A Platform Selection Framework for Blockchain-Based Software
Systems Based on the Blockchain Trilemma
Jan Werth
1
, Nabil El Ioini
2
, Mohammad Hajian Berenjestanaki
1
, Hamid R. Barzegar
1
and Claus Pahl
1
1
Free University of Bozen-Bolzano, Bolzano, Italy
2
University of Nottingham, Malaysia
Keywords:
Blockchain, Distributed System, Distributed Ledger, Blockchain Trilemma, Scalability, Security,
Decentralization.
Abstract:
Blockchains are used in many software systems to deal with trusted storage. The selection of the appropriate
software architecture stack in distributed systems is generally driven by scalability, security, and decentral-
ization as central qualities. In the blockchain domain, these are known as the blockchain trilemma, as they
oppose each other. We select the most popular blockchain platforms based on these trilemma properties and
other indicators to provide a platform review. Specific metrics will be derived from the overall goals and ap-
plied to the platform options. This serves as a basis to create a Selection framework to facilitate the choice of
the best possible platform for a given system architecture. The selection framework is evaluated through a use
case.
1 INTRODUCTION
For many software systems, specifically in distributed
architectures, blockchains can provide a trusted stor-
age layer. A central trust authority is no longer
required as the network participants manage the
blockchain state. Here, each node stores a (local)
copy of the blockchain, which is considered im-
mutable. In addition, other benefits of blockchain
technology are better security and enhanced privacy.
Nevertheless, common qualities such as scalability,
security, and decentralization often need to be con-
sidered. In the blockchain domain, these are known
as the blockchain trilemma, as they oppose each
other. Since blockchain platforms offer different per-
formances regarding these properties, decision sup-
port for the right platform in an overall system archi-
tecture is needed.
We propose a selection framework that allows us
to find the best possible blockchain for a given system
with respective scalability, security, and decentraliza-
tion requirements for instance, as a decentralized
exchange, for e-voting, or in the healthcare sector.
The first goal is to provide a substantial analysis of
a selected set of blockchain platforms and their con-
sensus protocols based on the blockchain trilemma
and subsequently draw out a selection framework for
choosing the right platform. For this, we map the
three goals scalability, security, and decentralization
onto detailed, measurable metrics to allow a detailed
assessment of the platforms. This in turn is used to
define the selection framework for the selection.
2 BLOCKCHAIN TRILEMMA
The blockchain is an immutable, decentralized ledger
of transactions, where multiple transactions are
grouped into a block which is then appended to the
chain of blocks. There is no need for a central au-
thority. Instead, blockchains reach consensus thanks
to their consensus protocol.
Consensus protocols form the pillar of
blockchains by defining a set of rules which
dictate how a distributed system and its parts operate
and interact. The consensus protocol determines
how the network can reach consensus on the future
state of the blockchain, i.e., most of a blockchain’s
participants have to agree on the same state to reach
consensus. Therefore it is essential to understand
that the consensus protocol and the blockchain
platform itself can be seen as two separate objects
where the consensus protocol defines the rules and
functionality of the blockchain and the blockchain
platform realizes the consensus protocol. Different
blockchains use different consensus protocols.
The Blockchain Trilemma states that a blockchain
can only fulfil two out of the three following contra-
dicting properties: Scalability, Decentralization and
Security. It is an extension of the scalability problem -
362
Werth, J., El Ioini, N., Berenjestanaki, M., Barzegar, H. and Pahl, C.
A Platform Selection Framework for Blockchain-Based Software Systems Based on the Blockchain Trilemma.
DOI: 10.5220/0011837300003464
In Proceedings of the 18th International Conference on Evaluation of Novel Approaches to Software Engineering (ENASE 2023), pages 362-371
ISBN: 978-989-758-647-7; ISSN: 2184-4895
Copyright
c
2023 by SCITEPRESS Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)
a problem arising from an increasing number of trans-
actions but a limited throughput in major blockchain
platforms.
2.1 Scalability
Scalability is one of the most important aspects of
many distributed systems such as blockchains. Here,
it refers to the speed at which participants of a peer-to-
peer network can reach consensus on the state of the
blockchain (Hafid et al., 2020). Mathematically it can
be represented as the maximum block size divided by
the block interval (Croman et al., 2016). Following
this, solving the scalability issue can be done by ei-
ther increasing the block size or decreasing the block
interval. However, external factors such as computing
power, bandwidth, and storage space (Buterin, 2021)
call for an internal solution to the problem. This is
where the blockchain trilemma arises, existing solu-
tions such as the Proof-of-Stake (PoS) consensus pro-
tocol trade in decentralization in favor of scalability.
