We Keep Coding

Need suggestions for reliable data broadcasting inside LAN using Java or android

We are working on a Android project with the below requirements.
The application should be able to send data to all the devices which are running our application which exists in the WiFi LAN.
Some payloads are expected to be of size >= 5MB.
The data shouldn't be lost and if lost the client should know the failure.
All the devices should be able to communicate with all other. There will be no message targeted to a specific device instead all the messages should be reached all the devices in the N/W.
No internet hence no remote server.
Study we have done:-
UDP Broadcasting - UDP doesn't guarantee the message delivery but this is a prime requirement in our case. Hence not an option.
TCP - TCP guarantees the message delivery but requires the receiver IP address to be known before hand and in our case we need to send the message to all the devices inside the LAN. Hence not a straight option.
Solutions we are looking into:-
A Hybrid approach - Name one of the devices in the N/W as Server. Post all the messages to a local Server. The Server keeps a open socket to all the devices(which have our application) & when there is a message from a device then it routes the message to all the devices. The disadvantages of this approach are,
Server having multiple sockets open each per device. But in our case we are expecting devices <=5 in LAN.
Server discovery using continuous UDP broadcast.
We want to have all the data in all the devices. So if we newly introduce any device into the LAN then that device needs to get all the data from the server.
So my question, have you any time worked on these kind of hybrid approaches? Or can you suggest any other approaches?

Your hybrid approach is the way to go.
Cleanly split your problem into parts and solve them independently:
Discovery: Devices need to be able to discover the server, if there is any.
Select server: Decide which of your devices assumes the server role.
Server implementation: The server distributes all data to all devices and sends notifications as necessary. Push or pull with notifications does not matter.
Client implementation: Clients only talk to the server. The device which contains the server should also contain a normal client, potentially passing data to the server directly, but using the same abstract protocol.
You could use mDNS (aka Bonjour or zeroconf) for the discovery, but I would not even recommend that. It often createsmore problems than it solves, and it does not solve your 'I need one server' problem. I would suggest you handcraft a simple UDP broadcast protocol for the discovery, which already tells you who the server is, if there is any.
Select server: One approach is to use network meta data which you have anyway, for example 'use the device with the highest IP address'. This often works better than fancy arbitration algorithms. Once you established a server new devices would use this, rather than switching the server role.
Use UDP broadcast for the discovery, with manual heuristic repeats. No fancy logic, just make your protocol resilient against repeated packets and repeat your packets. (Your WLAN router may repeat your packets without your knowledge anyway.)
You may want to use two TCP connections per client, potentially to two different server ports, but that does not matter much: One control connection (always very responsive, no big amounts of data, just a few hundred bytes per message) and one data connection (for the bulk of the data, your > 5 MB chunks). This is so that everything stays responsive.

Receive UDP Packets on Google Compute Engine

my task is to receive UDP packets on an Google Compute Engine. Given is the port, which is 300 and I already have a simple Java program which can handle these UDP packets.
My problem now. Get the UDP packets to my Java program.
For that I don't know how to do this.
What I have tried so far ist to let the Java program direct listen to the port, which didn't work. I suspect the traffic from outside the GCE must be routed to the inside?

First make sure you've added a firewall rule on your GCE VM instance's network which allows incoming traffic for UDP protocol on port 300.
You can do this by going to the Developers Console, select your project, then Compute-> Compute Engine-> Networks, click on the right network and verify firewall rules (and tags as well if you used them). If the rule is not added, then add a rule for the traffic.
"gcloud" is a command-line tool which you can use it to list and verify your firewall rules as well [1]:
$ gcloud compute firewall-rules list
If the firewall rules are good, then use a simple troubleshooting tool like netcat to test if traffic is being forwarded to your VM instance.
1. Listing Google Compute Engine firewall rules

I recommend not using Java to process UDP. I experienced inexplicable short periods where all UDP traffic arriving was being lost. This was caused by garbage collection.
The architecture that works for me is to use a GCE VM with a C++ app receiving UDP data, then using libCURL, convert the data to HTTP and dispatch it to Google App Engine for processing. To make the UDP receiver scalable and tolerant, use network balancing and multiple VMs listening for UDP.

UDP over Internet. How does it work?

As I am programming a network chat (java, but should not make a difference for the question), and wanted to use UDP, I ran into the problem of it not working over the internet. After a little research I found out that you have to have port forwarding for the specific port activated. So now it comes to my question:
Does UDP work over the Internet in a not configurable way?
For example, if I would program a whole Network Game would it make sense to use UDP? Or would I require the Player to activate Portforwarding and open the Port etc?
When would it make sense to use UDP then? And why?
I'm actually not understanding the whole point of UDP then.
For my programming point of view I would like to have a way to use it intuitive.
Like creating the DatagramSocket and the DatagramPacket, configure the Packet with the Data and the Destination and send it away over the internet.
As for my Users I don't want them to have to configure any specific things like opening the exact port they want to use etc. I just want them to use the program (server and client) and it should work.

