Optimizing Edge-Based Query Processing for Real-Time Applications
Kalgi Gandhi and Minal Bhise
Distributed Database Group, DA-IICT, Gandhinagar, India
Keywords:
Column Imprint, Edge Intelligence, Edge Query Processing, Resource Utilization, Workload-Aware.
Abstract:
The rapid growth of edge devices in large-scale systems presents challenges due to limited processing power,
memory, and bandwidth. Efficient resource utilization and data management during query processing are crit-
ical, especially for costly join operations. The Column Imprint-Hash Join (CI-HJ) accelerates hash joins using
equi-height binning but lacks real-time efficiency and scans unnecessary cachelines. This paper introduces
Workload Aware Column Imprint-Hash Join (WACI-HJ), a novel approach that leverages workload prediction
to optimize hash joins for real-time edge query processing. WACI-HJ comprises of two phases: the WACI-
HJ Generation Phase predicts query workloads and pre-processes data into bins using blocking and hashing
techniques, reducing overhead before query arrival. The Query Processing and Resource Utilization Phase
efficiently utilizes CPU, RAM, and I/O resources for runtime processing. Evaluations using Benchmark and
Real-World datasets demonstrate significant improvements in the Percentage of Cachelines Read PCR, Query
Execution Time QET, and Resource Utilization. PCR and QET show 18% and 5% improvement respectively.
The proposed technique has been demonstrated to work well for scaled and skewed data. Although PCR is an
indirect measure of energy consumption, direct Energy-Efficiency Experiments reveal gains of 1%, 23%, and
18% in CPU, RAM, and I/O utilization respectively. WACI-HJ provides an optimal and sustainable solution
for edge database management.
1 INTRODUCTION
Edge Computing (EC) performs computing at the
edge, i.e. the source of the network. In contrast to
Cloud Computing, EC brings computation and data
storage closer to the data sources (Awaysheh et al.,
2023). The typical edge system is only a few hops
away from the data layer. It enables real-time appli-
cations, minimizes data transfer, and facilitates rapid
decision-making. Ideal for resource-constrained en-
vironments and the Internet of Things (IoT), EC ana-
lyzes data locally before transmitting structured data
to the cloud.
In the context of the Internet of Things (IoT), if all
the large amounts of unstructured or semi-structured
data generated by the connected devices are transmit-
ted to the cloud, latency and bandwidth will increase.
EC is an intermediate step, as seen in Figure 1. It pro-
cesses the unstructured or semi-structured data trans-
ferred by IoT and sends the processed structured data
to the cloud for storage. This data-handling process
is crucial for efficient query processing, as it ensures
that only the processed data is sent to the cloud, re-
ducing latency and bandwidth usage.
EC involves storing, processing, and analyzing
data directly on edge devices or nodes rather than
sending all data to the cloud. This approach reduces
latency by enabling quick query responses and opti-
mizing bandwidth by transmitting only relevant in-
formation. It handles and analyzes data closer to
its source rather than sending it to distant servers.
This minimizes latency, conserves bandwidth, en-
ables real-time analytics, and reduces energy con-
sumption (Bilal et al., 2018).
1.1 Motivation
EC performs computing close to the source of the net-
work, and it is just a few hops away from the data
layer. The number of edge-based applications and the
associated data are growing rapidly. EC applications
in various fields, including Smart Manufacturing, In-
telligent Transportation Systems, Health Monitoring
Systems, and Smart Homes. These devices are ac-
tively deployed and utilized in these domains to en-
hance efficiency, safety, and convenience.
As smart applications continue to develop, the
number and variety of connected devices have surged,
resulting in rapid data growth. From 9.76 billion IoT
Gandhi, K. and Bhise, M.
Optimizing Edge-Based Query Processing for Real-Time Applications.
DOI: 10.5220/0013289300003929
Paper published under CC license (CC BY-NC-ND 4.0)
In Proceedings of the 27th International Conference on Enterprise Information Systems (ICEIS 2025) - Volume 1, pages 259-266
ISBN: 978-989-758-749-8; ISSN: 2184-4992
Proceedings Copyright © 2025 by SCITEPRESS – Science and Technology Publications, Lda.
