This paper describes the usage of encryption in FMFT (Fast Manager of File Transfer) program that is based on Xinan Liu's Reliable File Transfer Protocol. The system consists of a server and a client software pair utilizing UDP connection. The aim is to protect big data transfers by utilizing encryption maintaining the high transfer rates. Relying on Xinan Liu's solution a -based minimum viable product was developed and further enhanced later, which was even more improved by rewriting it in C++. By adding a graphical user interface to the client, it is more user friendly now. Furthermore, by using WebAssembly, the program is available for many platforms now. After presenting the performance hit of the usage of encryption on data packets, it will be discussed that thanks to multithreading, our application can utilize the CPU in a better way.
We experienced at file transfers forth and back with a server hosted in the Gyires
supercomputer’s data center [
A User Datagram Protocol (UDP) based software handles big data transfers a
way better than Transmission Control Protocol (TCP) based solutions. There are
UDP based systems as UFTP and UFTPD software pair that mostly accomplish
our needed features [
The structure of the paper is the following: in chapter two the development environment of the fast file transfer manager application is described. In chapter three encryption is presented. The multithreading work is detailed in chapter four. The achievements are enlisted in chapter five. Possible continuation and future research and development aims are described in chapter six.
For the platform independent development the Qt 5.15.1 stable version was
chosen [
The client is built for the necessary target system natively having a GUI (see
Figure 1.). The server is designed to run in headless mode. The client is planned
to be made usable from command line also. After testing our SCTP file transfer
system for desktop operating systems, we found that the performance of the TCP
based FTP transfer and our SCTP and UDP based file transfer implementations
need more investigation [
For testing purposes we used a gigabit network. The first solution to transfer
ifles was using the standard FTP protocol which runs over TCP. Similar to SABUL
[
For encryption Datagram Transport Layer Security (DTLS) was taken into
consideration [
A 100 MB test file was used to carry out the measurements, its data was transferred from the desktop workstation to the test server residing in the Gyires supercomputer’s server room. Results of the consecutive transfers starting from 200 bytes to 4000 bytes of chunk sizes can be seen in Figure 2. It can be observed that without encryption of data packets, the transfer rate stabilizes around 800 Mbps from chunk size of 700 bytes up to 4000 bytes.
1,000
800 ] s p b M [ e a 600 t r r e f s n a r T 400 200
200 1,000 2,000 3,000 4,000
The test machine has an Intel® Core™ i7 CPU 920 @ 2.67 GHz with 18 GB memory having gigabit Ethernet connection, connected to the academic Internet network running Ubuntu 18.04.5 LTS desktop version. The test server was a Cisco UCS C240 M5 having gigabit Ethernet connection, connected to the Gyires supercomputer. The applied virtual machine runs Ubuntu 18.04.5 LTS server version under VMWare’s vSphere 6.7 having 20 GB memory and 8 cores of an Intel® Xeon® Gold 6130 CPU @ 2.10 GHz.
By encrypting the data packets, the transfer rate cross the 800 Mbps boundary with 811 Mbps just at chunk size of 1100 bytes. The curve of results of transfer rates with encryption stays almost entirely under the curve of raw transfer of packages. Exception to this is chunk size of 2200 bytes where encrypted transfer rate of 836 Mbps was better by 1 Mbps over raw transfer rate of 835 Mbps. The diference is mostly under 1.5% at 900 bytes chunk size or higher. This result was consistent over various sized test data. For small chunk sizes as 600 bytes the diference can be as high as almost 20%. At a very small cost SimpleCrypt yields a good protection against attacks coming from third parties on the same network segment.
SimpleCrypt has a compression option and we tested it also. It was found that the packet size for text files could by reduced up to 50%. However, the typical data for our use case is binary, and for this kind of data usually the gain was generally around 5%. In the worst cases the compressed data was even bigger than original packet. Because of the increased stress on the CPU besides the small gain, this was not tested thoroughly. The next chapter describes our work toward multithreading. When it will be implemented more extensively in our programs, we will move this feature in a distinct thread and will test it deeply.
Multithreading was introduced into the client software first, because at long workload the graphical user interface became unresponsible. Separating the GUI from the work thread, the GUI elements as progress bar or buttons started to give appropriate feedback. This could be achieved by using the signal-slot system of Qt. The next essential change was to separate the connection setup from the data transfer. This was done for the server also. Having a distinct thread for the data transfers was a huge boost especially for the server program, because one thread has a practical physical limit for the number of the clients to be served. Modern servers are massively multithreaded nowadays, thus the number of served clients can be raised this way.
To optimize handling of new connections, the connection setup and data transfer was split into two data ports. For each port a distinct thread is started up. Consecutive data transfers from the same client can disturb each other if packets from the first transfer are so late that they arrive during the second data transfer. To minimize the packet mixing, the data transfers got a data port range of 256 elements. This range is used as a circular array, thus giving enough time for wandering packets to arrive. 256 is practical because an unsigned byte can be used to index it without computing modulo values each time. In this way, even one client can initiate multiple data transfers concurrently to the server at the same time. The server is also having a data port range having an own thread assigned for each of them.
Relying on Xinan Liu’s work a high-speed connectionless file transfer system was developed by using UDP control channel. On busy connections data transfers were sped up over 800 Mbps by utilizing multithreading among other factors. There is always a chance for a man-in-the-middle attack. Even malicious nodes anywhere in the network chain can tamper with packets or the integrity of the data transfers can be damaged by accidental bit alternations. CRC checking fights eficiently against packet modifications, but do not hide the transferred information. Encryption is a good tool for protecting sensitive data. By using SimpleCrypt to mitigate threats, the performance cost was generally under 1.5% at 900 bytes chunk size or higher. This way we recommend to use it all the time.
Datagram Transport Layer Security (DTLS) is a transport communication
protocol. It ensures security for datagram-based applications by letting them to
communicate in a way that is designed to prevent eavesdropping, tampering, or message
forgery [