Towards IT Workload Hybrid-Cloud Placement Advisory in Enterprise

Andr

e Hardt, Abdulrahman Nahhas, Hendrik M

uller and Klaus Turowski

Faculty of Computer Science, Otto von Guericke University, Magdeburg, Germany

{andre.hardt, abdulrahman.nahhas, hendrik.mueller, klaus.turowski}@ovgu.de

Keywords:

Enterprise Applications, SAP, Cloud Placement, BPMN.

Abstract:

Placement of IT workloads in a cloud or hybrid-cloud environment is not always straightforward and requires

taking into account various requirements, cloud offering capabilities, and costs. This fact has led researchers

and industry practitioners to develop various automation solutions to support this decision process. However,

the exact procedure for applying these solutions in practice, especially in the enterprise environment, is typi-

cally not discussed. In this work, we propose a formalized systematic business-centric process for delivering a

service that relies on a data-driven automation solution, as a tool for experts, for relevant data management and

placement optimization in a hybrid-cloud. We performed preliminary ﬁeld testing of the proposed approach on

real-world enterprise IT landscapes running SAP enterprise applications with the application of a user-friendly

placement optimization automation solution. Finally, the stakeholder feedback and key takeaways from the

ﬁeld testing are summarized, noting the feasibility and potential usefulness of the presented formalized pro-

cess.

1 INTRODUCTION

Selecting cloud services and workload placement, in-

cluding Enterprise IT workloads, in a public cloud

and hybrid-cloud environments can be a daunting

task. And when a cloud strategy is unsuccessful, it

can lead to signiﬁcant negative outcomes (Venkatra-

man and Arend, 2022). Tackling such challenges

can be difﬁcult, which, as discussed further in sec-

tion 2, has led researchers and practitioners to propose

and develop various tools that can assist in relevant

decision-making steps.

In this work, we aim to propose a business process

model derived from empirical observations and in

close collaboration with the industry experts. It relies

on an automation tool designed to support decision-

makers in selecting optimal placement for their en-

terprise IT workloads. Speciﬁcally, we focus on

the placement of the standard enterprise applications

(EA) in a hybrid-cloud environment. Our proposed

business process model incorporates the interaction of

non-technical decision-makers, domain experts, and

the placement recommendation tool.

Enterprise applications placement and selecting

infrastructure for these are complex tasks that require

considerable expertise and insight into business pro-

cesses to make a sound placement decision. There-

fore, we suggest that the centralization of this exper-

tise backed by an automated solution can lead to use-

ful outcomes. Furthermore, we argue that formalizing

a process of applying such a solution in an enterprise

environment may help to bridge the gap between aca-

demic research and industry, as it has the potential to

propose a clear procedure for applying algorithmic-

based (i.e., heuristic, metaheuristic, machine learn-

ing) automation solutions in real-world enterprises.

2 RELATED WORK

A number of tools were proposed over the years to

assist decision-makers in selecting the optimal place-

ment for various systems in the cloud. Dubbed

CloudAdvisor is such an example tool (Sahu et al.,

2024) that consists of a web frontend and a backend.

The rich web frontend is used to visualize and present

possible placement options in the public cloud, while

the backend relies on a combinatorial optimization

method to compute possible placements. Addition-

ally, the collection of user-deﬁned requirements and

constraints is done via the front as well. The combina-

torial optimization is focused on selecting the place-

ment option for the user’s data in the cloud so that

access latency is minimized. The stated target user

base of the tool is the consultants in the area of cloud

resource management.

CloudRecoMan (Mettouris et al., 2022) is another

830

Hardt, A., Nahhas, A., Müller, H. and Turowski, K.

Towards IT Workload Hybrid-Cloud Placement Advisory in Enterprise.