259

Figure 1: Query Processing at Edge.
devices in 2020, the count is projected to reach 34.6
billion by 2030 (iot, b), reflecting widespread adop-
tion across various industries. Correspondingly IoT
data, which was 44 zettabytes (ZB) in 2020, is ex-
pected to soar to 246 ZB by 2030 (iot, a). This ex-
plosive increase underscores the crucial role of ef-
fective data transmission and storage at the edge to
effectively manage and utilize this expanding digital
ecosystem.
1.2 Contributions
The main contributions of the work are as follows:
• To accelerate Query Processing in Edge
Workload-Aware Column Imprint-Hash Join
(WACI-HJ) is proposed.
• To enhance the real-time query processing perfor-
mance, Workload Prediction is incorporated with
WACI-HJ.
• WACI-HJ is implemented for benchmark Trans-
action Processing Performance Council (TPC)
TPC-H (tpc, b) and TPC-D (tpc, a) datasets along-
with, a real-world dataset Metropolitan Atlanta
Rapid Transit Authority (MARTA) (mar, ) aims
to validate the robustness and versatility of the al-
gorithm across diverse domains.
• WACI-HJ performance is compared with the
State-of-the-Art, Column Imprint-Hash Join (CI-
HJ) (Li and Xu, 2021) for the identified evaluation
parameters.
• WACI-HJ contributes to reducing latency, im-
proving energy-efficiency, and optimizing data
access, which can support sustainable database
management for diverse EC applications.
The rest of the paper is organized as follows: Sec-
tion 2 briefly summarizes the Literature Survey. The
Proposed Technique is elaborated in section 3, its Im-
plementation Details and Results are discussed in sec-
tions 4 and 5, and section 6 concludes the paper.
2 LITERATURE SURVEY
This section discusses the edge data management and
query processing domains. Additionally, merging Ar-
tificial Intelligence (AI) with EC predicts queries and
keeps track of the resources used in EC setups.
2.1 Data Management and Query
Processing at Edge
EC transforms data processing by decentralizing
computation and storage, reducing latency and en-
hancing responsiveness. Techniques such as Load
Balancing (Li et al., 2020) ensure balanced task distri-
bution across edge devices, optimizing performance
and preventing overload. Data Compression (Gandhi
et al., 2021) minimizes data size before transmis-
sion, maximizing bandwidth efficiency and acceler-
ating data transfer. Predicate Caching (Schmidt et al.,
2024) stores frequently accessed data locally, improv-
ing query response times, while Predictive Analytics
(Ma et al., 2018) enables proactive decision-making
based on historical data insights. Efficient resource
management strategies like Data Partitioning (Gandhi
and Bhise, 2019) and Prefetching (Chen et al., 2007)
enhance processing efficiency and reduce access la-
tency, while techniques such as Parallel Joins (Vi-
torovic et al., 2016) enhance scalability. Initiatives in
ICEIS 2025 - 27th International Conference on Enterprise Information Systems
260

Figure 2: WACI-HJ System Architecture.
energy-efficient EC, such as Green Cloud Network-
ing (GCN) (Sun and Ansari, 2017), optimize energy
consumption. Data encoding enhances memory usage
and reduce transmission volume (Symeonides et al.,
2019) while leveraging parallelism to split queries
into independent sub-queries helps optimize through-
put and latency (Xu et al., 2018). The CI-HJ strat-
egy (Li and Xu, 2021) further improves performance
in large-scale edge systems by pre-computing CIs for
join operations, accelerating hash processing and re-
ducing the data volume that must be scanned and
compared.
2.2 Edge Intelligence
Integrating AI with EC enhances processing and
decision-making at the network edge. Techniques like
Regression (Ma et al., 2018), Neural Networks (Ma
et al., 2018), and ARIMA (Calheiros et al., 2014) en-
able real-time data analysis and predictions, reducing
latency and bandwidth usage. However, current Edge
Intelligence (EI) models face challenges in predicting
complex queries, particularly in resource-constrained
edge environments.
3 WORKLOAD AWARE COLUMN
IMPRINTS-HASH JOIN
(WACI-HJ)
Cloud-based database systems often face scalability
and sustainability challenges due to bandwidth limi-
tations and high latency (Meruje Ferreira et al., 2024).
Edge computing mitigates these issues by bringing
storage and computation closer to the user, enabling
efficient real-time query processing. However, re-
source constraints in edge systems necessitate opti-
mized query execution, especially for join operations.
