- A study by the Research Center Munich and the Technical University of Vienna concludes that the IOTA mainnet with the Coordinator is not more performant than a blockchain.
- The study was conducted on the basis of historical data and shows the necessity of launching Coordicide.
Academics of the Munich Research Center, Huawei Technologies and the Distributed Systems Group, Vienna University of Technology have published a study entitled “Characterizing IOTA Tangle with Empirical Data”. In their scientific work, Fengyang Guo, Xun Xiao, Artur Hecker and Schahram Dustdar investigated the performance of the IOTA Tangle in the mainnet.
As the study states, IOTA organizes the transactions in a ledger as a Directed Acyclic Graph (DAG). As a result, the technology is considered “a promising platform to support Internet-of-Things (IoT) applications with its key features such as micro-payment support and absence of transaction fees”.
While, according to the researchers, there are extensive analyses based on synthetic data generated by simulations, an analysis based on empirical data on the IOTA mainnet is missing. Consequently, “many questions remain open, such as the real tangle topology, the actual transaction confirmation rate in the deployed system, and, in case of diverging findings, the reasons behind the present observations.”
The performance of the IOTA Tangle in mainnet
With this in mind, the researchers conducted a “comprehensive analysis” (322 GB of data sets) using real transaction data “officially published by the IOTA Foundation”. The researchers came to the following conclusion:
Our key finding is thatneither the tangle’s topological features nor the actual observed performance is consistent with the main conclusions from the literature.
To investigate the performance of transaction confirmation, the researchers considered a transaction as approved “as soon as this transaction is attached to the tangle by a new site. However, according to the definition of IF, a transaction is in reality only considered approved if it is approved by a milestone”.
In relation to this, the researchers noted that before May 2018, the rate of milestone emissions varied between several hundred and 1800 per day. Since September 2018, the rate of emissions has slowed to 500 per 12 hours and 1500 per day. Subsequently, the researchers also investigated the distribution of transaction confirmation times and found a median value of about 10 minutes. “This also applies to 25% to 75% of transactions in the quantile,” the researchers said.
In total, the researchers made three core observations. Firstly, the IOTA mainnet generates a Tangle that has different topological features compared to the simulated tangles. The “real IOTA Tangles show a power-law degree distribution rather than a Poisson distribution as in simulated cases”. Secondly, the investigation states that:
Nodes in the real IOTA indeed use various TSAs to attachnew sites into their local tangles. Milestones are selectedmore often than usual sites. Abnormal sites were observedin reality (cf. the result of EW Analysis), which are notsimulated in most of the prior art, where nodes perfectlyfollow the IOTA specification.
Finally, the researchers also found that the transaction confirmation rate is not as high “as usually” assumed when milestone confirmation is required, which requires the assistants of milestone sites. Ultimately, therefore, the researchers conclude that the “usual assumption” that IOTA can provide a much faster transaction rate than traditional blockchains is not true.
In particular, most of transactions take roughly 10minutes to be officially confirmed, which is not exactly instantas commonly assumed; yet, what is arguably worse is that there isa certain amount (5%) of transactions experiencing exceptionallylong confirmation time […]
This is far behind therequirement to support lightweight, rapid and instant IoTapplications that are delay-sensitive, especially consider-ing those exceptionally delayed transaction cases.
In the overall result of the study, the researchers therefore come to the following conclusion:
According to the presented results,our key findings are that the features of the real IOTA tanglesare topologically different from the simulated tangles; moreimportantly, the transaction confirmation time largely dependson the milestones issued by COO.
In addition, it is inefficientto rely on the mechanism of using site cumulative weight inthe random walk of TSA, which aligns with the recent plans ofthe IOTA Foundation. We hope that the presented results canprovide a better understanding of the nature of the real IOTA andmotivate to continue further analysis in the IOTA community.
However, it is important to note that the study is based on historical data from the IOTA mainnet (before the implementation of IOTA 1.5 phase 1). With the abolishment of the coordinator and the already proposed changes for the upgrade to IOTA 2.0, the collected data will be irrelevant and, according to the simulated data of the IOTA Foundation, will make the Tangle operational for use in the corporate sector and in IoT.
