-In recent years, microservice architecture (MSA) has become popular. Emerging from the agile community, MSA implies a number of small, independently deployable microservices. They are characterized by low coupling, high cohesion, low complexity, and are more flexible and convenient in application, saving resource efficiency.However, there is limited research on quality models for MSA. Although MSA is a special form of service-oriented architecture, there are still some differences between the two that are hard to ignore, such as decentralization, smaller service size and encouraging technical heterogeneity. Therefore, it is difficult to directly apply the traditional quality model research of service-oriented architecture to MSA. Based on the above considerations, this paper proposes a quality model for MSA, which reflects the quality characteristics of microservices through 16 measures for each of the Functional Suitability, Flexibility, Interactivity, Performance Efficiency, Reliability and Security. At the same time, to address the shortage of theoretical validation of current research, this paper uses static analysis and dynamic analysis to validate the proposed measures and prove the rationality of its theory.
A microservice architecture (MSA) is a variant of the
Service- Oriented Architecture (SOA) structural style. As to
Martin Fowler, it is to build an application as a group of
services each designed for a specifific business capability, and
intercommunicate via lightweight mechanisms while being
independently deployable [
Most of the software quality measures used in the early
process-oriented paradigm of software systems were oriented
towards underlying software measures, such as the number of
codes, the complexity of functions or control flows, etc.After
object-oriented programming became the dominant
programming paradigm, existing measures and standards may
not be fully applicable to object-oriented software systems,
which has led researchers to reevaluate existing measures and
propose new ones at the same time. Among them, the most
influential ones are the C&K measure set proposed by
Chidamber and Kemerer [
The remainder of this paper is organized as follows. Section Ⅱ analyzes and quantitatively expresses the MSA, Section Ⅲ proposes an MSA-oriented quality model, Section Ⅳ validates the quality model results, and the summary and future work are in Section Ⅴ.
In order to clearly and accurately describe the MSA and define the measures proposed subsequently, this subsection abstracts the microservice architecture of interest by applying the approach of Yang et al. [1] to this paper and formalizes it accordingly. The conceptual software architecture of the entire microservice system can be represented in the form of Figure 2.1. Based on Yang et al.'s study, we added the relationship between the microservice system and external programs as an attribute .
represents the dependency relationship between modules in Microservices . , ), ℎ . , . ∈ is the module that depends on other modules in the corresponding dependency, and is the module on which other modules depend in the corresponding dependency. The whole microservice system can be formalized as a collection of microservices, and dependencies between microservices. Integrated system(IS) refers to the sum of microservice system and external programs
For a single microservice in a microservice system, formalize each microservice in the following representation. = ( , , ) denotes a collection of modules in a microservice . denotes the set of interfaces in a microservice .
= { | ∈ 0} 0 means natural number.
= { | ∈ 0} = (
, ) = { | ∈ 0} represents dependencies between microservices and other microservices. is the microservice that depends on other microservices in the corresponding dependency, is the microservice on which other microservices depend in the corresponding dependency, and is the interface involved in the corresponding dependency.
EPR represents the relationship between the microservice system and external programs.
MSA consists of one microservice unit, each of which has the ability to perform functions independently; at the same time, microservices show the characteristics of low coupling and high cohesion, making the whole MSA more flexible, and each module and each microservice unit has the ability to be reused; in addition, due to the low complexity and small scale of microservice system, the resources required for execution are lower and more efficient. To address the advantages and characteristics of MSA, we developed a quality model of MSA by combining the research of Michel-Daniel et al.[1]
Software quality models can help us to better propose and
apply measures, and usually measure models describe the
entities, attributes, and relationships of measures, and
currently common measure models in the field of software
engineering include the GQM (Goal Question Measure)
model [
The GQM model is the "Goal-Problem-Measure" model, which is one of the common measures models in software engineering practice, and is based on the idea that measures are measured by answering specific questions about the goal of the measure. The GQM model consists of three main layers: the conceptual layer, the operational layer, and the data layer. The conceptual layer is the goal to be measured, and the operational layer decomposes the abstracted goal into concrete questions. The data layer will give specific measures for these questions. By deriving the measure values of the measures and thus answering the questions, the specific measure of the target is finally obtained. The framework of the approach is shown in Figure 3.1.
The QMOOD model is a four-level hierarchical quality
model proposed by Bansiya and Davis [
The final quality model for the microservice-oriented architecture is shown in Figure 3.3. The first layer is the sum of the quality attributes to be evaluated in this paper. The second layer is the software attributes that can reflect the characteristics of the microservice architecture, which are Functional Suitability, Flexibility, Interactivity, Performance Efficiency, Reliability and Security. The third layer is the specific measures, which are 16 in total. These 16 measures each of the six software attributes in different aspects. The fourth layer is the object to be measured by the proposed measures. Considering that the microservice architecture advocates technical heterogeneity, the selected measures are to a certain extent independent of the programming language and technical implementation, this paper selects modules, interfaces, microservices and their related relationships as the measures of microservices themselves based on the characteristics of microservices. At the same time, we add the external system as the measure object because we have to consider the interaction between microservice system and other systems. C. Study On measures Of MSA Quality Model
This section measures MSA by Functional Suitability, Flexibility, Interactivity, Performance Efficiency, Reliability and Security. Each of the six software attributes and their measures will be described below.