Hash joins are a key strategy for improving perfor-
mance and scalability in such environments (Barber
et al., 2014). CI-HJ (Li and Xu, 2021), a state-of-the-
art approach for large-scale edge systems, leverages
pre-computed CIs to accelerate hash joins by reduc-
ing the data volume involved, thus enhancing query
performance and minimizing resource utilization.
WACI-HJ enhances hash joins in large-scale edge
systems by leveraging workload-aware binning and
predictive optimization. Unlike traditional equi-
height binning, it uses workload patterns and value
weights, such as cachelines or value frequencies,
to construct histograms aligned with query require-
ments. The approach pre-computes CIs for each
table and generates optim ized imprint vectors for
probe columns, reducing cache misses during hash
joins. Additionally, WACI-HJ integrates a forecaster
to analyze historical data trends, predict upcom-
ing workloads, and allocate resources efficiently, en-
abling real-time query processing and improved per-
formance in edge systems.
Phase I includes Pre-processing, Prediction, and
Blocking and Hashing modules, which compute bins
ahead of query arrival. Module 1 loads the dataset,
generates queries using Zipf’s Law, and constructs
a Usage Matrix to identify attribute usage patterns.
Module 2 computes cluster usage frequency and ap-
plies the ARIMA technique to predict future usage.
Module 3 assigns bins, sets borders, and divides data
into blocks, storing them as WACIs.
In Phase II, Module 4 updates the workload and
WACI-HJ Output, loading necessary cachelines into
main memory for query processing. Module 5 moni-
Optimizing Edge-Based Query Processing for Real-Time Applications
261

tors CPU, RAM, and I/O utilization to ensure efficient
resource management during query execution.
3.1 WACI-HJ Data Structures and
Algorithm
As depicted in Algorithm 1, WACI-HJ for Module
1 involves the Usage Matrix Basket (UMB), which
lists queries and their associated attributes. For Mod-
ule 2, the Query Workload List (QWL) holds the to-
tal query frequency count. The Cluster-Range Table
(CRT) comprises of ranges. The Cluster-Frequency
Table (CFT) contains the query range and weekly ar-
rival frequency. The workload and WACI-HJ Output
are updated upon the query arrival. WACI-HJs, stored
in secondary memory, bring the necessary cachelines
to the main memory for query output. Further, CPU,
RAM, and I/O utilization are monitored during query
execution.
The time and space complexities of WACI-HJ
are O(2
x
(n ∗ t) ∗ 64) and O(n) respectively, where
x is the log of the number of bins, n represents
the total number of tuples in the probe column, i.e.
no of cachelines*64, and t is the size of the data type.
4 IMPLEMENTATION DETAILS
This section provides a comprehensive overview of
the hardware and software setup, dataset and query-
set specifications, and evaluation parameters used for
the implementation. It outlines the technical environ-
ment and resources utilized to demonstrate the effec-
tiveness of the WACI-HJ algorithm.
4.1 Hardware and Software Setup
The machine hardware configuration included a quad-
core Intel i3-2100 CPU clocked at 3.10 GHz. The sys-
tem has 32 GB of RAM, temporarily storing data that
the CPU needs to access quickly. It has 500 GB of
hard disk space for storing data and files. Addition-
ally, the system includes hardware with L1, L2, and
L3 caches sized at 64 Kilobyte (KB), 512 KB, and 3
MB respectively, enhancing processing speed by stor-
ing frequently accessed data close to the CPU.
The software required to implement WACI-HJ is
the VS Code IDE platform (vsc, ) for C language de-
velopment, and MonetDB (mon, ), an open-source
column store database for implementing CIs. Google
Colab (goo, ) for Python, supporting machine learn-
ing computing are used for prediction tasks.
Algorithm 1: Workload Aware Column Imprint-Hash Join
(WACI-HJ) Generation Algorithm.