DOI: 10.5220/0013361800003929

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 830-839

ISBN: 978-989-758-749-8; ISSN: 2184-4992

similar tool, similarly presenting a web-frontend and

a backend with a recommender system. Within the

proposed methodology on which CloudRecoMan is

based, the recommendation for cloud placement se-

lection is made based on the company proﬁle entered

into the system by a decision-maker who does not

necessarily possess sufﬁcient technical background.

Another company-oriented approach was pro-

posed earlier than the two aforementioned ones and

based on a muti-criteria-decision-method. The ap-

proach is named (MC

)

(Menzel et al., 2013), with

a web-frontend and a backend prototype architecture.

The user is required to supplement placement alterna-

tives, which can either be determined by the experts

and supported by integrating external data sources.

The user must also supply criteria and requirements,

in the speciﬁc format, that will be used for ranking the

placement alternatives. Certain aspects of cloud alter-

natives, such as performance, are evaluated by inte-

grating external measurement benchmarks.

Another tool named CloudAdvisor, with no rela-

tion to the aforementioned tool of a similar name, was

proposed as a service-oriented solution (Jung et al.,

2013) for recommending cloud placement conﬁgura-

tion based on the application workload. Similarly to

all aforementioned solutions, the proposed solution

has a frontend view for the users as well as the back-

end where the necessary calculations are performed,

including a combinatorial search of suitable alterna-

tive placement solutions. The solution is corporate-

oriented and focused on VM placement. The user

must specify the workload information via the fron-

tend, as well as the performance, energy consump-

tion, and budget information. After that, the user is

presented with a set of alternatives and their costs for

cloud placement for the given workload.

All of the aforementioned examples of workload

placement selection methodologies and tools follow

different approaches but generally present the tool

that is designed to assist the decision-makers using al-

gorithmic solutions. However, most of these require

both business and technical understanding to oper-

ate. The only exception is CloudRecoMan (Mettouris

et al., 2022), which is a recommendation tool based

on the company’s business proﬁle, rather than details

of the IT workload.

The delivery model of the solutions discussed

above presumes end-user-oriented use, where the user

interacts with the solution via a graphical frontend.

However, the exact business process of delivery and

application for these automation tools or services in

an enterprise environment is not discussed in detail

for any of the proposed methods. In this work, we

aim to lay the foundation for bridging this gap.

3 PROPOSED PROCESS

In this work, we propose a formalization of a busi-

ness process for delivering an advisory service fo-

cused on optimization of enterprise application place-

ment in a hybrid-cloud. It is aimed at larger enter-

prises. We argue that automation solutions, such as

discussed in section 2, can be used to support the ex-

perts by automating the tedious tasks of data man-

agement, cost estimation, placement optimization cal-

culation. However, it is important to clearly under-

stand how such solutions can be applied in real en-

terprises, where various stakeholders and domain ex-

perts must be involved to achieve the goal of hybrid-

cloud adoption or sizing for existing IT infrastructure.

We believe that formalization of the application might

positively affect the adoption of automation solutions

based on state-of-the-art algorithms to support experts

in enterprise environments.

To the best of our knowledge, there is no pub-

lished literature proposing a business-centric formal-

ization of a process for providing an advisory service

of enterprise IT systems placement optimization in a

hybrid-cloud, while relying on a data-driven automa-

tion optimization system for supporting the experts

in the ﬁeld. While there are proposed tools, there is

a lack of discussion on how precisely these can be

applied in a real enterprise environment. Such envi-

ronments typically involve various experts and stake-

holders, which might beneﬁt from being supported by

a data-driven automation solution for placement op-

timization. The proposed formalization is the result

of collaboration with ﬁeld experts and empirical ﬁeld

testing.

The focus of the proposed process lies in the in-

volved business activities. Therefore, we present a

high-level overview of this process as a Business

Process Model and Notation (BPMN) (OMG, 2010)

model in Figure 1.

As seen from the aforementioned ﬁgure, the pro-

cess encompasses two main actors. First is the

hybrid-cloud Placement Optimization Service (HC-

POS), which in turn encompasses the placement ex-

perts as well as the automation system for placement

selection optimization. This service relies heavily on

the automation provided by an Optimization Automa-

tion System (OAS), which is discussed in more detail

further in subsection 3.2. The OAS is operated by the

domain experts, whose tasks and roles are discussed

in subsection 3.3.