Using only a limited number of validators, a partic-
ular type of node allowed to create and confirm new
blocks, PoS protocols can decrease required network
communication and increase its scalability. Proof-of-
work (PoW) protocols do not differentiate between
different types of nodes, and everyone has the same
rights.
Low blockchain transaction rates lead to a prob-
lem where transactions can no longer be processed
immediately. Therefore, in the context of blockchain,
scalability refers to the ability to support high trans-
actional throughput while maintaining performance.
Croman et al. (Croman et al., 2016) identified key
metrics to measure scalability of blockchain plat-
forms: maximum throughput, latency, bootstrap
time, and cost per confirmed transaction, where the
first two measurements are the most important for a
user who intends to use a blockchain without being a
miner or a validator.
Maximum throughput refers to the above-
explained concept of transactions per second. La-
tency is the time it takes for a blockchain to create
a new block, append it to the blockchain, and regard
it as confirmed. It can be divided into two parts which
are the block time and the time to finality. The for-
mer refers to the time needed to create a block and
add it to the blockchain. In contrast, time to final-
ity can be once again subdivided into deterministic
and probabilistic. Deterministic means that a block is
considered final once it is appended to the blockchain.
In other words, the block is no longer changeable
once it has been published. Probabilistic refers to the
blockchains in which a block is still subject to change
once it has been added to the blockchain, i.e., due to
the network not having reached consensus on the fu-
ture state of the blockchain. Bootstrap time refers to
the time it takes to download a blockchain and con-
firm all the blocks and transactions. Costs per transac-
tion are external factors such as setup cost, hardware
cost, storage cost, and power usage.
2.2 Decentralization
Decentralisation is the central ethos and given nature
of the blockchain technology, but also a massive bot-
tleneck regarding scalability and security. It describes
the transfer of control and decision-making rights
from a central authority to a distributed network. A
characteristic of decentralisation in blockchains is the
distrust between its participants, which is desired and
required for it to work correctly.
Measuring a network’s decentralisation depends
on the type of blockchain. Two types of blockchains
exist or rather two types on how decentralisation must
be measured. One type uses the Proof-of-Work con-
sensus protocol, while the other type uses Proof-of-
Stake or a similar consensus protocol where the rights
to create a new block are given to a node based on
staked capital. The decentralisation (and security)
of a Proof-of-Work blockchain depends on the net-
work’s hash rate and how distributed it is. A net-
work’s hash rate is the cumulative hash rate of all
the (mining) nodes participating in the block creation
competition. Therefore, the higher the network’s hash
rate, the harder it is to disrupt it.Decentralisation of
a Proof-of-Stake or similar blockchain can be mea-
sured in the number of validators, the distribution of
staked capital among the validators, and the percent-
age of token supply that has been staked. Another
metric to measure decentralisation is the Initial To-
ken Allocation. It can create unfair advantages for
a group that receives many tokens and determine the
next block and chain governance. For both Proof-of-
Work and Proof-of-Stake (or similar blockchains), it
is important to measure how many nodes or pools (a
pool is a group of miners or validators which join to-
gether to increase their chance of creating the next
block) control the majority of the network. his met-
ric is also called Superminority or Nakamoto Coef-
ficient. The Nakamoto Coefficient is defined as the
minimum number of nodes required to get 51% of the
total capacity (either in computing power or staked
capital) (Srinivasan, 2017). However, for networks
with a lower Byzantine Fault Tolerance, it is only re-
quired to control one-third of the network’s comput-
ing power or staked capital.
A Platform Selection Framework for Blockchain-Based Software Systems Based on the Blockchain Trilemma
363
2.3 Security
Security is the third aspect of the blockchain
trilemma. A framework used in cybersecurity is the
CIA triad which stands for confidentiality, integrity,
and availability.
Some other security features come from the CAP
theorem, which stands for consistency, availability,
and partition tolerance and states that none of these
three characteristics can be achieved simultaneously.
This theorem is especially popular among blockchain
developers and often mentioned in the whitepapers
of different platforms, documentation, or online dis-
cussions. Zhang et al. (Zhang et al., 2019) applied
the aforementioned CAP properties to a distributed
ledger: Consistency: All nodes keep an identical
ledger with most recent updates. Availability: Any
transaction generated at any time will be accepted in
the ledger. Partition Tolerance: Even if part of the
network fails, it can still operate normally.
It seems that the blockchain implementation has
violated the CAP theorem by achieving not only con-
sistency but also availability and partition tolerance.
