Future Internet Architectures Report
Professor Shweta Jain
STAR – Storage Aware Routing Protocol
The storage aware routing protocol or STAR as it is more commonly referred to was researched and developed by Dr. Shweta Jain of CUNY York College, Snehapreethi Gopinath and Dipankar Raychaudhuri of Rutgers WINLAB. It was designed to improve the efficiency of generalized delay tolerant networks or gDTN as it more commonly referred to; and to facilitate with future internet architectures as mobility becomes more essential.
STAR or anything similar will prove to be vital due to the fact that it supports mobility throughout the network in which it will be implemented in. The current internet architecture was designed with IP being the narrow waist that everything goes through from the application layer to the physical layer. But this internet architecture was not designed to support mobile devices that move and change location dynamically. And as we know mobile devices which are connected to the internet are becoming a huge part of global internet, vastly out numbering computers and other devices which remain at one location. This trend will continue and mobile devices will generate the majority of the future internet traffic load.
Wireless networks face many challenges to implement and functions successfully at all times due to the fact that there is no hard-wire connection and devices in the network move around. Many things can happen to disrupt communication; wireless signals like wifi can be lost due to distance as devices move throughout the network, other frequencies my cause channel disturbances, and link outages due to the drop in link speed. Also some devices may disconnect from the network when their batteries are running low in order to conserve power. The term gDTN which means generalized Delay Tolerant Networks was coined to described these network scenarios. Generally when these interruptions take place the entire piece of data that was being transferred at that time would be lost and would have to be retrieved all the way from the host again. This means that the routers would have to find the path back to host or server where the original content came from, because the data wasn’t stored anywhere on its path to the content receiver. Although it may not take that long to reconnect to the host the more disconnections happening the time will definitely add up and be noticeable. Worst yet, the connection between the original content creator or server cannot be re-established when the content requestor reconnects to the network, this causes serious problems and lack of functionality in the network. But STAR will remedy these problems.
Currently future internet researchers are have been working to develop new technologies that will further accommodate wireless networks and integrate them as a part of the future internet architecture. So far they have proposed a new transport protocol that would provide hop by hop routing instead of end to end routing. But these innovations at the upper layer will only work efficiently if the network routers and protocols are created or upgraded to support wireless networks and mobility throughout the network. So this is where STAR comes in with its new network protocol that makes routers smarter to provide the best quality data transfer throughout the network.
Star will put into effect policies that will work seamlessly between disconnected wireless networks to well-connected wired networks. The objective is to always find the best end-to-end path to the destination and knowing how and when to transmit the data. The main elements needed in order for star to function are: a storage router, a transport layer, a routing cost metric, network topology discovery, path computation, and a store and forward algorithm.
Star introduces the design of a storage router that will be used to allocate storage for all layers of the protocol stack. Routers implementing star would be built with large storage capacities which would be used to store large content in transit. These routers will keep the information on all the data requests up stream and all the actual data coming down stream before it is forwarded further. Different content can be stored and cached if necessary for future use if the router constantly requests it; this means that the routers can be smart keep data the nodes closest to it requests regularly instead of going to the host or original content creator all the time. This can be implemented easy as the cost of storage is decreasing rapidly and companies are creating bigger hard disks.
The transport layer currently being used which is TCP has been proven to be in efficient in general wireless networks due to the fact that the connections between wireless devices terminate at the access point(AP). As mobile devices move around they may connect to another AP which would cause the data transfer to stop and need to restart from the original server unless the wireless device can find a way to connect to the original AP that was being used when the transfer initially started. Therefore star uses the hop by hop protocol which is reliable and doesn’t require connections. In hop by hop the data is transferred by using the selective repeat ARQ scheme. Instead of sending an acknowledgement every time a frame is transferred a batch acknowledgment(BACK) is used when everything is done transferring. The back contains a field that tells the server what frames need to be retransmitted if any.
The routing cost metric is important as it is used to determine when it is best to the link between the server and the client is best and optimized for data transfer. Star uses a two dimensional routing cost metric which has both short term and long term route costs. Star constantly checks the state of the link to know when the bandwidth and signal is at its best, it keeps a record of the average highest a link was to know that it was capable of that before, so the router would wait for the link to get to its optimal strength or close to it before sending the data. The routing cost metric also factors in the distance and time it would take for each transfer. The less time a transfer takes the lower the cost, the highest the link strength is the lower the cost, the shorter the distance is the lower the cost. Star implements tables in the router that keeps the records of all the nodes directly connected to it, and their neighbor nodes. This table is always being updated so the router can know information about its neighbors and the path to other nodes in the vicinity. The tables are updated by using the information received from a broadcast of hello messages. These hello messages contains information such as a list of the node that sent the message’s neighbors and best transmission rate for each neighbor, as well as the available storage on itself and its neighbor nodes if that router already has such information. Using the information received from the broadcast messages each node creates their one and two hop neighbor tables.
With this information star then chooses the path with the average lowest costs when all the variables are factored in and taken into consideration and uses that path to send the data. Star uses a modified breadth first search to find multiple paths with low costs to each destination. Star works smarter and not harder, by finding the path with the lowest to one destination very far it can also learn the best routes to various destination nodes along the way. Star saves that information for later use if necessary.
After figuring out the best possible route to a node, star then has to make the decision to either forward or store the data. Before forwarding data each router first looks up in its table the available storage on the multiple routers downstream in route to the destination node. If there isn’t sufficient storage capacity on the next router downstream or the next few, the router will not forward the data. It will instead keep the data in its own storage to be forwarded later when there is sufficient storage space on the routers downstream. Routers also store data when it doesn’t know how to get to the destination node, if the node at the next hop is disconnected for any reason, or if the signal/link between the nodes is not optimal for data transfer.
Overall I think star is going to be a huge part of the future internet architecture as it mends and deals with quite a few issues of the current internet architecture. It works genius to find the best ways to transfer data to its destination. It adds versatility to the internet to support mobile devices moving throughout the network and disconnecting from time to time. If the devices are disconnected or are unable to be reached the data being sent to them can be stored to be distributed later.
I think the biggest issue star will face is storage management on the routers. How long should the data be stored? What data is important to keep and what is not? As much as storage is getting cheaper, in a large network routers storage may get filled from time to time. So I think there should be a system in place to determine what data is important and what can be deleted in such a situation when storage is minimal. What if the data being stored was going to a device once connected to that router and that device has since moved and the device is better off getting the data from the original content creator. As long as this part is figured out star will work great and improve the internet architecture substantially.
STAR: Storage Aware Routing Protocol for Generalized delay Tolerant Networks by Shweta Jain, Snehapreethi Gopinath, and Dipankar Raychaudhuri.