The experts interact not only with OAS but also

communicate with the second actor of the proposed

process. This actor is the organization, which is a

consumer of the provided service and the ﬁnal recip-

Towards IT Workload Hybrid-Cloud Placement Advisory in Enterprise

831

ient of the IT workload placement recommendation

for the IT landscape. In the business environment,

it can be seen as essentially a ﬁrst actor’s customer

according to the proposed process. The assumptions

and tasks of the organization are discussed further in

subsection 3.1.

3.1 Organization

As mentioned before, under the term organization, in

our proposed process, we understand the ﬁnal recip-

ient of the service, or in other words, the customer.

As such, the stakeholders within the organization ini-

tiate the entire process. Speciﬁcally, it is expected that

the organization must state a goal for its IT landscape

transformation and (hybrid) cloud adoption or resiz-

ing. The expectation is that the organization involves

the relevant stakeholders on the business site in the

process. It is worth noting that in a sufﬁciently large

enterprise, the organization (customer) might as well

be a subdivision instead of an external service con-

sumer.

The proposed process is oriented on IT landscape

transformations in organizations that already have

running IT landscapes. The initiation of such IT

landscape transformation can be driven by a variety

of business goals, which, in the end, directly inﬂu-

ence the functional and non-functional requirements,

as well as the outcome of such a transformation.

The primary motivation for designing this process

speciﬁcally for the situation where the IT landscape

already exists, instead of providing a ”green ﬁeld”

assessment, is that the existing infrastructure can be

measured. Speciﬁcally, we rely on the data acquired

by recording performance counters reﬂecting the sys-

tem performance over a pre-determined, representa-

tive set of time. These performance counters include

the capacities and consumption of the available com-

putational resources (i.e., CPU, main memory, stor-

age, network) as well as the application-speciﬁc mea-

surements (e.g., number of transactions, number of

users, etc.).

These measurements allow us to construct a repre-

sentative workload proﬁle. Such a proﬁle reﬂects the

capacity needs of the existing IT landscape within the

real-world load, reﬂecting the real business needs of

the organization. Furthermore, these measurements

might reveal capacity deﬁciencies of the existing in-

frastructure, which might hinder the performance of

the systems within the IT landscape and lead to neg-

ative consequences for the dependent business pro-

cesses. This data and insights derived from it can

Planning and implementation of an IT infrastructure

from ground up without existing legacy systems.

serve as a basis for decision-making in an IT land-

scape hybrid-cloud transformation or resizing, as it

reﬂects the existing business processes and needs in

terms of computational resources.

3.2 Optimization Automation System

One of the main assumptions of our proposed pro-

cess is the existence of an optimization automation

system (OAS). Such a system is a ready-to-use solu-

tion that is accessible to its users by means of a graph-

ical user interface (GUI) and requires no technical or

software engineering knowledge from the end users.

Typically, such OAS is a solution consisting of fron-

tend and backend components.

The backend component consists of the databases,

data processing routines, and business rules imple-

mented in the logic of a chosen programming lan-

guage, as well as possibly complex computational al-

gorithms (heuristic, metaheuristic, or machine learn-

ing). The backend can also provide various appli-

cation programming interfaces (API), for integration

with other information systems. However, the com-

plexity or intertwined components that are critical to

the functionality are obscured from the end users. The

internal or external IT teams carry out the hosting and

maintenance.

The frontend is the component that provides the

user-friendly GUI and is the main layer of interaction

with the OAS by the end users. Within our proposed

process of overall service delivery, the GUI should

be oriented toward the end users who are experts in

the primarily business-oriented ﬁelds but with enough

technical background to formulate inputs and inter-

pret outputs of the OAS. Deep knowledge of the in-

ternal functionalities of the OAS and its components

(e.g., internal optimization algorithms, cost models,

requirements model) is not a prerequisite for the end

users.