Input: Tuples T
i
, Queries Q
j
Output: WACI-HJ Output
//Module 1: Pre-processing
Initialize UMB
for query Q
j
do
A
j
= Extract Attributes From Query(Q
j
)
for attribute A
i
in A
j
do
if Ai exists in U MB then
Add Query Index(A
i
, Index o f (Q
j
), UMB)
else
Set Query Index(A
i
, [Index o f (Q
j
)], UMB)
end if
end for
end for
//Module 2: Prediction
Create CRT C
i
and CR
i
State CFT = number of
weeks * CRT
Cluster week f req =
∑
no o f queries
j=1
QF
j
CF
i
= array(cluster week f req)
model = auto arima(CF
i
, suppress warnings = True)
PredictedQWL = model × predict(n periods =
1, return conf int = True)
//Module 3: Blocking and Hashing
imp vec = no o f cachelines ×
sizeof(ImprintVector)
for i = 0 to no o f cachelines − 1 do
imp vec[i] = no o f bins × sizeof(datatype)
end for
for j = 0 to no o f cachelines − 1 do
for val = 1 to cachelineSize do
bin = getBin(val)
imp vec[ j][bin] = 1
end for
end for
for val = 0 to build col size − 1 do
HashArray[hashFunction(col[val])] = val[i]
end for
for i = minimum bin required to
maximum bin required do
while list temp[i] do
q = list temp[i] → data
for w = q × cacheline size to
min(q × (cachelines
size + 1), size probe col)
do
if probe col[w] ≤ bin borders[i] or
probe col[w] ≥ bin borders[i + 1] then
continue
end if
if HashArray[hashFunction
(probe col[w])] ̸= −1 then
store BAT Values corresponding to
both columns in WACI-HJ Output
end if
end for
list temp[i] = list temp[i] → next
end while
end for
ICEIS 2025 - 27th International Conference on Enterprise Information Systems
262

4.2 Dataset and Queryset
The implementation of WACI-HJ is demonstrated us-
ing a benchmark dataset, a standardized dataset that is
generally well-documented, public, and widely used
in research. Alongwith, real-world datasets contain
authentic data from sources like sensors.
TPC-H (tpc, b)- TPC-H a benchmark dataset in the
supply chain domain, simulates a data warehousing
environment with eight tables. The orderkey attribute
is used for experimentation, joining l orderkey with
o orderkey. The 1 GB dataset is scaled upto 10x for
testing. Out of 22 OLAP queries, 9 involve joins
with orderkey, and 11 custom range queries with joins
were added, totalling 20 queries. TPC-D, a skewed
version of TPC-H, is also referenced.
MARTA (mar, )- MARTA, a real-world dataset from
the transportation domain, includes 13 tables. The
route id attribute is used for experimentation. The
original 93.1 MB dataset is scaled to 10x, and 10 cus-
tom range queries with joins were created for evalua-
tion.
4.3 Set of Experiments
The experiments are organized into distinct sets de-
signed to thoroughly evaluate the algorithm perfor-
mance, robustness, and adaptability across various
scenarios.
Set 0: System Configuration- Initially, Set 0 focuses
on configuring the system by determining the optimal
bit size and number of clusters to enhance the algo-
rithm effectiveness. Once the system is configured,
the subsequent experiments are divided into four sets.
Set I: Basic Experiments- Set I involves basic tests
conducted on the standard dataset size, with TPC-H
and APM datasets initially set at 1 GB.
Set II: Data Scaling Experiments- In Set II, the ex-
periments assess how the algorithm performs with
varying data sizes, scaling the datasets from x upto
10x.
Set III: Varying Skewness Levels Experiments -
Real-world data is often not evenly distributed.
Hence, Set III examines the algorithms robustness
by evaluating its performance on datasets with skew-
ness levels ranging from 50% i.e. uniformly skewed
to 99%, addressing real-world data distribution chal-
lenges.
Set IV: Energy-Efficiency Experiments- Set IV di-
rectly measures energy-efficiency by monitoring CPU
and RAM utilization and I/O efficiency (Patel and
Bhise, 2023). It aims to evaluate the systems energy
consumption during query execution across different
data types and workloads.
4.4 Evaluation Parameters
The evaluation of WACI-HJ contains Input, Output,
and Resource Utilization parameters.
Input Parameters- Input Parameters include Bit Size
(ranging from 8 to 128 bits), Data Size (scaled from
x to 10x of a base 1 GB for TPC-H and 0.97 GB for
APM), and Skewness Levels (ranging from 50% to
99%).
Output Parameters- Output Parameters measure sys-
tem performance, PCR which indicates the proportion
of blocks accessed by a query compared to the total
number of blocks, QET measures the time taken for
query execution in seconds, and AET denotes the to-
tal duration of algorithm execution, also in seconds.
Resource Utilization Parameters- Resource Utiliza-
tion Parameters assess the use of computational re-
sources, including CPU Utilization, RAM Utilization,
and I/O Efficiency, to measure the system efficiency
and performance under various conditions.