However, this is not the case as a block’s latency
plays a role in consistency. This was also identified by
Zhang et al. (Zhang et al., 2019) who state that consis-
tency is not achieved simultaneously with availability
and partition tolerance but only after a period of time.
Given this weak characteristic and the existence of a
higher-level term in the CIA triad, we can categorize
consistency as part of integrity. We can also group
availability and partition tolerance, where the latter
is a sub-category of the former: Confidentiality con-
cerns the privacy of a user. Integrity includes consis-
tency, authenticity, accuracy and tamper-resistance of
data. Consistency means that all nodes store the same
state of the blockchain. Availability means that the
blockchain is always available to be read and accepts
transactions. Partition Tolerance means that it works
even if part of the network fails.
According to Zhang et al. (Zhang et al., 2019)
a blockchain platform does not only have to fulfill
the above properties of the CIA triad, it also must be
resistant against numerous different types of attacks,
such as resistancd to DDoS Attacks, to majority at-
tacks (51% attack and single shard attack), to double
spending and to transaction flooding. These are some
examples of attacks a blockchain platform, yet these
properties represent the most crucial security proper-
ties a blockchain must satisfy (Zhang et al., 2019).
Another measure of security is the Nakamoto Coeffi-
cient.
Byzantine Fault Tolerance. (BFT) refers to the
ability of a distributed system to keep working cor-
rectly even when a fraction of its nodes fail or act ma-
liciously. Blockchains reach BFT through their con-
sensus protocols which dictate the rules. If a node is
no longer following the consensus protocols’ rules, it
is a malicious node that does not act in the network’s
interest. Furthermore, most of the blockchains’ at-
tacks occur inside the network, implying the pres-
ence of byzantine nodes. A higher byzantine fault
tolerance also means that a network is more secure.
Other outside attacks (e.g., DDoS attacks) attack sin-
gle nodes in the network and try to shut them down.
Such an attack can not only be prevented by having a
high byzantine fault tolerance so that fewer nodes are
required to keep the network operating, but also high
decentralization fends off DDoS attacks. Yet again,
another example of how two aspects of the blockchain
trilemma are connected.
3 SELECTION CRITERIA AND
METRICS
We provide an analysis of different blockchain plat-
forms, which will then be used to create a selec-
tion framework to facilitate the selection of which
platform to use among the selection of platforms.
The reason is that every platform takes a different
approach to the blockchain trilemma with a differ-
ent objective. This poses a challenge for prospec-
tive blockchain developers and users as a particular
blockchain platform may not fulfil their requirements,
i.e. the application requires high, intermittent trans-
actional throughput, which only a few platforms sup-
port. Given the numerous existing blockchain plat-
forms, an analysis of all platforms would go beyond
the scope here and is not feasible. Therefore, only a
selection of blockchain platforms will be included in
the selection framework.
3.1 Platform Selection
The selection of which blockchain platforms to
compare was made in April 2022 based on
their blockchain trilemma properties, their type of
blockchain, their initial token allocation, their To-
tal Value Locked (TVL), and the number of min-
ers/validators participating in the consensus.
The most important selection criteria to ensure di-
versity among the selected blockchain platforms were
their blockchain trilemma aspects. In other words, it
was essential to include blockchains that focus on the
trilemma’s different aspects. This was done by look-
ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering
364
ing at different metrics which define the platform’s
scalability, decentralization, and security. The se-
lected metrics for scalability are the platform’s max-
imum transactions per second, their block time and
their time to finality. For decentralization the number
of nodes, their type and whether the number of nodes
is fixed is essential. The number of nodes, their type,
and the distribution of computing power and staked
capital is also of interest for security, as well as the
Byzantine Fault Tolerance of the network.
Another criterion which had to be considered is
the type of blockchain. To ensure that the selected
blockchains are programmable on only public per-
missionless blockchain platforms (Pahl et al., 2018),
which also support Smart Contracts, were included.
The selection resulted in 9 blockchain platforms
1
with one Layer 0 solution (Cosmos), one Layer 2
solution (Polygon), and one platform which was in-
cluded for its approach to the blockchain trilemma,
Harmony. Harmony is already applying a method
called sharding to its blockchain, which Ethereum,
the platform with the most protocols and highest
TVL, will apply in 2023. The selection is presented
in Table 1.
We summarize the consensus protocols used by
the selection of blockchain platforms and additional
consensus protocols which may be necessary to un-
derstand more complex ones in Table 2.
3.2 Analysis of Platforms
The metrics chosen to analyze the blockchains are
generally available for all public blockchains. Table 3
includes transactions per second. Some platforms are
not yet fully implemented but will reach higher tps in
the future. These metrics are identified in (Croman
et al., 2016) to compare blockchain scalability.
current TPS: measure the average throughput of a
blockchain platform.
max TPS: states the maximum transactions per sec-
ond a blockchain can process.