3.3 Cloud Placement Experts

As mentioned before, within our proposed process,

under the term experts, we mean the cloud and the IT

software experts on the side of the service provider.

The speciﬁc focus of our service delivery process is

the placement of the standard off-the-shelf IT appli-

cations in the hybrid-cloud infrastructure. Therefore,

the typical cloud infrastructure knowledge of the ex-

pert users will lie in the business-process-focused do-

mains surrounding the target enterprise application

(EA). Such areas include cloud services and architec-

tures, capacity sizing of the server solutions, licensing

issues, operational costs, and interpretation of busi-

ICEIS 2025 - 27th International Conference on Enterprise Information Systems

832

Organization (customer)

Collect performance

measurments data

Evaluate Placement

Recommendation

Placement

approved?

Determine the goals

of the IT landscape

transformation

Internal gathering of

the requirements

and cost information

More infromation

requested?

yes

Data

requested?

yes

Hybrid-Cloud Placement Optimization Service

Expert(s)

Retrieve previosly

collected information

and results (if exists)

Initial inspection of

performance

measurment data

Analyze and finalize

the required for

placement

information

Configure and

schedule placement

optimization

Sufficient

information?

Interpret and select

the optimization

results

Initiate the assesment

and IT landscape

transformation

(hybrid-cloud

adoption) project

Data Correction

Is data

corrected?

Customer's

input required?

yes

Understand the

existing environment

and information

yes

Import cloud offerings

Information

Prepare cost models

based on the

requirements

Execute Optimization Prepare Results

Data Storage

Save the requirements,

constraints, and cost

information

Measurment data

import

yes

Initialize the

optimization based on

the collected data,

information, and

configuration

Finalization

procedures

Placement recommendation (optionally with additional clarifications)

Request data

Request

placement

advisory

Optimization Automation System

Await available

computational resources

Collected

data

Requirements and

cost information

Request more

information

Notify approval

Placement

(re)assesment

request

Data

approved

Figure 1: BPMN model of the service delivery process.

Towards IT Workload Hybrid-Cloud Placement Advisory in Enterprise

833

ness requirements. These experts are typically not fa-

miliar with state-of-the-art algorithmic solutions, so

the goal of AOS is to provide these as tools for the

experts.

3.4 Process Flow

As depicted in Figure 1, the business process pro-

posed in our process of the IT landscape hybrid-cloud

transformation starts and ends with the consumer of

the service (i.e., the customer organization). The pro-

cess ends with the organization’s stakeholders receiv-

ing and approving a new IT landscape placement rec-

ommendation. If the approval can’t be achieved, part

of the entire described process is restarted, hence in-

troducing a partially iterative aspect.

3.4.1 Assessment of the Existing IT Landscape

The organization initiates the process and determines

the intent and goals for the planned IT landscape

transformation project. After that, the organization

engages HCPOS and hands over the stated goals. The

HCPOS experts process these goals and may assist

the organization in collecting the required measure-

ment data from the running IT landscape.

Suppose the company already has collected histor-

ical data describing the IT landscape workload proﬁle

in sufﬁcient detail and length. In that case, this data

can be used directly and should be sent to the experts.

If such a dataset does not yet exist, the process of col-

lecting measurements is conducted according to the

customer’s requirements as well as the stated goals. It

ranges from multiple weeks to multiple months. The

goal is the construction of a reliable workload pro-

ﬁle for the existing IT landscape. After the collection

period has elapsed, the gathered data is sent to the

HCPOS experts for initial pre-processing.

The experts then import the data into the OAS,

where the collected data is further automatically pro-

cessed, cleansed, and analyzed. However, at this

stage, the OAS can indicate that the data is not suit-

able for further processing according to various rules

and checks (e.g., erratic workload proﬁle, too many

missing values, errors). If that occurs, the expert can

attempt to correct the data or initiate further data col-

lection.