5 RESULTS AND DISCUSSIONS
This section evaluates indexing techniques, including
the state-of-the-art CI-HJ (Li and Xu, 2021) and the
proposed WACI-HJ with (Shah et al., 2024) and with-
out workload prediction, based on input, output, and
resource utilization parameters. The results are de-
rived using benchmark datasets TPC-H (tpc, b), TPC-
D (tpc, a), a skewed TPC-H and the MARTA dataset
(mar, ), both representing high growth edge data ap-
plications. Results are averaged across all queries in
the queryset.
These evaluations involved four experiments, op-
timizing parameters like bit size and cluster count for
efficient PCR and QET. Bit size, being system depen-
dent, remains constant across datasets, while cluster
count varies with data size larger datasets generate
more clusters. Further sets use a 32-bit size, with 4
clusters for TPC-H (tpc, b) and TPC-D (tpc, a), and 2
clusters for MARTA (mar, ).
5.1 Set I: Basic Experiments
In the Basic Experiments, PCR is measured based
on query selectivity and frequency. Results shown in
Figure 3, reveal a significant improvement in queries
with lower selectivity due to more non-distinct values
in the orderkey attribute, especially when compared to
CI-HJ. WACI-HJ reduces scanned cachelines by 14%
compared to CI-HJ, with an additional 4% improve-
ment when workload prediction is included. This re-
duction is attributed to workload-aware optimization,
Optimizing Edge-Based Query Processing for Real-Time Applications
263

Figure 3: TPC-H Set I: PCR.
Figure 4: TPC-H Set I: QET.
leading to lower RAM and I/O usage and decreased
system latency.
Figure 4 shows the QET results for the Basic Ex-
periments, with queries categorized by the number
of joins. WACI-HJ delivers better performance for
queries with more joins, demonstrating its efficiency
in hash join acceleration. Compared to CI-HJ, WACI-
HJ achieves a 3% improvement in QET, with an addi-
tional 2% enhancement when workload prediction is
integrated. This results in faster query execution and
reduced CPU usage.
5.2 Set II: Data Scaling Experiments
Data Scaling Experiments execute the proposed tech-
nique with data sizes ranging from x, 2x, 5x, 7x, and
10x.
Figure 5 shows the results of the PCR for Data
Scaling Experiments. The overall gain was 7% to
18% when scaled from x to 10x. The trail indicates
that the PCR for WACI-HJ is constant wrt. data size.
Figure 6 shows the results of the scaled QET, with
WACI-HJ achieving an overall improvement of 2% to
5% in Data Scaling Experiments, demonstrating con-
sistent QET enhancement for large-scale query sce-
narios.
WACI-HJ demonstrates a 2% higher AET com-
Figure 5: TPC-H Set II: PCR.
Figure 6: TPC-H Set II: QET.
pared to CI-HJ, but achieves a significant 5% im-
provement in QET. When data scales from x to 10x,
AET improves by up to 19% due to optimized work-
load bins, making WACI-HJ more suitable for larger
datasets.
5.3 Set III: Varying Skewness Levels
Experiments
Varying Skewness Levels Experiments are performed
using the TPC-D (tpc, a) dataset. WACI-HJ is evalu-
ated at different levels of data skewness.
Figure 7: TPC-D Set III: PCR.
WACI-HJ PCR is constant for various skewness
levels, as shown in Figure 7. An 5% to 50% improve-
ICEIS 2025 - 27th International Conference on Enterprise Information Systems
264

Figure 8: TPC-D Set III: QET.
ment is recorded for WACI-HJ over CI-HJ as the data
gets skewed. A peak is observed for 80% skewed
data, and we are further investigating it.
As shown in Figure 8, WACI-HJ becomes 2% to
10% faster with a further increase in skewness level.
It shows that WACI-HJ is robust for different skew-
ness levels.
5.4 Set IV: Energy-Efficiency
Experiments
Although PCR indirectly reflects energy consump-
tion, the energy-efficiency of WACI-HJ is directly
measured using Resource Utilization tools like htop
and iotop to track CPU, RAM, and I/O consump-
tion. This section compares the resource utilization
of WACI-HJ with the state-of-the-art technique, using
the TPC-H dataset for evaluation.
