Echo_rd

An Erlang library that implements the chord P2P system. For more details see there https://pdos.csail.mit.edu/papers/ton:chord/paper-ton.pdf.

Build

$ rebar3 compile

Interface

The library allows to abstract the location of a service and distribute it. All the commands of the service are not changed by Echo_rd, it just routes them to the node handling it correctly. The library will need a block service providing an interface compatible with gen_bm.erl file.

An example of an implemented block service can be seen here: https://github.com/GiacomoZiffer/Eblock_rd

The full list of methods complete with a brief explanation can be seen in the gen_app.erl file.

Modules

• application_manager: it will handle the communication with the upper level (application and block manager);
• checker: it will handle the communication with the predecessor on the network. It will also handle the address (and id) of the predecessor;
• communication_manager: it will encode and decode incoming/outcoming messages and route them properly;
• fixer: it will make lookups in order to update the finger table entries;
• hash_f: it will provide methods for the hashing of addresses and resources;
• join_handler: it will handle the status of the node regarding join/create/leave;
• lager_sinks_handler: it will provide methods for the correct execution of lager logging;
• link_manager: it will create/manage tcp sockets;
• logging_policies: it will handle the filters on the logging based on the chosen policy;
• lookup_request: it will handle a single lookup request waiting for response;
• naming_handler: it will handle all naming operations;
• naming_manager: it will handle the naming table when naming_handler doesn't exist or is restarting;
• normalizer: it will handle all the operations on nodes id;
• params_handler: it will handle values needed for the execution after entering a network;
• request_gateway: it will handle and create all the lookup_request present;
• router: it will handle the finger table and the lookup mechanism;
• socket_handler: it will handle a single tcp connection;
• socket_listener: it will listen for tcp connections;
• stabilizer: it will handle the successor list and handles the communication with the successor;
• statistics: it will provide useful metrics to analyze the network.

Configuration

Echo_rd can be configured using

$ application:set_env(echo_rd, #VAR#, #VAL#).

The following table shows the possible configurations (in square brackets default values):

#VAR#	#VAL#	Meaning
ip	public	Set this to use public ip address instead of the private one
ip	[private]	Use private ip address
log	all	Set this to use the maximum level of logging
log	logic	Set this to log only the events strictly related to chord and join mechanisms
log	comm_only	Set this to log only the communication events
log	chord_only	Set this to log only the events strictly related to chord
log	joiner_only	Set this to log only the join mechanisms
log	naming_only	Set this to log only the naming operations
log	[undefined]	Log only errors
lager_log	lager_on	Logs are written into file and printed on screen (may make it hard to use the terminal)
lager_log	lager_only	Logs are written only on file
lager_log	lager_off	Logs are written only on terminal (may make it hard to use the terminal)
lager_log	[undefined]	No logging
connect	NUMBER ([7000])	Maximum amount of time that can be waited during a tcp_connect without giving timeout
delay	{constant, NUMBER}	Add NUMBER ms of latency to communication
delay	{normal, MEAN, VAR}	Add latency to communication based on a normal distribution with mean MEAN and variance VAR
delay	[undefined]	No artificial delay
fix	off	Stops the fix process (cannot be restarted)
fix	[undefined]	Fix as normal
stabilize	off	Stops the stabilize process (cannot be restarted)
stabilize	[undefined]	Stabilize as normal
fail_detection	off	Stops the fail detection system (Needs also the stabilize off option for proper use)
fail_detection	[undefined]	Fail detection as normal
size	NUMBER ([50])	Dimension of the buffers for statistics
timing_increase	NUMBER ([2])	After each failed lookup, average timing is multiplied by this number

Messages

Message	Params	Created by	Forwarded to	Notes
lookup_for_join		join_handler	router
lookup_response	ID, Address, Length	router	request_gateway, join_handler	ID contains the requested id, Address will points to the closest successor of the requested id, Length the number of hops for the lookup
ready_for_info		join_handler	join_handler
join_info	Nbits, SuccessorList, Resources	join_handler	join_handler	Contains all the info needed for the join
ack_info		join_handler	join_handler	Confirms the reception of join_info
abort	Reason	join_handler	join_handler	Aborts current join operation
ack_join		join_handler	join_handler	Ends the join operation
leave_info	Resources	join_handler	join_handler	Contains the resources of the leaving node
ack_leave		join_handler	join_handler	Confirms the reception of leave_info
ask_pred		stabilizer	checker
pred_reply	Predecessor, SuccessorList	checker	stabilizer	Responds to the ask_pred message
lookup	ID, Hops	router, lookup_request	router	Hops indicates the max number of nodes that can be visited
command	Address, Command	application_manager	application_manager	Address contains the ip address of the node sending the command
get_stats	Number	statistics	statistics	Number contains the number of already visited nodes
stats	Statistics	statistics	statistics

• 1 is Api (l), Router, Fixer(l)
• 2 is Api (any)

Truth-Table of the CM's state machine

ID	Curr_State	Action	Re-Action	New_State
a	init_joiner	join	lookup_for_join	look
b	init_joiner	create	start	init_provider
c	look	look_resp	ready_for_info	pre_join
ti	look	timeout	hard_stop	init_joiner
d	pre_join	info	ack_info	j_ready
e	pre_join	abort	lookup_for_join	look
ti	pre_join	timeout	hard_stop	init_joiner
f	j_ready	ack_join	start	init_provider
g	j_ready	abort	lookup_for_join	look
ti	j_ready	timeout	hard_stop	init_joiner
h1	init_provider	ready_for_info	join_info	not_alone
h2	init_provider	ready_for_info	abort	init_provider
i	init_provider	leave_info	ack_leave	init_provider
j	init_provider	leave	leave_info	leaving
l1	not_alone	ready_for_info	abort(joiner)	not_alone
l2	not_alone	ready_for_info	abort(curr), join_info(joiner)	not_alone
m	not_alone	ack_info	ack_join, drop_many_resources	init_provider
n	not_alone	leave	abort(curr), leave_info	leaving
o	not_alone	leave_info	abort(curr), ack_leave	init_provider
tj	not_alone	timeout	hard_stop	init_provider
p	leaving	ack_leave	stop	init_joiner
q	leaving	ready_for_info	abort	leaving
tl	leaving	timeout	hard_stop	init_joiner

Where:

• The ID field column the id of the event
• The Curr_State column represents the current state, i.e. the state from which the transition comes from
• The Action column represents the incoming request
• The Re-Action column represents the state-machine’s response w.r.t. the corresponding action
• The New_State column represents the new state after the event

Observation:

• For the id of the timeout scenarios is used the following notation:
  • ti for the default interval
  • tj for the join interval
  • tl for the leaving interval
• For the ready_for_info message we have the IDs h1,h2 and l1,l2 according to the priority of the joiner who has to be served.