When OAS determines that a sufﬁcient amount

of data with acceptable quality is achieved, the pro-

cessed data is stored in internal data storage for reuse.

Data collection is a time-consuming procedure. Thus,

we assume the data collection is performed only once

within the frame of the same IT landscape transfor-

mation project. The following part of the proposed

process is, however, iterative.

3.4.2 Requirements Collections

In the following activities, the expert can access the

collected data via the frontend provided by OAS. The

presentation of the data must be sufﬁcient for the ex-

perts to efﬁciently understand and analyze it in or-

der to fulﬁll their further activities. These activi-

ties include the collection of the function and non-

functional requirements, as well as constraints, for the

IT landscape and its components subject to the trans-

formation or resizing project. The requirements are

collected by interacting with the stakeholders on the

organization’s side in a format required by the OAS

for automatic processing. Furthermore, for hybrid-

cloud projects, the cost information related to the cus-

tomer facilities (e.g., private data centers) must also

be estimated.

Calculation of running costs for the components

of the IT landscape in the private data center of the

customer is difﬁcult to estimate outside of the organi-

zation itself (Greenberg et al., 2009; Kashef and Alt-

mann, 2012; Altmann and Kashef, 2014; Brogi et al.,

2019) and, therefore, must be assisted by the organi-

zation. The role of the expert in this case is to guide

the customer’s representatives in this information col-

lection process according to their own expertise in the

domain.

This information collection process can be done in

an iterative way according to the expert’s judgment.

After the expert concludes that a sufﬁcient amount of

information is collected, it is imported to the data stor-

age of OAS for future reuse and associated with the

previously collected measurement data.

3.4.3 Automated Processing

In the next activity, the expert hands over control of

the process to the automation procedures and busi-

ness rules encoded into the OAS. Within this activ-

ity, any additional conﬁguration parameters might be

given by the expert (e.g., select a speciﬁc preset). It

is done via the GUI provided by the OAS’s frontend.

When it’s done, a task is scheduled to process the col-

lected measurement data with the collected require-

ments and customer-speciﬁc cost information.

The exact technical implementation of the au-

tomated placement recommendation processing can

vary from one use case to another, however a few

key activities, that are speciﬁcally relevant to our pro-

posed service delivery process are highlighted.

Firstly, since the IT landscape transformation

project involves at least one public cloud provider

as a target for the placement, offerings of these

cloud placements must be acquired for processing and

decision-making. Depending on the provider, it can

ICEIS 2025 - 27th International Conference on Enterprise Information Systems

834

be done either by importing speciﬁc price lists or ac-

cessing provider-speciﬁc APIs. Either way, the goal

of this activity is to acquire sufﬁcient information

about the cloud offerings with technical and pricing

information, such that decision-making is possible by

the algorithm.

The amount of cloud-speciﬁc information and

composition of the offerings can vary depending on

the IT landscape, as well as requirements and con-

straints. For example, if one of the requirements is

that one of the systems in the IT landscape requires a

particularly high availability level, this would require

the retrieval of a set of high-availability offerings, or

composition of such offerings (Salapura and Mahin-

dru, 2016), suitable for the speciﬁc enterprise system.

The decision of which offerings are required accord-

ing to the speciﬁc requirements is derived from the

business rules encoded into the automation processes

of OAS and will differ from use case to use case, from

one enterprise IT system vendor to another.

In the next activity, the OAS must prepare sufﬁ-

cient cost models for the comparison purposes of var-

ious placement compositions of the IT landscape in a

hybrid-cloud environment. Expenses associated with

running speciﬁc components of the IT landscape in

the private data centers of the organization are reused

from the information collected by the expect from the

organization stakeholders. However, the cost estima-

tion for the public cloud can also be daunting because

of the wide variety of pricing models available, even

for the same cloud service (Wu et al., 2019). There-

fore, it is imperative that the encoded business rules

of cost estimation for all cloud providers are use-case-

speciﬁc and validated prior to the rollout of the OAS.