Figure 9: TPC-H Set IV: IO Efficiency.
The CPU utilization of WACI-HJ shows 2% gain
is observed in CPU usage compared to CI-HJ, reduc-
ing the latency. A 27% gain is observed in RAM us-
age compared to CI-HJ due to the optimal number
of cachelines read. Due to the effective number of
cachelines read, I/O operations were reduced by 18%,
making the system more faster and energy-efficient,
as seen in Figure 9.
5.5 Comparison with State-of-the-Art
CI-HJ (Li and Xu, 2021) improves query performance
by using CIs to avoid scanning irrelevant data blocks,
but it may incur significant processing overhead with
large datasets due to unnecessary data scans and lacks
real-time workload adaptation.
In contrast, WACI-HJ adopts a workload-aware
approach, creating bins based on workload informa-
tion to minimize cache misses and predicting future
workloads for effective real-time query processing.
5.5.1 Quantitative Comparison
In experiments on the TPC-H (tpc, b) dataset, WACI-
HJ outperforms CI-HJ, with an 18% improvement in
PCR and a 5 second reduction in QET, demonstrat-
ing faster query execution. Despite a 2% increase in
AET due to workload prediction overhead, WACI-HJ
proves effective in optimizing query performance.
WACI-HJ also shows significant energy-
efficiency, achieving a 23% improvement in RAM
utilization and an 18% enhancement in I/O opera-
tions, reducing memory and data access overhead.
CPU utilization sees a minor 1% gain, highlighting
WACI-HJ effectiveness in resource optimization for
edge environments.
5.5.2 Qualitative Comparison
CI-HJ (Li and Xu, 2021) handles scaled and skewed
data but scans unnecessary cachelines, lacks real-time
processing, predictive features, and energy-efficiency,
performing better on synthetic datasets than bench-
marks like TPC-H.
WACI-HJ being workload-aware, handles scaled
and skewed data avoids unnecessary scans, supports
real-time processing, includes predictive capabilities,
and is energy-efficient. It excels with skewed data and
demonstrates robust performance on benchmarks like
TPC-H and real-world datasets like MARTA, making
it more versatile and effective.
6 CONCLUSIONS
WACI-HJ optimizes real-time edge query processing
by accelerating hash joins using the workload-aware
approach, where the upcoming workloads are pre-
dicted well in advance. Experimental results show
significant reductions in PCR and QET, with 18%
and 5% improvements for the benchmark dataset,
respectively. The algorithm scales well, maintain-
ing stable PCR across data sizes upto 10x and vary-
ing data skewness. In resource-constrained environ-
Optimizing Edge-Based Query Processing for Real-Time Applications
265

ments, WACI-HJ achieves gains of 2% in CPU, 27%
in RAM, and 18% in I/O utilization. Testing with a
real-world dataset shows a 54% improvement in PCR
and 10% in QET. Energy-efficiency experiments con-
firm gains of 1% in CPU, 38% in RAM, and 49% in
I/O. By optimizing data access and incorporating pre-
dictive capabilities, WACI-HJ reduces query latency
and conserves energy, making it ideal for resource-
constrained edge applications.
REFERENCES
Google Colab. Available: https://colab.research.google.
com/ Accessed: June, 2024.
IoT Data Growth. Available: https://www.gartner.com/en/
documents/3996804 Accessed: June, 2024.
IoT Devices Growth. Available: https://www.statista.com/
statistics/1183457/iot-connected-devices-worldwide/
#:
∼
:text=The%20number%20of%20Internet%20of,
billion%20IoT%20devices%20in%202030. Ac-
cessed: June, 2024.
MARTA Dataset. Available: https://data.world/
brentbrewington/marta-hackathon Accessed: June,
2024.
MonetDB. Available: https://www.monetdb.org/ Accessed:
June, 2024.
TPC-D Dataset. Available: https://github.com/gunaprsd/
SkewedDataGenerator Accessed: June, 2024.
TPC-H Dataset. Available: http://www.tpc.org/tpch/ Ac-
cessed: June, 2024.
VScode. Available: https://code.visualstudio.com/ Ac-
cessed: June, 2024.
Awaysheh, F. M., Tommasini, R., and Awad, A. (2023). Big
data analytics from the rich cloud to the frugal edge.
In 2023 IEEE International Conference on Edge Com-
puting and Communications (EDGE), pages 319–329.
IEEE.