This attribute measures degree to which MSA provides functions that meet stated and implied needs when used under specified conditions.
a) FAM(Functional appropriateness measures): This measure measures degree to which the functions facilitate the accomplishment of specified tasks and objectives. The user is counted to demonstrate the steps necessary to complete a task as well as any unnecessary steps. The ratio of the former to the latter is the measure for this section.
b) FCM(Functional correctness measures): This measure measures degree to which MSA provides the correct results with the needed degree of precision.
This attribute is an upgraded version of maintainability, which reflects the concept of high maintainability of microservice systems ---- coupling, cohesion and complexity, and also reflects the ability of microservice systems to be developed twice, which is the concept of reusability we proposed.
a) CPM(Coupling measures): The coupling measure measures the degree of dependency between elements within a microservice and between a microservice and other microservices. Loose coupling has a positive impact on flexibility.The measures proposed in this paper quantify the coupling at two main levels, the microservice implementation element level and the microservice level, respectively. The main concerns include the dependency relationships between modules belonging to the same microservice and the invocation relationships between microservices generated through interfaces.
• Internal Microservices Coupling of Model，IMSCM.
The IMSCM calculates the sum of the dependencies
between the modules inside the microservice, i.e., the
connections between modules [
= | |
Weighted Microservices Coupling of Interface ，
WMSCI. WMSCI measures the degree of dependency
between microservices and microservices in a system,
and in this measure, this paper abstracts the coupling
to the microservice level, and the specific measures
include the invoking and invoked relationships. In this
paper, we refer to the study of Kulesza et al [
•
Microservices Cohesion of Interface Data，MSCID. MSCID quantifies the cohesiveness of a given microservice by measuring the degree of similarity of the parameters passed in the interfaces exposed by its microservices; a microservice is highly cohesive if all interfaces work on the same type of input parameters.
Microservices Cohesion of Interface Usage，MSCIU. MSCIU quantifies the cohesiveness of a given microservice by
measuring the invocations of that
interface by other microservices. A microservice is
considered highly cohesive when each user of the
microservice (microservice consumer) invokes all the
public interfaces of the microservice, which can be
considered highly relevant for implementing a certain
functionality.
[
module lead to potential changes in other modules within the microservice. This is expressed as the possible information flow and dependencies between modules, which can be obtained through the passing of dependencies. If a module is directly or indirectly dependent on a large number of other modules, the more likely it is that changes to this module will affect other modules.
MSMPC is calculated as follows.
= ∑
∑ be connected by the passing of dependencies, i.e., whether the value of the passing closure matrix at their locations is not zero. •
Microservices Parameter Count ， MSPC.
MSPC mainly measures the
number of data structures appearing in the microservice, and in this paper we mainly consider the number of parameters of the interface, which includes its own interface and the interface of invoking other microservices. MSPC is calculated as follows.
= ⋃
, ∈ | =1
|
d) RM (Reusability measures): Reusability is a very
desirable quality measure for industry [
• •
The degree of reuse of microservices in secondary development.
Counts the number of times a microservice module is reused in the execution of different services to characterize its degree of reuse. Time efficiency of reuse of microservices in secondary development. Characterize the reuse time efficiency of a microservice
module by counting the time it consumes to reuse it during the execution of different services.
This section examines the interaction between the MSMPC refers to the visibility theory from the study the monetary cost of the resources used for running the = |{ ∈ | .
= }| | | ∗ | | •
Microservices Cohesion of Model ， MSCM. This measure abstracts the methods and properties in a class to the module level in a microservice. If the number of connectivity dependencies graphs formed by
all of a microservice is modules and 1, then the microservice is cohesive to a certain extent. However, considering that many microservice architectures manage the operations related to interfaces through a unified module, which often leads to the module becoming the hub of an otherwise unconnected graph, the module and its dependencies are removed from the connected graph when using this measure for cohesion measurement. ( − 1) and are modules in the microservice, the same connectivity graph, and its return value is 1 or 0. ( , ) is to calculate whether and belong to
Complexity primarily
measures the complexity of the microservice implementation and execution functions, and more broadly includes the ease with
which developers can perceive, understand, and modify them. High complexity can have a negative impact on flexibility.In this paper, complexity is measured by the following two measures, taking into account the complexity of the microservice implementation itself and the complexity of developer awareness and understanding. •
Microservices Model Propagation Cost ， MSMPC.
