lab: review labs 1-6

source/Lab01/Instructors Guide/assignment.tex

@@ -2,7 +2,7 @@
 % AUTHOR: Rafael Belchior
 % Thanks to Prof. RUI SANTOS CRUZ for providing the template
 %
-\usepackage{helvet} 
+\usepackage{helvet}
 \renewcommand{\familydefault}{\sfdefault}
 \usepackage{a4wide}
 \usepackage{ucs}
@@ -17,8 +17,8 @@
 \usepackage{hyperref}
 \usepackage{cite}
 \usepackage[absolute]{textpos}
-\usepackage{tabularx} 
-\usepackage{tabulary}                 
+\usepackage{tabularx}
+\usepackage{tabulary}
 \usepackage{fancyhdr}
 \usepackage[table]{xcolor}
 \pagestyle{fancy}
@@ -68,16 +68,16 @@
 \lhead{}
 {\renewcommand{\arraystretch}{1.1}
 \fancyhead[C]{\begin{tabularx}{1.0\textwidth}{|l|X|l|l|}
-\hline 
+\hline
 % In the following line change Course Name: PPIII, PPB
 \textbf{EB 20/21} & \textbf{Enterprise Blockchain Technologies} & \textbf{Number:}  &  1 \\
 \hline
 % In the following line insert your Name and IST ID
-\multicolumn{2}{|l|}{Module I - Introduction} & \textbf{Issue Date:}  &  14 Sept 2020 \\ 
+\multicolumn{2}{|l|}{Module I - Introduction} & \textbf{Issue Date:}  &   \\
 \hline
 % In the following line insert the Activity CODE and Title (abridged)
 %\textbf{WP n.} (99) & (Subject) & \textbf{Group:} & (99) \\
-\multicolumn{2}{|l|}{Background: Distributed Systems} & \textbf{Due Date:} &  21 Sept 2020\\ 
+\multicolumn{2}{|l|}{Background: Distributed Systems} & \textbf{Due Date:} &  \\
 \hline
 \end{tabularx}}
 \rhead{}
@@ -92,63 +92,61 @@
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%,sdist2017
 
 \section*{Instructors Guide}
-This document provides a proposal of solution for LAB\#01, which concerns the an introduction to distributed systems and the RAFT consensus algorithm. Most of the solutions can be found at \cite{raft_paper, Ousterhout_presentation, raft_viz,raft_homepage}.
+This document provides a proposal for a solution for LAB\#01, which concerns an introduction to distributed systems and the RAFT consensus algorithm. Most of the solutions can be found at \cite{raft_paper, Ousterhout_presentation, raft_viz,raft_homepage}.
 
-This class aims to illustrate what is consensus, and provide an intuition on how does it work. For that, we present the RAFT algorithm. Hoewver, this class does not present RAFT in an exhaustive manner. For details, check the paper \cite{raft_paper}.
+This class aims to illustrate what consensus is and provide intuition on how does it work. For that, we present the RAFT algorithm. However, this class does not present RAFT in a thorough manner. For details, check the paper \cite{raft_paper}.
 
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 % YOUR TEXT STARTS HERE
-%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%   
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
 \section{RAFT}
 \label{sec:RAFT}
 
-\subsubsection*{Exercise 1: To consolidate your knowledge, watch the a guided visualization of the RAFT algorithm \cite{raft_viz}. 
-Elaborate on how the RAFT algorithm can contribute to eliminate failures from a client perspective.}
+\subsubsection*{Exercise 1:} To consolidate your knowledge, watch the guided visualization of the RAFT algorithm \cite{raft_viz}.
+Elaborate on how the RAFT algorithm can increase the system's availability from a client perspective.
 
-RAFT is a crash fault tolerant consensus algorithm - meaning that it tolerates when some servers in a cluster crash. When a client makes a request to the system, the leader disseminates the request, making every follower execute it. If a fault occurs in one of the followers, leading to an inconsistent state (error). Assuming that such error leads to a failure, and the system crashes, the resiliency of the whole the system as a whole will not probably fail, as that fault is tolerated by other nodes. This way, the system is more likely to be available. 
+RAFT is a crash fault-tolerant consensus algorithm that it tolerates when some servers are in a cluster crash. When a client requests the system, the leader disseminates the request, making every follower execute it. If a fault occurs in one of the followers, leading to an inconsistent state (error). Assuming that such error leads to a failure and the system crashes, the whole system's resiliency will not probably fail, as that fault is tolerated by other nodes. This way, the system is more likely to be available.
 