The next activity executed by OAS is actual

decision-making by employing a state-of-the-art algo-

rithm (e.g., metaheuristic, machine learning). The al-

gorithm and its implementation will strongly depend

on the target use case. The central goal of this ac-

tivity is to execute an automation solution that seeks

viable IT landscape placement solutions and selects

the most suitable one according to the encoded busi-

ness rules, technical speciﬁcations, data, customer re-

quirements, and constraints. It’s a typical case of

combinatorial optimization or multi-criteria-decision-

methods applications, depending on the complexity

of the problem, similar to many solutions discussed

in section 2.

The next and ﬁnal activity of OAS within a single

iteration is responsible for the processing of the out-

put generated by the selected algorithm in the preced-

ing activity. This step is critical because, depending

on the type, the optimization algorithm might gen-

erate a large volume of data, which must be trans-

lated into human-readable form. In this case, it has

to be an interpretation of the algorithm output to spe-

ciﬁc placement solutions of the given IT landscape

according to the requirements, constraints, and mea-

sured data. Therefore, it is imperative that the output

contains the exact speciﬁcs of the placement compo-

sitions for the cloud solutions with the pricing infor-

mation.

Furthermore, the prepared results contain not only

a sufﬁcient amount of information but are also pre-

sented in a form that is understandable not only to

the expert but also to the stakeholders in the customer

organization. This typically means providing the ex-

pert with the interactive view in the frontend of the

OAS. Additionally, it is prudent to automate the gen-

eration of the reporting material for dispatching to

the stakeholders in a format that the organizations in

the given domain typically accept. Automation of the

executive-level visualizations, speciﬁcally with a con-

cise explanation of the decision-making process made

by the employed algorithm, can have a positive effect

on the ﬁnal approval stage (Dimara et al., 2022). The

automation of generating such materials can consid-

erably reduce the efforts and time required by the ex-

perts in the following activity focused on preparing

these for dispatching to the customer organization for

approval.

3.4.4 Iteration Finalization

Finally, the prepared results and recommendations for

placement of the IT landscape are sent back to the

organization for approval, optionally supplemented

with additional proposals or clariﬁcations from the

experts. If the results of the process and the placement

recommendation are found to ﬁt the requirements of

the stakeholders, the process of our proposed process

ends. If the approved recommendation is accepted,

the experts are notiﬁed, and the project is ﬁnalized

appropriately. This might include the archival or even

removal of the customer’s data from the OAS to en-

sure the security of the customer’s infrastructure. At

this point, it is assumed that the organization can en-

act the proposed IT transformation plan.

However, if the ﬁnal results are not approved, the

process is returned to the expert, who attempts to ad-

just the requirements, constraints, and optimization

settings while taking into account the feedback from

the previous interaction of the process execution. The

expert can also choose to request further input from

the customer. This additional information from the

company stakeholders is further supplemented to the

data storage for OAS. This iterative process continues

until the ﬁnal reﬁnements produce the results that are

approved by the customer organization.

Towards IT Workload Hybrid-Cloud Placement Advisory in Enterprise

835

4 PROOF OF CONCEPT

We empirically test the overall feasibility of the pro-

posed process in a real-world environment by pro-

viding advisory support to stakeholders in the organi-

zations operating standard IT enterprise applications.

The goal is to verify the proposed process’s overall

usefulness and applicability in the real-world enter-

prise environment. The ﬁeld testing was based on

production SAP enterprise landscapes with the goal

of assessing the viability of the hybrid-cloud transfor-

mation of the existing IT landscape.

