260

S. Das et al.

Fig. 10 Graphicalcomparisonofcomputationtimeperblockgeneratedofdifferenthashalgorithms

for different tiers of devices

of bandwidth starts becoming apparent. Also, lack of software support, as well as

comparatively less security and restrictions, in its design makes it hard to optimize

existing software for lightweight hash functions. The need for higher bandwidth

thresholds and the additional security of more complex hash algorithms like Blake2s

and Keccak makes them more suitable from this tier onwards with computational

time per block less than 10 s. We opted for Blake2s, as it was very well optimized

for devices of this tier, many of which use Intel Skylake processors, which is also

reflected in the sampled data.

For Tier IV, two custom desktops with configurations (4.9 GHz AMD Ryzen™

9 5900X, 32 GB RAM, Nvidia Geforce RTX 3080) and (5.0 GHz Hexa Core

Intel® Core™i7-8700 K OC, 16 GB RAM, Nvidia Geforce GTX 1060) are used

in the sampling process. At this level of computational power, most secure algo-

rithms will have very respectable computational time per block due to the abun-

dance of processing power. We opted for Keccak256 in this tier based on the lowest

computational time for 256 bits.

If we look back at the resulting output in our implementation from Figs. 5, 6,

7, and 8, we can see the computational time per block generated achieved in the

simulation are as follows:

Tier I: 3.0035 ms, Tier II: 1.4865 ms, Tier III: 9344.6439 ms, Tier IV:

2062.9772 ms.

These values are well within of scope of the analytic results achieved in the

statistical comparison of different algorithms per tier. Hence, our implementation of

the proposed model of blockchain architecture is proven to be efficient for IIoT.