Barber, R., Lohman, G., Pandis, I., Raman, V., Sidle, R.,
Attaluri, G., Chainani, N., Lightstone, S., and Sharpe,
D. (2014). Memory-efficient hash joins. Proceedings
of the VLDB Endowment, 8(4):353–364.
Bilal, K., Khalid, O., Erbad, A., and Khan, S. U. (2018).
Potentials, trends, and prospects in edge technologies:
Fog, cloudlet, mobile edge, and micro data centers.
Computer Networks, 130:94–120.
Calheiros, R. N., Masoumi, E., Ranjan, R., and Buyya, R.
(2014). Workload prediction using arima model and
its impact on cloud applications’ qos. IEEE transac-
tions on cloud computing, 3(4):449–458.
Chen, S., Ailamaki, A., Gibbons, P. B., and Mowry, T. C.
(2007). Improving hash join performance through
prefetching. ACM Transactions on Database Systems
(TODS), 32(3):17–es.
Gandhi, K. and Bhise, M. (2019). Affinity-based fragmen-
tation for sensor data. In 2019 IEEE 16th India Coun-
cil International Conference (INDICON), pages 1–4.
IEEE.
Gandhi, K., Pandat, A., and Bhise, M. (2021). Ex-
periments on static data summarization techniques.
In 2021 IEEE International Women in Engineering
(WIE) Conference on Electrical and Computer Engi-
neering (WIECON-ECE), pages 17–20. IEEE.
Li, G., Yao, Y., Wu, J., Liu, X., Sheng, X., and Lin, Q.
(2020). A new load balancing strategy by task alloca-
tion in edge computing based on intermediary nodes.
EURASIP Journal on Wireless Communications and
Networking, 2020:1–10.
Li, Y. and Xu, W. (2021). Utilizing the column imprints
to accelerate no-partitioning hash joins in large-scale
edge systems. Transactions on Emerging Telecommu-
nications Technologies, 32(6):1–17.
Ma, L., Van Aken, D., Hefny, A., Mezerhane, G., Pavlo,
A., and Gordon, G. J. (2018). Query-based workload
forecasting for self-driving database management sys-
tems. In Proceedings of the 2018 International Con-
ference on Management of Data, pages 631–645.
Meruje Ferreira, L. M., Coelho, F., and Pereira, J. (2024).
Databases in edge and fog environments: A survey.
ACM Computing Surveys, 56(11):1–40.
Patel, M. and Bhise, M. (2023). Muar: Maximizing uti-
lization of available resources for query processing.
In 2023 IEEE/ACM 23rd International Symposium
on Cluster, Cloud and Internet Computing Workshops
(CCGridW), pages 176–183. IEEE.
Schmidt, T., Kipf, A., Horn, D., Saxena, G., and Kraska,
T. (2024). Predicate caching: Query-driven secondary
indexing for cloud data warehouses. In Proceedings
of the 2024 ACM SIGMOD International Conference
on Management of Data, pages 347–359.
Shah, K., Gandhi, A., Gandhi, K., and Bhise, M. (2024).
Workload prediction for edge computing. In Pro-
ceedings of the 25th International Conference on Dis-
tributed Computing and Networking, pages 286–291.
Sun, X. and Ansari, N. (2017). Green cloudlet network:
A distributed green mobile cloud network. IEEE net-
work, 31(1):64–70.
Symeonides, M., Trihinas, D., Georgiou, Z., Pallis, G., and
Dikaiakos, M. (2019). Query-driven descriptive ana-
lytics for iot and edge computing. In 2019 IEEE In-
ternational Conference on Cloud Engineering (IC2E),
pages 1–11. IEEE.
Vitorovic, A., Elseidy, M., and Koch, C. (2016). Load bal-
ancing and skew resilience for parallel joins. In 2016
IEEE 32nd International Conference on Data Engi-
neering (ICDE), pages 313–324. Ieee.
Xu, R., Palanisamy, B., and Joshi, J. (2018). Queryguard:
Privacy-preserving latency-aware query optimization
for edge computing. In 2018 17th IEEE International
Conference On Trust, Security And Privacy In Com-
puting And Communications/12th IEEE International
Conference On Big Data Science And Engineering
(TrustCom/BigDataSE), pages 1097–1106. IEEE.
ICEIS 2025 - 27th International Conference on Enterprise Information Systems
266