Block Time: measures the time in seconds it takes
for a blockchain to create a new block.
Number of Nodes: measures the decentralization of
a blockchain. The number only includes nodes
responsible for block creation.
Time to Finality: measures of block latency. In
probabilistic networks, a block is not considered
1
Please note that we do not list all individual sources
of information separately due to their large number. All in-
formation has been gathered from the documentation made
available by the providers.
final even after it was created due to the risk of
forks and other. However, some platforms in the
selection offer deterministic finality, which means
that a block is final the moment it was produced.
Table 4 shows a comparison of each platform’s
decentralisation and includes the Nakamoto Coeffi-
cient. The metrics to measure decentralization (Con-
way, 2022) (Srinivasan, 2017) are:
Number of Nodes: participating in the block cre-
ation process.
Type of Nodes: Different consensus protocols use
different types of nodes for block creation. How-
ever, in this selection, most platforms use valida-
tors or some sort of validators
2
.
Fixed Number of Nodes: states whether the number
of nodes participating in the block creation is fixed
or can scale.
Hashrate / % of Supply Staked: reports the net-
work’s total hash rate (for Ethereum) and percent-
age of how much of the token supply of a net-
work’s crypto-currency is staked.
Nakamoto Coefficient: measures how many entities
are in control of 51% or 34% (depending on
BFT) of the network’s power (either in comput-
ing power or staked capital).
Table 6 depicts the security aspect of the blockchain
platforms. The metrics are:
Byzantine Fault Tolerance measures the threshold
of failed or adverse nodes a network can with-
stand.
Availability blockchains is imported as transactions
always occur.
Anonymity shows if complete anonymity or
pseudonymity where a transaction can be tracked
and linked to an address is offered.
Note that the Nakamoto Coefficient is also a metric
used to measure the security of a network as large en-
tities.
3.3 Design of the Framework
To create a selection framework, we have to establish
the metrics and rules which the selection framework
will follow. The metrics are used in scientific articles
and in the blockchain community to compare and an-
alyze the performance of different platforms (Hafid
2
For example, in BSC, validators have to be approved
by Binance. Trons’ Super Representatives are also valida-
tors. Only Ethereum does not use validators but uses com-
peting nodes as participate in the Proof-of-Work consensus.
A Platform Selection Framework for Blockchain-Based Software Systems Based on the Blockchain Trilemma
365
Table 1: Selected Blockchain Platforms.
Consensus Structure Architecture
Smart Contracts
Ethereum PoW Chain Single Chain YES
Cosmos Tendermint PoS Chain Cosmos Parachain YES
BSC PoSA Chain Cosmos Parachain YES
Avalanche SoA DAG
X-Chain, P-Chain, C-Chain
YES
Solana PoH Chain Single Chain YES
Fantom LCA DAG Main Chain (Atropos) YES
Tron DPoS Chain Single Chain YES
Polygon PoS Chain Sidechains YES
Harmony FBFT Chain Sharding YES
Table 2: Comparison of Consensus Protocols.
Throughput
Transaction
Finality
Decentralization
BFT
Energy
Consumption
Proof of Work Low Probabilistic High =50% High
Proof of Stake Low Probabilistic Medium =50% Low/Medium
Tendermint PoS Medium Deterministic Medium =33% Low/Medium
Delegated-Proof-of-Stake High Probabilistic Low =33% Low/Medium
Proof-of-Staked-Authority Medium Probabilistic Low =33% Low/Medium
Snowflake-to-Avalanche Medium/High Probabilistic Medium =50% Low/Medium
Proof-of-History High Deterministic Medium =33% Low/Medium
Fast Byzantine Fault Tolerance High Deterministic Medium =33% Low/Medium
Lachesis Consensus Protocol High Deterministic Medium =33% Low/Medium
et al., 2020) (Buterin, 2021). In the presentation of
the blockchain trilemma (Buterin, 2017) Buterin de-
scribed the three trilemma aspects as follows:
Scalability: Defined as being able to process O(n) >
O(c) transactions
Decentralization: Defined as the system being able
to run in a scenario where each participant only
has access to O(c) resources
Security: Defined as being secure against attackers
with up to O(n) resources
3.3.1 Scalability
Maximum throughput and latency of a network are
the most decisive indicators for scalability for users
who do not actively participate in the network (Cro-
man et al., 2016). Therefore, the maximum through-
put (how many transactions per second a network can
handle), the block time, and time to finality are se-
lected to measure the scalability of a network. Time
to finality is assessed based on a network being deter-
ministic or probabilistic. Block time is rated by the
seconds it takes for a network to create a new block,
which, is not used in this selection framework as all
selected blockchain platforms present a similar block
time. Thus, only the maximum transaction per second
as some blockchains are not fully implemented yet
and the time to finality (deterministic or probabilistic)
is used to measure the scalability of the blockchain
platforms.
