistic approach employing overcommit (i.e., possibly allocating VMs so that the total of requested resources is actually higher than the available physical resources in the hosts/nodes).

What is the correct set of overcommit ratios used by the Nova Scheduler for CPU and Memory?

(select the option that best fits the question).

There are three properties referred to by the CAP theorem (consistency, availability, and partition-tolerance).

Which two of them have been selected to be enforced in Amazon Dynamo and why? 

(select the option that best fits the question).

In Azure, programmers can deploy applications comprising sets of instances executing services with different profiles.

What are these profiles?

(select all the correct options).

Consider a scenario with large scale Java-based web application with large number of users and whose interactions with the application result in the invocation of vast number of functions. 

In general, the application owners want to trace potential bottlenecks or inefficiencies with significant impact on the customer base. 

Specifically, the application owners want to know:

 - how many distinct users have a significant number of invocations to a function (i.e., more than 50 invocations) that, while successfully completed (HTTPStatusCode within the 200s range), are taking too long to complete (more than 10 seconds). 

The input consists of a text file where each line stores the relevant data to maintain about each individual request/invocation:

- <requestID> <timestamp>, <className>, <methodName>, <userID>, <HTTPStatusCode>, <responseTimeMs>

Assume the standard signature for the mapper: List<k2,v2> map(int, string) and reducer:  List<k3, v3> reduce(k2, List<v2>) functions, where map has the default MapReduce input (cursor/line number and text line content) and k2, v2, k3 and v3 can all be replaced with types of your choice.  

Write the pseudo​code for a MapReduce application, with a single ​map ​and a single ​reduce ​method.

The final output of the reduce function should produce, when adequate, a single record identifying each application function that may be causing a potential bottleneck, as described, and the number of users that may be affected, if any.

 (if you need, you can briefly state any assumptions at the end of the answer after the pseudocode.)

(No penalty for wrong answer.)

 Note: MapReduce is not natural language. Pseudocode needs not be Java but it must include clear operating steps well identified, it should not be a paragraph or set of paragraphs rephrasing the question. But do express your reasoning, nonetheless.