Peerster v2.0 is a peer-to-peer file sharing application based on structured overlay network. It allows downloading files either by knowing the file owner's ID and the filename or by a keywords-based search (WIP).
For the implementation of the overlay network, we use Chord. Peerster v2.0 has now the following properties of chord:
- Every peerster will store in its finger table at most
O(logN)nodes, where N is the number of nodes in the network. - As well, every search for ID will take at most
O(logN)"messages" (seetests/test4.sh)
In the picture below there are four Peersters. One of them is the genesis node, let's say A. Nodes B, C and D will join the ring network knowning one existing node. We can interact with a particular node using a client program or a web interface.
Since Chord original algorithm uses Remote procedures calls, we decided to use gRPC, a high performance, general purpose Remote Procedure Call library. For Chord control messages this project uses Unary RPC, but for file download and upload we are using the Client/Server streaming RPC (see gRPC concepts ). This library provides a simple interface and a fast development environment. Because the RPCs are using TCP, our Peersters will reliably transmit files between each other and it will also avoid congestions.
Our implementation is offering a O(logN) ID search time complexity in a Chord ring network, the graph below was obtained by runnning multiple (N) Peerster processes on the same localhost. Considering this, we can see a lower bound of just 0.5ms for the Find Successor query. Testing for various values of N, reveals the logarithmic curve below:
We create a distributed hash table using our Chord overlay. Every node
will be responsible for keys, such that hash(node.predecessor.IP) < hash(key) <= hash(node.IP).
In order to support keyword-based search, we will be implementing the algorithm sketched out in this paper. We will build a (distributed) inverted index tree. In other words, we map each keyword to a node in the DHT, which will store a list of documents containing that keyword.
In order to build the Peerster itself (as a node), you need to run go build in terminal and run the program. Use --help to get information
what flags you need to provide to the Peerster:
$ go build
$ ./CS438_Project --help
Usage of ./CS438_Project:
-checkPredecessorInterval int
Number of seconds between two runs of CheckPredecessor Daemon (default 1)
-create
Pass this flag to create a new Chord ring
-existingNodeId string
The id to which this node should join
-existingNodeIp string
ip:port for the existing Peerster in the Chord ring to join
-fixFingerInterval int
Number of seconds between two runs of FixFingers Daemon (default 1)
-ipAddr string
ip:port for the Peerster (default "127.0.0.1:5000")
-join
Pass this flag to join to an existing Chord ring
-m int
Number of bits in one node's id; max = 256, min = 4 (multiple of 4) (default 8)
-name string
name of the Peerster
-stabilizeInterval int
Number of seconds between two runs of Stabilize Daemon (default 1)
-v more verbosity of the program
In order to give commands to your node, e.g., "look up this file" or search "hello world.txt", you need to build the client. From the root folder of the project run in terminal:
$ cd client
$ go build
$ ./client --help
Usage of ./client:
-ID string
Download: File owner's ID / FindSuccessor: ID for which the IP is requested
-PeersterAddress string
Peerster address to connect to
-command string
Command to be sent to Peerster: download/upload/findSuccessor/search
-file string
file name at owner
-nameToStore string
Name used to store the downloaded file
-query string
Search query (required for search command)
-v verbose mode
-withDownload
Used with search command if you want to download one of the found results
Thank you for taking time to check out our project!