3.3.2 Decentralization
According to Conway (Conway, 2022), decentraliza-
tion of Proof-of-Work networks is measured by its
hash rate and its distribution among the participants of
the network. A Proof-of-Stake network (and similar
blockchains) is measured by the number of validators,
the percentage of token supply staked, and the distri-
bution of the token supply across its validators (Con-
way, 2022). Following this, we calculate a decentral-
ization index for the selected Proof-of-Stake (or sim-
ilar) blockchain platforms by their average ranking
for the number of nodes, their percentage of supply
staked, and their Nakamoto Coefficient. This decen-
tralization index is used to determine how decentral-
ized a network is as the number of nodes may be mis-
leading due to the Nakamoto Coefficient. Ethereum’s
decentralization will be measured along with its peers
(other Proof-of-Work blockchains). A lower decen-
tralization index is favourable as it indicates that the
network is more decentralized.
3.3.3 Security
Security can be measured by the Byzantine Fault Tol-
erance of a network. An asynchronous network can
not provide safety (guarantee that all malicious nodes
will eventually agree to the new state) and liveness
(ability to process transactions) if the number of ma-
licious nodes exceeds the BFT threshold (Bracha,
1987). For networks with a Byzantine Fault Toler-
ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering
366
Table 3: Scalability of Blockchain Platforms.
current TPS max TPS Block Time Time to Finality
Ethereum 10 12-15 12-14 seconds 60 seconds
Cosmos /
10,000
per zone
˜6 seconds Instant
BSC 40-60 160 ˜3 seconds 75 seconds
Avalanche 5-10
5,000+
per subnet
˜2 seconds ˜1 second
Solana 1,500-2,500 710,000 ˜0.7 seconds Instant
Fantom 10-15 300,000 ˜1 second Instant
Tron 50-200 2,000 3 seconds 60 seconds
Polygon 30-50
65,000
per sidechain
2.3 seconds ˜2 seconds
Harmony ˜10
500
per shard
2 seconds Instant
Table 4: Decentralization of Blockchain Platforms.
Number
of Nodes
Type of Nodes
Fixed
number
of validators
Hashrate /
% of supply
staked
Nakamoto
Coefficient
Ethereum ˜6,000 Competing No 913.74 TH/s
1
3
*
Cosmos 175 Validators Yes 62.23% 7
BSC 21
Authorized
validators
Yes 81.47% 8
*
Avalanche ˜1,250 Validators No 60.82% 52
Solana ˜2,000 Validators No 73.79% 27
Fantom 92 Validators No 47.01% 3
*
Tron 27
Super
Representatives
Yes 45.81% 8
2
Polygon 100 Validators Yes 30.89% 13
*
Harmony
250
per shard
Validators Yes 42.48% 5
*
1
Ethereum is the only PoW platform, thus hashrate is used to measure decentralization.
*
Is estimated since no central source exists for the Nakamoto Coefficient for some.
Table 5: Decentralization Ranking of Blockchain Plat-
forms, with Number of Nodes NoN, Hashrate / % of sup-
ply staked HR, Nakamoto Coefficient NC, Decentralization
Index DI.
NoN HR NC DI
Ethereum 2 2 2 2
Cosmos 4 3 6 4.33
BSC 8 1 4 4.33
Avalanche 2 4 1 2.33
Solana 1 2 2 1.66
Fantom 6 5 8 6.33
Tron 7 6 4 5.66
Polygon 5 8 3 5.33
Harmony 3 7 7 5.66
ance of 33% a number of malicious nodes between
33% and 50% can already halt the blockchain so that
it can no longer produce new blocks (Bunin, 2022).
In addition, to Ethereum, Avalanche, and Polygon,
where >50% of the network needs to be malicious to
bring it to a stop, we also consider BSC secure all the
validators have to be approved by a central authority
and must publish their identity.
3.3.4 Resources and Repeatability
To analyse the different platforms we studied
whitepapers and online documentation. As those
sources mainly focus on the conceptual aspects,
such as maximum throughput or the functions of
their consensus protocol, block explorers are used
to get real-time and historical information on a
blockchain. Block explorers are mainly developed
by the blockchain foundation, reputable community
members, or former blockchain developers.