-\subsubsection*{Exercise 2: When does a term start? }
-A new term starts whenever a new leader is elected, or if there is a tie at the leader election process.
+\subsubsection*{Exercise 2:} When does a new term start?
+A new term starts whenever a new leader is elected or a tie at the leader election process.
 
 
+\subsubsection*{Exercise 3:} How do candidates act if two of them attempt to become leaders at precisely the same time?
 
-\subsubsection*{Exercise 3: How does candidates act if two of them attempt to become leaders at exactly the same time?}
 
-When votes are split in a way there is no majority of any candidate, the outcome of a leader election is ``(c) a period of time goes by with no winner'' \cite{raft_paper}. In this case, each candidate will time out and start a new election. A new election includes incrementing the term number and to initiate a new round of voting. There is at most one winner per term; and candidates eventually win, as the new timeouts are chosen randomly.
+When votes are split in a way, there is no majority of any candidate, the outcome of a leader election is ``(c) a period of time goes by with no winner'' \cite{raft_paper}. In this case, each candidate will time out and start a new election. A new election includes incrementing the term number and to initiate a new round of voting. There is at most one winner per term, and candidates eventually win, as the new timeouts are chosen randomly.
 
 
-\subsubsection*{Exercise 4: If a follower receives a message has a different term than the one recorded, does it accept the message? }
+\subsubsection*{Exercise 4:} If a follower receives a message that has a different term than the one recorded, does it accept the message?
 
-If a message received has a greater term than the term recorded by a node, that node updates its term and becomes a follower. If a message received has a lower term than the recorded .
+If a message received has a greater term than the term recorded by a node, that node updates its term and becomes a follower. If a message received has a lower term than the recorded.
 
 %\subsubsection*{Exercise 5: Explain why a client cannot obtain inconsistent messages, depending on the machine that is queried.}
 
-\subsubsection*{Exercise 5: Suppose that a leader receives a command from a client, and notifies all followers. Upon sending the second transaction, one of the followers crashes. How does RAFT guarantee that the follower is updated?}
+\subsubsection*{Exercise 5:} Suppose that a leader receives a command from a client and notifies all followers. Upon sending the second transaction, one of the followers crashes. How does RAFT guarantee that the follower is updated?
+
+After the leader receives a command, the leader propagates that new command information to the followers (uncommitted). Only then he executes the function and returns the result to the client. The followers are then notified, and they execute the command in their state machine. The transaction is considered committed at this point. The follower is updated by following the Log matching property.
 
+\subsubsection*{Exercise 6: Consider the following RAFT cluster, composed of a leader and four followers. The leader has sent eight messages, corresponding to eight commands. Those eight commands translate into eight log entries. The current term is term three. What are the commands, represented by each log index, that are safe to be executed by each node?}
 
-After the leader receives a command, the leader propagates that information of a new command to the followers (uncommitted). Only then he executes the function and returns the result to the client. The followers are then notified, and they execute the command in their state machine. The transaction is considered committed at this point. The follower is updated  through following the the Log matching property.
+The log entries that are considered to be committed are log entries up to log index 7, because they are replicated across most nodes (3 out of 5). Two servers are out of sync.
 
-\subsubsection*{Exercise 6: Consider the following RAFT cluster, composed by a leader and four followers. The leader has sent eight messages, corresponding to eight commands. Those eight commands translate into eight log entries. The current term is term three. What are the commands, represented by each log index, that are safe to be executed by each node?}
+\subsubsection*{Exercise 7:} Consider the following RAFT cluster, composed of a leader and four followers. There are inconsistencies in the logs across nodes. Assume that node S4 crashed on term 1. Node S5 became the leader on term 2, but was only successful in replicating logs 4 and 5, before crashing. Node S3 is the leader in term 3.
 
