Network services: connectivity and reliability

Network services are characterized by two attributes:

The first attribute indicates whether the network service is connection-oriented or connectionless (Fairhurst et al., 2017).
The second attribute indicates whether the service is reliable or unreliable (Fairhurst et al., 2017).

Based on these attributes, a service can be (Fairhurst et al., 2017):

A Connection-Oriented and Unreliable Service
An example of a connection-oriented but unreliable service is real-time audio or video communication, such as a video call (Schulzrinne et al., 2003).
The service is connection-oriented because a session is established between the two devices before data is exchanged: the call is initiated, accepted, and kept active until it ends.
However, the service can be considered unreliable because it does not necessarily guarantee the retransmission of every lost audio or video segment (Eggert et al., 2017; Schulzrinne et al., 2003).
In real-time communication, retransmitting data that arrives too late would be of little use. If a small audio packet is lost, the system may prefer to ignore or compensate for it rather than interrupt the communication to recover it (Eggert et al., 2017).

A Connectionless and Unreliable Service
An example of a connectionless and unreliable service is the transmission of a real-time location in a mapping, delivery, or tracking app (Eggert et al., 2017).
The device periodically sends small location updates, for example: “I am at this location,” followed a few seconds later by “I am now at this new location.” Each update is independent of the others and is mainly useful at the moment it is sent.
If one of these updates is lost, the system may not retransmit it because newer information will arrive shortly afterward. In this case, receiving the updated location quickly is more useful than recovering an outdated one (Eggert et al., 2017).

A Connection-Oriented and Reliable Service
An example of a connection-oriented and reliable service is uploading a file to a cloud service such as Google Drive or Dropbox (Eddy, 2022).
When we upload a file, our device establishes a logical connection with the service’s server (Eddy, 2022). 
During the transfer, data is sent in a controlled manner so that the server can receive it correctly (Eddy, 2022).
This service is reliable because the file must arrive complete and intact.
 If part of the data is lost or arrives incorrectly, the system must detect the problem and retransmit the missing information. Otherwise, the file may be corrupted or impossible to open (Eddy, 2022).

A Connectionless and Reliable Service
An example of a connectionless and reliable service is an app sending an important notification, such as a banking or delivery notification (Thomson et al., 2016).
The message can be sent as an independent unit: continuous communication between sender and receiver does not need to be maintained. Each notification contains the information required for delivery to the correct device.
However, the service can be considered reliable because the system attempts to ensure that the notification is actually delivered. It may use delivery acknowledgments, retransmissions, or temporary storage: if the device is unreachable at that moment, the notification can be stored and sent as soon as the device comes back online Thomson et al., 2016).

In some respects, connection-oriented services resemble circuit switching: before data is exchanged, a logical connection is established between the sender and the receiver (Fairhurst et al., 2017):

Connectionless services, by contrast, are modeled on the postal system: each letter travels independently of the others; it arrives when it arrives, and it may not arrive at all.
Furthermore, two letters with the same sender and receiver may travel along different routes (Fairhurst et al., 2017; Postel, 1980).
In the image below, a connection-oriented service is shown at the top and a connectionless service at the bottom.

A service is generally characterized as either reliable or unreliable.
A reliable service attempts to ensure that data arrives complete, correct, and in the expected order. To achieve this, it uses acknowledgments, error checking, and retransmissions when necessary (Eddy, 2022; Fairhurst et al., 2017).
Generally, a reliable service requires the receiver to send an acknowledgment to the source for every packet received (Eddy, 2022). 

This reliability introduces overhead because ensuring reliability requires additional computational resources. In some situations, this may be undesirable, for example in a real-time application such as video streaming (Eggert et al., 2017; Schulzrinne et al., 2003).

Conversely, an unreliable service does not guarantee that specific data will actually reach its destination (Fairhurst et al., 2017; Postel, 1980).