The process of analyzing and assessing the
blockchain platforms is repeatable for most of the
part. Scalability and Security of a network is mea-
sured with objective metrics. Only for decentraliza-
tion, we compared the selected blockchain platforms
to one another according to (Conway, 2022).
A Platform Selection Framework for Blockchain-Based Software Systems Based on the Blockchain Trilemma
367
Table 6: Security of Blockchain Platforms.
BFT Availability Anonymity
Ethereum 50% Transaction with low fees can become stuck (as miners re-
ceive the fee, a low fee does not offer any incentive to pro-
cess the transaction over other transactions with higher fees)
Data availability is achieved by full nodes
Pseudonymity
Cosmos 33% Validators are penalized for inavailability Pseudonymity
BSC 33% Validators are penalized for inavailability Pseudonymity
Avalanche 50% SoA can adaptably change byzantine fault tolerance for availability
Block and Transaction Data are simultaneously stored on Kyve
Pseudonymity
Solana 33% Horizontal scaling gives up network availability for scalability Pseudonymity
Fantom 33% Validators and delegators are penalized for inavailability Pseudonymity
Tron 33% Fees to prevent transaction flooding Pseudonymity
Polygon 50% Validators and delegators are penalized for inavailability
Achieves data availability by the means of an additional data layer on the
blockchain
Pseudonymity
Harmony 33% Shards store only 1/n of the global state, new blocks from shards are crosslinked
to the beacon chain.
Pseudonymity
4 SELECTION FRAMEWORK
This selection framework aims to facilitate deciding
which blockchain platform to use. It does not provide
any method to decide whether the use of a blockchain
(for a particular application) is reasonable or not. The
user should have already made this decision. It is also
assumed that the user knows their specifications re-
garding needed and desired scalability, decentraliza-
tion and security.
4.1 Development Process
The development of the selection framework started
with finding attributes which best split the selection
of platforms into two homogeneous parts. However,
finding a starting attribute that allowed all users to fol-
low their wanted aspect of the blockchain trilemma
was impossible.
Thus, the final selection framework starts with
the question ”What is most important” with all
three aspects as answer possibilities. This split no
longer produced pure nodes but mixed nodes where a
blockchain platform could be part of two or all three
pathways. Therefore, some platforms can be reached
through different paths as they fulfil multiple crite-
ria. The subsequent nodes in the selection framework
were used to further down-sample the selection set.
Generally, at each decision node, the user can
choose between one of the three trilemma aspects
where scalability includes all platforms with a max-
imum tps rate greater than Visa’s 1,700 tps. Decen-
tralization is measured with the help of the Decen-
tralization Index presented in Table 5, where all net-
works with an index <4 are considered to be more
decentralized than others. This split was chosen due
to the (large) gap in the decentralization index be-
tween Avalanche (index of 2.33) and the next best
platforms, BSC and Cosmos (index of 4.33). Secu-
rity embraces all platforms with a byzantine fault tol-
erance of 50% and BSC where the validators are
publicly known. Section 4.2 gives a detailed overview
of all decision nodes.
Figure 1 shows our selection framework where
multiple questions must be answered before the selec-
tion framework terminates with a suggestion. Further-
more, for some pathways, the selection framework
terminates with two possible blockchain platforms.
This is due to their similarity when measured in terms
of the blockchain trilemma. In this case, further study
of the differences between the two platforms is rec-
ommended.
4.2 The Selection Nodes
a: What is most important? The first question
asked is what is most important for the user. The
three choices offered are the three aspects of the
blockchain trilemma.
b: More scalability, or decentralization or secu-
rity? Following the scalability path from the first
decision, the next question is what is the second
most important to the platform. Yet again, the user
is confronted with all three blockchain trilemma
aspects.
c: Scalability or Security? The 2nd question for de-
centralization is whether the platform should fo-
cus more on scalability or security.
d: Validators publicly known? This question is to
differentiate BSC from other secure platforms, as
ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering
368
Figure 1: Blockchain Platform Selection Framework.