microservice. The cost can be estimated at design time and
corrected after the implementation is evaluated at execution
time.[
b) RT (Response time): the anticipated delay between
the time when a request to a microservice is issued and the
time when the result is delivered.The average response time
of all requests over a period of time is measured, and the
smaller the average response time, the faster the processing
speed and the more efficient the service.[
n=Number of responsed measures 5) Reliability: degree to which a system, product or component performs specified functions under specified conditions for a specified period of time.
a) SER (Successful execution rate): the ability of a service provider to successfully fulfil the requests within a given period of time. It is measured as a number between 0 and 1 or a percentage calculated as the ratio between successful requests and the total number of requests.
b) Sca (Scalability): the ability of a microservice to function correctly (as designed) irrespective of the changes in size (amount of resources) without inquiring performance penalties. the degree of scalability can be calculated by analyzing the distribution of synchronous requests provided by the exposed interfaces, a high diversity of requests indicating poor scalability.
management): a quality attribute describing the ability of a microservice to cope with failures. A
microservice complies to this property by saving the
internal state, and restarting automatically while loading the
most up-to-date state prior to the failure. It is a binary
attribute with ”yes” or ”no” values and it is verified via
instance graphs or type graphs [
6) Security: degree to which a product or system protects information and data so that persons or other products or systems have the degree of data access appropriate to their types and levels of authorization.
a) CM (Cofidentiality measures): degree to which a product or system ensures that data are accessible only to those authorized to have access.
= 1 − / A=Numbers of cofidentiality data items that can be accessed without authorization B=Number of data items that require access control b) IM (Integrity measures): degree to which a system, product or component prevents unauthorized access to, or modification of, computer programs or data.
A=Number of data corruption prevention methods actually B=Number of data corruption prevention methods available
= / implemented and recommended
IV. VALIDATION OF QUALITY MODELS
measure-based evaluation is compulsory for assessing the
quality attributes of microservices. A vast number of quality
criteria
measuring
a
variety
of
aspects
concerning
microservices exist [
It is a technique that does not require execution of the analyzed software. Scanning the code of a microservice before being linked to other microservices can aid the identification and correction of vulnerabilities without incurring the costs of running the code.
It is a technique that requires the execution of the analyzed software, usually employed when the application code is not available. The
most common type of dynamic analysis consists of Unit Tests and it has the benefit of validating static analysis findings or identify new flaws.
C. Validation test result Analysis:
a) Functional Suitability Validation Analysis: The analysis of Functionality Suitability is a dynamic analysis. To verify the functionality of the microservice, the program of the microservice unit is executed and then compared with the expected result, if the result meets the expectation, it means the Functionality Suitability is good.
b) Flexibility Validation Analysis: The analysis of Flexibility belongs to a combination of static and dynamic analysis.
By studying the modules and interfaces of microservices and their relationships, we can determine the coupling, cohesion, and complexity of microservices. These belong to static analysis; however, when studying the reuse of microservices in secondary development, we need to run microservices in different development environments, which belongs to dynamic analysis.
c) Co-existence measures Validation Analysis: Coexistence
measures means that the interaction between the microservice architecture and the external program does not affect the execution of their respective functions, and that no exceptions occur in either program during the execution of the test. Obviously, we need to make the microservices system and the external program run simultaneously, which is a dynamic analysis .
d) Confidientially measures Validation Analysis: The extent to which the microservices system ensures that only authorized people have access to the data. The detection of such measures is a dynamic analysis and requires verification that unauthorized users have access to the data.
Microservices is a young research area, and there is still relatively little work on quality models for microservices architectures. Although At the same time, the current research on quality models for services often lacks theoretical validation, which may be more important than quantitative analysis. Based on this, this paper conducts quality modeling research for microservice architectures, and its main work and contributions include.
A quality model is proposed for the microservice architecture, which reflects the quality characteristics of microservices by measuring the six software attributes of microservices Functional Suitability, Flexibility, Interactivity, Performance Efficiency, Reliability and Security through 16 measures. The former considers the compatibility between the microservice architecture and external programs, and the latter adds the examination of the reusability of the microservice architecture on the basis of maintainability.
We designed a combination of dynamic and static analysis to validate and analyze the MSA quality model with key indicators to increase the reliability of the results.
Although this paper establishes a quality model under microservice architecture and conducts theoretical validation to prove its effectiveness, there are still some shortcomings and areas that need further research. For example, other quality model measures will be added, such as cohesion attributes that are relatively incomplete, and subsequent research can measure cohesion through semantic analysis to achieve a more accurate assessment of flexibility. Moreover, a more comprehensive quantitative study will be conducted subsequently after more industrial data are collected to further prove the validity of the model.