-The log entries that are consider to be committed are log entries up to log index 7, because they are replicated across the majority of nodes (3 out of 5). Two servers are out of sync.
+RAFT assumes that the current leader is correct (holds the most up-to-date log of commands). S5 will discard its version of logs 6-9, and receive the updated log from S3. S1 and S2 eventually receive logs 8 and 9. As most nodes have the log complete up to log entry 9, the commands present in the log entries can be executed.
 
-\subsubsection*{Exercise 7: Consider the following RAFT cluster, composed by a leader and four followers, in which there are inconsistencies on the logs across nodes. Assume that node S4 crashed on term 1. Node S5 became the leader on term 2, but was only successful to replicate logs 4 and 5, before crashing. Node S3 is the leader on term 3. What will be the final state of the system, assuming no new commands are received, and that S5 recovers from crash?} 
+\subsubsection*{Exercise 8: During the leader election process, a mechanism that ensures consistency across logs is the rejection of candidates that have outdated logs. What are the checks the follower does against the candidate's proposal? On which occasion does it reject voting on the candidate?}
 
-In RAFT, it is assumed that the current leader is correct (holds the most up-to-date log of commands). S5 will discard its version of logs 6-9, and receive the updated log from S3. S1 and S2 eventually receive logs 8 and 9. As the majority of nodes have the log complete up to log entry 9, the commands present in the log entries can be executed.
+The node checks if the candidate's last log entry is in the same term as the candidate's last log entry. If the node's last log entry is in a superior term, it does not vote on the candidate. If the terms are the same, and the node's log index is higher (more complete log), it does not vote on the candidate. This process certifies the elected leader has all the committed entries.
 
-\subsubsection*{Exercise 8: During the leader election process, a mechanism that ensure consistency across logs is the rejection of candidates that have outdated logs. What are the checks the follower does against the candidate's proposal? In which occasion does it reject voting on the candidate?}
+\subsubsection*{Exercise 9: Consider the following RAFT cluster, composed of seven nodes. Given that the leader of term eight has ten log entries and interacts with six followers, how did the system get into this state?}
 
- The node checks if the last log entry from the candidate is in the same term as the candidate's last log entry. If the node's last log entry is in a superior term, it does not vote on the candidate. If terms are the same, and the node's log index is higher (more complete log), it does not vote on the candidate. This process certifies the elected leader has all the committed entries.
- 
- \subsubsection*{Exercise 9: Consider the following RAFT cluster, composed by seven nodes. Given that the leader of term eight has ten log entries, and interacts with six followers, how did the system get into this state?}
- 
-  From RAFT's paper: When the leader at the top comes to power, it is
-possible that any of scenarios (a–f) could occur in follower
-logs. Each box represents one log entry; the number in the
-box is its term. A follower may be missing entries (a–b), may
+From RAFT's paper: When the leader at the top comes to power, it is
+possible that any of the scenarios (a–f) could occur in follower
+logs. Each box represents one log entry; the number in the box is its term. A follower may be missing entries (a–b), may
 have extra uncommitted entries (c–d), or both (e–f). For example, scenario (f) could occur if that server was the leader
 for term 2, added several entries to its log, then crashed before
 committing any of them; it restarted quickly, became leader
@@ -159,8 +157,8 @@ \subsubsection*{Exercise 8: During the leader election process, a mechanism that
 \subsubsection*{Exercise 10: How does the system re-organize itself if a node leaves the system?}
 
 configuration changes must
-use a two-phase approach: the first phase disables the old consortium configuration, disabling the processing of client requests; the second phase enables the new configuration. ``In Raft the cluster first switches to a transitional configuration we call joint
-consensus; once the joint consensus has been committed, the system then transitions to the new configuration''. 
+use a two-phase approach: the first phase disables the old consortium configuration, disabling client requests' processing; the second phase enables the new configuration. ``In RAFT, the cluster first switches to a transitional configuration we call joint
+consensus; once the joint consensus has been committed, the system then transitions to the new configuration''.
 For more details, see Section 6 of \cite{raft_paper}.
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 % YOUR TEXT ENDS HERE