BSC is the only platform in this selection where
the validators have to publicize their identity. So
if the validators should be known, we come to our
first result. Else the selection framework contin-
ues.
e: Decentralization or Scalability? The third ques-
tion for a secure platform also concerns the other
two trilemma aspects.
f: Finality? This question follows the question of b:
More scalability, or decentralization or secu-
rity? and divides the subset of scalable platforms
based on their block finality.
g: Is also some decentralization needed? If the
platform should offer instant finality, the last
question in the selection framework is the
question of decentralization. Suppose less decen-
tralization is required, the selection framework
points to Fantom and Harmony. If, however, the
platform should also be more decentralized, it
terminates with Solana and Harmony as a result.
h: Pure scalability? If the answer to question f: Fi-
nality? is no and probabilistic finality is suffi-
cient, the next question is whether the platform
should offer some decentralization or security. If
the answer is no and only scalability is required,
the result of the selection framework is Tron. Else
question k: Decentralization or more scalabil-
ity? follows.
i: Security or more scalability? This question can
be reached from decision b: More scalability,
or decentralization or security? if chosen de-
centralization or from question c: Scalability or
Security? if chosen Scalability and yields three
different results. If the platform should be secure,
it points to Avalanche. If it should be scalable,
suggestions are Solana and Cosmos.
j: More decentralization or scalability? Just like
the last decision, this question can also be reached
from two former questions and confronts the user
with the question of whether more decentraliza-
tion or scalability is required. If the answer is
decentralization, the selection framework points
to Ethereum, the only platform in this selection
where every node has the same rights. Else the
selection framework results with Avalanche.
k: Decentralization or scalability? This question
can be reached if both security and scalability
have been chosen before and queries whether
decentralization or scalability is more important.
If the former is chosen, the selection framework
terminates with Avalanche, otherwise, with
Polygon.
4.3 Observations
The presented selection framework provides only a
limited summary of a subset of selected blockchain
platforms. Many platforms appear to be similar
to others but utilize different underlying technology
(i.e., consensus protocol, the structure of chain, data
storage, etc.). Therefore, solely relying on this selec-
tion framework is not sufficient and all the technical
details should be researched. This is particularly ev-
ident if we look at the result of ”Fantom, Harmony.
Cosmos” when selecting ”no” at node g: Is also some
decentralization needed?. While all three are re-
sults of the same pathway, Fantom is a DAG-based
blockchain with a maximum of 300,000 transactions
per second, Harmony is a sharded blockchain where
A Platform Selection Framework for Blockchain-Based Software Systems Based on the Blockchain Trilemma
369
each shard supports only up to 500 transactions per
seconds but can reach higher transactions per seconds
according to its whitepaper, whereas Cosmos is a plat-
form to build blockchains on which can horizontally
scale and reach thousands of transactions per second
per zone. All three platforms are also similar in terms
of block time, deterministic time to finality, decentral-
ization (index of 6.33 for Fantom vs index of 5.66 for
Harmony vs index of 4.33 for Cosmos) and security
( 33%BFT ).
5 EVALUATION OF THE
FRAMEWORK
We apply the selection framework to a real-world ap-
plication setting. A common field where blockchain
technology is applied is public services and govern-
ment. We apply our selection framework to a decen-
tralized exchange platform and a conceptual example
by means of e-voting.
Following this, the presented selection framework
could have been of use for SushiSwap to decide
which blockchain platform to build their application
on. However, given the fact that SushiSwap is a fork
of another decentralized exchange, UniSwap, which
launched in November 2018. Back then, Ethereum
was one of the only blockchain platforms to exist that
offered Smart Contract support. Most of the platforms
presented here only launched after UniSwap’s initial
launch.
In our local 2018 provincial elections a total of
284,361 voters cast their votes from 7am to 9pm on
the 21
st
October 2018. This corresponds to a voter
turnout of 73.87%. In the morning (until 11am) only
19.6% voted but this number increased in the after-
noon as 50.5% of all eligible voters casted their vote
until 5pm . At the closure of the election stands a total
of 73.87% voted.
Table 7: 2018 Provincial Elections in numbers.
Time Voters % votes/h votes/s
7-11:00 75,241 19.6% 18,810 5.23
11-17:00 118,200 30.9% 19.700 5.47
17-21:00 89,479 23.4% 22.370 6.21
Total 282,920 73.9% 20.209 5.61
From Table 7 we can see that on average 5.61
votes were cast by second, alternating slightly de-
pending on the time of the day. Therefore, an e-
voting application for provincial elections must be
able to process at least six votes per second. However,
this number does not take into account that with e-
voting, the distribution of votes cast may change and
the number of votes increases, which is expected with
e-voting (Anane et al., 2007).
Security is important in e-voting systems (Anane
et al., 2007) (Abuidris et al., 2019). Both see
anonymity as one of the most important points for an
e-voting system, something that no platform in this
selection offers as they all offer only pseudonymity
(see table 6).
Disregarding this concern, we can move from
node a: What is most important? to node d: Val-
idators publicly known? following the security path.
