Request IPFS network
Last updated
Was this helpful?
Last updated
Was this helpful?
Request uses IPFS to store transaction data in a decentralized way. Transactions are actions done on a request, for example: create, accept, reject...
When files are stored in IPFS, they are kept locally on the IPFS node, and are accessible by any node connected to the network. IPFS creates a unique hash used as an identifier to access a file. We store this hash on the Request smart contract to have a trustless list of transactions.
The main IPFS network has tens of thousands of nodes and many files. All of the Request transactions are a tiny fraction of the IPFS network. To find a transaction file, the Request Node IPFS has to on the network. Content retrieval is currently relatively slow on IPFS.
By creating a dedicated IPFS network for Request, our network is isolated from the rest of the IPFS network. All the Request IPFS nodes will only communicate with other Request Nodes. By keeping the network small, we can make sure most nodes are connected between themselves, and asking for files can be done directly to a node instead of through a traversal.
This is a big advantage for us because our nodes' end goal is different from that of a normal IPFS node: all the IPFS nodes used by a Request Node should have all the files on the network. This makes the DHT pointless, and the most critical factor in discovery time becomes how many nodes every node is connected to.
These are the main reason why we created the Request IPFS network.
IPFS has a feature called private network. It allows IPFS nodes that share a private key to communicate only among themselves and keep their files private from the open IPFS network. We use this feature to create an IPFS network separate from the open one, but we keep this key public. This way, we have a separate public network from the open network.
We also changed some default IPFS configurations on our network, to improve performance and responsiveness. The main change we did is disabling the DHT, so instead of traversing the network to find a file, the nodes will only ask to their neighbor nodes for those files. Since on our network every node is supposed to have every transaction file, those responses tend to be a lot faster (on most of our test cases the response time went from seconds to a few hundred milliseconds).
We distribute a Docker image of our configured IPFS node. To run it, you can use the command:
There are two easy ways to connect an IPFS node to the Request Network:
If you want to set up your IPFS node yourself, here is the information you would need:
The swarm key
you should put this file at your IPFS path (usually in $IPFS_PATH)
swarm.key file content:
The configurations
Use the docker image to run your IPFS. It comes pre-configured and you just need to run it.
script available from the or the .