Blockchain sharding networks partition the network into
parallel processing units to overcome the throughput
limitations of monolithic chains. Configuring such a
network requires balancing two conflicting objectives –
transaction throughput and latency, that cannot be
optimized independently. This paper applies the
Non-dominated Sorting Genetic Algorithm II (NSGA-II) to
approximate the Pareto front of sharding configurations
defined by three decision variables: number of shards,
committee size, and cross-shard transaction ratio. The
performance model is calibrated against measurements from
RapidChain and OmniLedger. Ten independent runs yield a
stable Pareto front approximation of 100 solutions covering
30–1,500 Transactions per Seconds (TPS) at 96–390 ms
latency. NSGA-II achieves Inverted Generational Distance
(IGD) of 0.0048 ± 0.0003 and hypervolume convergence above
90% within 40 generations, with near-zero inter-run
variance (Coefficient of Variation, CV < 0.1%) across ten
independent runs, confirming robust convergence to a
well-distributed approximation of the Pareto front.

Choosing an appropriate search algorithm for
multi-objective optimization problems requires
understanding both the structural properties of the problem
and the mechanisms through which different algorithms
exploit them. This paper compares Non-dominated Sorting
Genetic Algorithm II (NSGA-II) against two baseline methods
– Random Search and Grid Search, on the problem of
configuring a blockchain sharding network with respect to
throughput and latency. The comparison is conducted over
ten independent runs per stochastic algorithm and evaluated
using Inverted Generational Distance (IGD) and a coverage
density analysis. NSGA-II achieves IGD = 0.0048 ± 0.0003,
an 8-fold improvement over Random Search (IGD = 0.040,
Wilcoxon p < 0.001) and a comparable improvement over Grid
Search (IGD = 0.038). The results show that the advantage
of NSGA-II stems from directed, population-based search
that concentrates effort near the Pareto front, rather than
from a larger evaluation budget alone.

In this study, we investigate the recharging of nodes in
wireless rechargeable sensor networks(WRSNs). We first
formulate a novel node recharging scheduling problem to
maximize the number of surviving nodes, and further propose
an adaptive recharging scheduling algorithm for WRSNs. In
this scheme, wireless charger(WC) calculates calculate the
charging priority of nodes based on their lifetime,
importance, and distance from itself, and select the node
with the highest priority to replenish energy. When its
energy is insufficient, it can switch to partial charging
mode. The simulation results show that the proposed
algorithm performs better than the other algorithms in
terms of the proportion of dead nodes, network lifetime,
and charging delay. The proposed algorithm can effectively
reduce the mortality rate of the nodes.

The typical approach to cross-language data exchange
between programming languages consists of serialization or
using immutable shared buffers, which do not allow in-place
modification, and impose copying overhead for each
mutation. We present Mutable Shared Memory (MSM), a memory
architecture that allows any C-compatible language (C++,
Python, Rust, and etc.) to directly read and write the same
data without serialization or duplication of buffers. MSM
uses a deterministic 8-byte aligned object layout, a
centralized native allocator exposed via foreign function
interface (FFI), and a (conditionally) lock-free
singlewriter/multiple-reader protocol on x86-64.
Experimental evaluation shows that MSM achieves 5.6×
speedup over Apache Arrow on in-place array mutation, 2.0×
faster end-to-end throughput on cross-language graph
pipelines, and 6.1× improvement over MsgPack on iterative
exchange workloads. The findings indicate that it is not
necessary for shared memory to be immutable in order to
achieve safe cross-language communication. Direct mutable
shared memory access under a clearly defined concurrency
protocol delivers substantial performance gains.

The use of BDD (Binary Decision Diagram) packages have
become essential in implementation of many CAD (Computer
Aided Design) algorithms, especially in the field of logic
synthesis. There are currently many different BDD packages
available, but all of them are built according to the basic
principles and recommendations for implementing these
packages. These packages usually have unique and computed
hash-based tables to provide a way to efficiently access
BDD nodes and the results of partial operations on them.
Selection of a hash function can have impact on the
performance of these tables. MurmurHash3 hash function is a
non-cryptographic hash function known for its excellent
performance and distribution characteristics. Therefore,
this paper presents an analysis of the impact of using the
MurmurHash3 hash function in BDD packages. Experiments on
the BDD benchmarks show slight improvement in performance
of BDD packages.

The implementation of the ISO/IEC 42001 standard in
organizational environments enables the effective
integration and use of modern information systems (IS),
contributing to improved personalized learning and
operational efficiency. By aligning standardized processes
with advanced digital platforms, organizations can
systematically collect and analyze data, monitor
performance, and adapt resources to meet specific user and
organizational needs. This study examines the impact of
ISO/IEC 42001-compliant IS on learning outcomes and
organizational processes. The results indicate that such
systems support more informed decision-making, enhance
overall efficiency, and improve alignment between
technological solutions and strategic objectives. The
application of ISO/IEC 42001 in modern IS represents an
important step toward optimizing organizational performance
and enabling sustainable digital transformation.

Poverljivo računarstvo (eng. Confidential Computing)
omogućava hardversku zaštitu podataka tokom aktivne obrade
u memoriji virtuelne mašine, čime se zaokružuje zaštita
podataka u tri stanja: u mirovanju, u tranzitu i u
upotrebi. Ovaj rad kvantifikuje uticaj na performanse AMD
SEV-SNP (eng. Secure Encrypted Virtualization – Secure
Nested Paging) mehanizma na Microsoft SQL Server, koristeći
HammerDB sa TPROC-C (transakcionim) i TPROC-H (analitičkim)
radnim profilima. Merenja su sprovedena na identično
konfigurisanim Google Cloud instancama sa AMD EPYC
procesorima, pri čemu se jedina razlika odnosila na
aktivaciju SEV-SNP mehanizma, dok su transparentno
šifrovanje podataka u mirovanju (eng. Transparent Data
Encryption, TDE) i šifrovanje transportnog sloja (eng.
Transport Layer Security, TLS) bili aktivni u oba
scenarija. Rezultati ukazuju da stepen performansnog
opterećenja u značajnoj meri zavisi od tipa radnog profila
i nivoa konkurentnosti. Za transakcioni profil opterećenje
je izraženo pri niskoj konkurentnosti, ali se značajno
smanjuje pri većem broju korisnika, dok je za analitički
profil razlika između platformi zanemarljiva.