At this point, choosing BSC as the platform would
also fulfil the provincial election transactions per sec-
onds requirement. But also, the paths which lead to
Ethereum, Avalanche, and Polygon are imaginable.
Another aspect identified by Abuidris et al. is scalabil-
ity (Abuidris et al., 2019) which leads us to node k:
Decentralization or more scalability? from where
one can choose between high scalability or decen-
tralization. At this point, a precise suggestion of a
blockchain platform for an e-voting application is dif-
ficult to make as we are faced with the blockchain
trilemma. Is it more convenient to trade decentral-
ization for scalability and choose Polygon, or is also
decentralization required for an e-voting application
and Avalanche is the better blockchain platform to
choose?
In terms of scalability, both platforms are quite
similar, whereas Avalanche is slightly faster in cre-
ating blocks and reaching consensus, but Polygon
can handle more transactions per second. An as-
sumption at this point is that the decentralization and
adaptive changeable Byzantine Fault Tolerance from
Avalanche are advantageous.
The objective of the selection framework is to fa-
cilitate the selection of the correct platform. In the
case of the presented selection framework, it helps
to decide which platform from the selected and an-
alyzed platforms should be considered for usage. For
this, the selection framework uses the blockchain
trilemma. To answer the second question, the users
must know their specifications and what they expect
from the blockchain platform. If the users do not
know if they prefer scalability over security or de-
centralization, the selection framework is not helpful.
However, it is also not designed for that purpose. If
the selection framework is helpful to the users in any
way of narrowing down their final selection or elim-
inating some blockchain platform, it is considered to
be helpful and adds value to the user.
The e-voting case presented above is a concep-
tual example of how the selection framework could be
used to choose the best possible blockchain platform
for a decentralized application. By knowing what is
ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering
370
important to their application, the user can narrow
down the selection of platforms.
6 CONCLUSIONS AND FUTURE
WORK
In order to support a system architecture by trusted
storage, the decision on the right blockchain plat-
form depends on a range of required system quali-
ties. Blockchain platforms offer more trust, security,
and privacy as principal benefits, but system proper-
ties such as scalability or degree of decentralisation
are equally important in distributed systems. We com-
pared selected blockchain platforms in terms of scal-
ability, decentralization, and security. The selection
was made on multiple criteria, such as trilemma prop-
erties, their type of blockchain, and their initial to-
ken allocation. For the analysis of each platform, we
studied their respective whitepaper and documenta-
tion and also interacted with the platform’s commu-
nity on social media platforms such as Twitter, Red-
dit, and Discord. However, as such sources cover only
the platforms’ conceptual aspects, we also utilized
websites which track the analytics of each blockchain,
so-called blockchain explorers. The results of the
analysis indicated that the blockchain trilemma holds
true.
We developed a selection framework based on the
trilemma properties where each split asks for the most
important aspect of the trilemma. Only later splits
ask specific questions about the required scalability
of the blockchain platform taking specific metrics into
account.
REFERENCES
Abuidris, Y., Kumar, R., and Wenyong, W. (2019). A survey
of blockchain based on e-voting systems. In Intl Conf
on Blockchain Technology and Applications.
Anane, R., Freeland, R., and Theodoropoulos, G. (2007).
E-voting requirements and implementation. In Pro-
ceedings CEC-EEE 2007.
Bracha, G. (1987). Asynchronous byzantine agreement pro-
tocols. Information and Computation, 75(2):130–143.
Bunin, V. (2022). Proof of stake’s security model is being
dramatically misunderstood. Medium.
Buterin, V. (2017). Sharding faq. https://vitalik.ca/general/
2017/12/31/sharding faq.html.
Buterin, V. (2021). The limits to blockchain scalability.
https://vitalik.ca/general/2021/05/23/scaling.html.
Conway, L. (2022). Measuring decentralization: Is your
crypto decentralized? Blockwors.
Croman, K. A., Saxena, P., Shi, E., G
¨
un Sirer, E., et al.
(2016). On scaling decentralized blockchains. In Intl
conf on financial cryptography and data security.
Hafid, A., Hafid, A. S., and Samih, M. (2020). Scaling
blockchains: A comprehensive survey. IEEE Access,
8.
Pahl, C., El Ioini, N., and Helmer, S. (2018). A decision
framework for blockchain platforms for iot and edge
computing. In IoTBDS.
Srinivasan, B. S. (2017). Quantifying decentralization.
Medium.
Zhang, R., Xue, R., and Liu, L. (2019). Security and privacy
on blockchain. ACM Computing Surveys (CSUR),
52(3):1–34.
A Platform Selection Framework for Blockchain-Based Software Systems Based on the Blockchain Trilemma
371