IP's role to get a packet/datagram from the source host to the destination host. IP is a packet delivery system; the packet carries some payload in its Data field; this could be an ICMP message but usually is something more interesting and important than error messages or Echoes; usually, it is the data/message of some networking application like Web, Email, Telnet; this message is not carried raw in IP's Data field but is carried by another layer of envelope: TCP or UDP. So IP usually carries a TCP or UDP envelope. Inside the Ethernet frame is the IP packet inside of which is a TCP segment inside of which is the application's "message" which is usually also an envelope of the application.
IP is an unreliable and connectionless datagram service. Packets can be lost, delivered out of order, even duplicated. IP is a best-effort protocol: it tries to deliver its payload, if it succeeds, good; if it fails, it doesn't care or do anything about it! (except for maybe an ICMP error message from a router). IP is essentially a one-way fire the packet into the network and forget about it. Doesn't know what its carrying.
So some additional value-added service/protocol is needed to deal with these issues. That's TCP. (For realtime and streaming applications e.g. VOIP and video, and for some short request and response protocols like DHCP, DNS, SNMP, NTP they aren't dealt with, these applications use UDP). TCP turns the network into a reliable and connection-oriented service that guarantees delivery of message by using a complex system of acknowledgements and sequence numbers. It also performs a flow and traffic congestion role (even more complex).
Positive acknowledgment with retransmission is used to guarantee reliability of packet transfers. The receiver responds with an acknowledgment message as it receives the data (or after it receives some multiple segments worth of data; this is the flow control feature of TCP specified by the automatically-adjusted "window" size). The sender keeps a record of each packet (or, more precisely, segment) it sends. The sender also maintains a timer from when the packet was sent, and retransmits a packet if the timer expires before the message has been acknowledged. This timeout value is adjusted by TCP to accommodate fluctuating network traffic congestion and is a key factor in the functioning of the network (and Internet).
Web, email, telnet/ssh remote login, ftp are some application protocols that use TCP. The goal is accurate "file transfer" activity, where every bit and byte of the message must be delivered. All of the (multi-packet) message must arrive and be reassembled before being given to the application. TCP is another 'envelope' and so has a header with various fields, including a Data field that carries the payload (the [portion of] web page, email, file etc). To identify the application/protocol of the payload, ports are used. (These are software constructs that are completely different from hardware NIC ports.) TCP header's Source port field indicates the sending application, the Destination port field indicates what application at the destination host this TCP segment is intended for.
Analogy is that IP network is the street, IP host is the house, the ports are the doors (up to 64K of them!). Behind a door is a particular application. Servers listen at a well-known port (e.g. web HTTP server at port 80, telnet server at port 23, ftp at port 21, SMTP email server at port 25) so that a client knows to whom (i.e. which server/listener) to send the message. Clients such as a web browser or telnet client are assigned an unused port number that will be used just for the duration of this networking activity. These short-lived ports are called ephemeral ports.
The TCPs on the two hosts establish a connection so they "know" they are communicating back and forth.They can determine that all data sent was received.
Analogy of establishing a conversation, conversing back and forth, then ending the conversation. A host can be in many simultaneous conversations (i.e. TCP connections) with another or many other hosts.
netstat                  show all connections
netstat -n                  numerically rather than with port names and DNS names
netstat -f               don't truncate the names to fit column width
netstat -a               show connections and listening servers
netstat -aob               w/PID and program
netstat -ano             show all connections and listeners numerically w/PID
netstat -p TCP           restrict to TCP (no UDPs)

TCP segment

Options: 0-40B
Checksum: on header and data.
SYN flag Let's synchronize sequence numbers. Only the first packet sent from each end has this set.
ACK flag All packets after the initial SYN packet sent by the client have this set.

TCP 3-way handshake

Example of how two hosts' TCPs establish a connection. Each application on hosts has separate connection. Creates a kind of virtual circuit between the two TCP peers.

1. SYN: The active open is performed by the client sending a SYN to the server. The client sets the segment's sequence number to a random value A.
2. SYN-ACK: In response, the server replies with a SYN-ACK. The acknowledgment number is set to one more than the received sequence number i.e. A+1, and the sequence number that the server chooses for the packet is another random number, B.
3. ACK: Finally, the client sends an ACK back to the server. The sequence number is set to the received acknowledgement value i.e. A+1, and the acknowledgement number is set to one more than the received sequence number i.e. B+1. Both the client and server have received an acknowledgment of the connection.
Steps 1 and 2 establish the connection parameter (sequence number) for one direction and it is acknowledged. Steps 2 and 3 establish the connection parameter (sequence number) for the other direction and it is acknowledged. With these, a full-duplex communication is established.
NB. Wireshark reports the initial sequence numbers as 0 for our ease-of-use.