The problem you've run into is not one of UDP vs TCP (although using the unreliable, unordered UDP as the basis of a chat application seems like an odd choice to me).
The problem is that of NAT traversal. In a nutshell, home routers perform a network function called NAT - Network Address Translation. They do it in order to use a single public IP address for all machines inside the NAT (which are given private addresses - usually 10.0.0.0 or 192.168.0.0). The router then switches the source IP address in all packets sent from inside the LAN from the private address to the public one. It uses port numbers to "remember" which machine sent what to what address, in order to perform the backwards translation when the response arrives.
The problem arises when someone wants to initiate a connection to a machine behind a NAT. Without seeing an outgoing connection first, the NAT doesn't know to which internal computer and port it should forward the packet. This is what happens to you.
There are various fixes for this issue, with the simplest one being manual port forwarding (as you've discovered), but it's a well known problem faced by any peer-to-peer application. If you need to contact a machine behind NAT (i.e. contact most home users) and you want your application to work out-of-the box (without your users fiddling with their routers) you need to research NAT traversal techniques, implement them in your application, and hope that the user's home routers support them. It's a huge pain in the neck.

EDITED: with Joachim Pileborg's correct suggestions!
UDP is often a better choice for action-based games, where it's vitally important to have updates to the client or server with the latest data about a player, player input, or the game world.
TCP begins with a 3-way handshake to establish a connection (which takes time). If your game communication protocol is via TCP, all packets in a message have to arrive before the message becomes available. Even a small amount of Internet congestion could cause your game to lag.
TCP is good for communications that MUST arrive in full.
With UDP, the client or server can send the latest player/game state in individual packets that do not depend on arriving in order. If a packet is late, or arrives out of order... it should be ignored.
UDP is good for communications that need to be fast, but losing individual packets is OK.
Both should be available in your Java platform.
Here's some further reading:
http://gafferongames.com/networking-for-game-programmers/udp-vs-tcp/

Netty two-way and n-client UDP

I am trying to write an application which will run in many machines, and these machines will communicate with each other through sending a stream of UDP packets. Usually one machine will be sending many packets to another machine, until they switch and another one will do the same to another machine and so forth. Choosing UDP is due to the nature of the application (real-time/speed more important than reliability). I also understand that UDP is a connectionless protocol.
By doing a quick prototype, I managed to create a NIO Datagram Server (bind), and a NIO Datagram Client (connect). However, I realized that I can only send one way only from the client to the server (or maybe I am missing something?). I couldn't send in the opposite direction.
Also, since UDP is a connectionless protocol, I thought that it was supposed to accept many clients sending packets to it (n-client to one server), and the other way around (server sending to clients one-by-one or multi/broad-cast).
Should I create a server in the client listening to a different port to acheive two-way? Can n-client send packets to one server at the same time?
I just want someone to clear this thing to me. No need for sample code (although it will be greatly appreciated), you can give me some pointers.
Thank you.

For what you are describing, you need one server thread and one more client thread in EACH and EVERY instance of your program running on different machines.
Create a multi threaded program, with a serving thread, and a client thread - the client thread needs to know about all servers - IP and Port. The server simply listens on a port on the current machine.
If you run this same program on multiple machines, You will get a p2p group.
You can also set up a tracker on a free server in the internet. A tracker program will rest in a well known URL, and will maintain a list of participating machines. Each instance of your program when started on a machine will update the tracker of details necessary to connect to it, and the tracker can maintain a list of this data and share it all with any new instance coming up later.

Discovering clients on a wifi network

I'm writting a java application, and I need to quickly discover any other running clients on any wired or wireless local networks in order to establish a TCP connection.
what's the best way of doing this? Are there libraries or code snippets that would do this?

Multicast UDP is a good way of doing this. It's used in a couple of technologies that support automatic discovery of networked devices over local IP networks (UPnP and ZeroConf).
Multicast UDP is not TCP, but it is still based on IP and, so, uses the same addressing mechanism i.e. IP addresses. Quite often it is compared to radio broadcasting i.e. a multicast sender only needs to send 1 message (i.e. it is like a broadcast) but only clients that are "tuned-in" to the multicast channel will receive it.
You can do a quick search on google or wikipedia for these as a starter, but the basic idea is as follows:
when a client starts, it sends out a multicast UDP "hello" message to some pre-specified multicast address and port (e.g. UPnP uses 239.255.255.250:1900)
existing clients are listening for incoming multicast "hello" messages on the specified address and port - when a client receives one, it sends a response to the sender
the client sending the "hello" message receives a response from each existing client on the network and is now aware of the IP address of each client
If you are looking for libraries to use, UPnP libraries can tend to be a bit heavyweight and a lot of folk generally don't like them, so ZeroConf might be a little more suitable. I don't know of any java implementations of such things but I'm sure you can find some with a little digging around.

A network scan can be very long, even longer on wireless networks. If you need them quickly thru Java you may implement a "meeting point" server on your network. This server listen to a predefined port, clients register on the server on startup and the server can distribute information about the clients on request.
HTH.

I guess you need to do a scan on your application's port on all IPs in your subnet.
Just what are the available IPs - or what is your subnet for that matter?
I'm afraid determining that could turn out to be impossible as the network is designed to be transparent to your application.
So, i'd use brute force: pick your IP and change the last byte. Might be too much, might be not enough though.
Or you send a broadcast (which usually would be targeted at x.x.x.255) and see who answers.
See Datagram Broadcasting and Multicasts. But i think that's not TCP/IP anymore.

There is a JGroups toolkit for reliable multicast communications. It allows automatic discovery of additional clients using Multicast techniques as described in other answers.
It also provides communication APIs on top of multicast sockets.
It is used in a number of projects such as JBoss, Tomcat and more to provide an infrastructure for distributed cache. See more here.