Selection of the placement conﬁguration in a

hybrid-cloud for SAP landscapes can be a complex

task and subject to a variety of considerations and

constraints (Berhorst et al., 2021). In our ﬁeld test,

we focused on the Infrastructure-as-a-Service (IaaS)

(Badger et al., 2012) options of the public cloud

providers as placement options, as well as private (on-

premise) data centers of the IT landscape owner. For

the cloud placement options in the hybrid environ-

ment in our ﬁeld test we relied on Azure (B

ogelsack

et al., 2022), their SAP-certiﬁed VMs, required net-

working and storage infrastructure, and APIs to auto-

mate offering selection and costs estimation.

It is shown in previously published work that this

task can be solved by using metaheuristic algorithms

(Kharitonov et al., 2023), which also served as a ba-

sis for the ﬁeld testing discussed in this work. How-

ever, in principle, any suitable optimization mecha-

nism can be applied. Therefore, the speciﬁcs of op-

timization are beyond the scope of this work, as we

concentrate on how these can be actually used in a

business environment. For the purposes of this proof

of concept, the aforementioned metaheuristic-based

approach was wrapped into a specially developed and

hosted containerized prototype solution, consisting of

frontend (web-UI) and backend (relational database,

queues, various workers), as discussed in section 3.2.

The performance metrics data collection from a

real-world SAP IT landscape was carried out using a

software solution that integrated with the SAP sys-

tems via the standard APIs and did not impact the

overall performance of the system. The data col-

lected contains the performance counters collected as

time series over a number of weeks. The perfor-

mance counters include both the utilization and ca-

pacity standard metrics (e.g., CPU, RAM, network,

storage), as well as the SAP-speciﬁc metrics (e.g.,

number of transactions, SAPS(Marquard and G

otz,

2008)).

The goal is to take advantage of the data and pre-

vent overprovisioning by selecting the placement op-

tions according to the real workload proﬁle. Choos-

ing the right size of the target cloud environment can

signiﬁcantly reduce future operational costs (Aloysius

et al., 2023) of an SAP system.

The collection of the requirements and constraints

was carried out by a team of domain experts with

close cooperation with the owners (organization) of

the SAP IT landscape. The collection and processing

of the requirements are supported by the frontend of

our prototype OAS, thus ensuring the correctness and

consistency of the collected values.

4.1 Feedback

At the ﬁnalization phase of the ﬁeld test, we have col-

lected feedback from the participating stakeholders.

It is important to note that in this work, we focus on

the feedback related to the overall business process

validity and usefulness, while the results of the place-

ment optimization itself are consistent with the work

we relied upon to develop the proof of concept proto-

type (Kharitonov et al., 2023).

Two real-world companies participated in the ﬁeld

test. Company A is a manufacturing business that

maintains its own SAP IT landscape for internal use.

Company B is a service provider with a higher degree

of internal competence in the ﬁeld of public cloud.

The management-level stakeholder, in the role of

the head of the SAP infrastructure department in com-

pany A, declared that the primary objective of the

proof of concept is to acquire fact-based executive-

level answers about cloud adoption within the given

business realities. The aforementioned stakeholder of

company A remained satisﬁed with the assessment re-

sults achieved in the proposed business process sec-

tion 3.

The technical-level personnel in the same de-

partment demonstrated initial skepticism regarding

the application of the proposed process and voiced

doubts it would provide useful insights beyond what

is known already. However, the ﬁnal feedback was

positive and concentrated on the number of relevant

insights presented as a result, underlying the useful-

ness of the proposed formalized process for integrat-

ing state-of-the-art algorithmic approaches to real-

world enterprise processes. Speciﬁcally, it was noted

that the presented cost-driving factors are hard to cal-

culate just within the day-to-day department activity.

Within Company B, two groups of stakeholders

can be viewed according to their expectations. First

are the stakeholders in the ﬁeld of strategy and tech-

nological transformation, including strategic manage-

ment in the ﬁelds of data centers, cloud, and infras-

tructure. These initially expressed an interest in the

validation of their own efﬁciency, costs, and innova-

ICEIS 2025 - 27th International Conference on Enterprise Information Systems

836

tion advantages in the ﬁeld of cloud adoption.

Second is the technology-oriented executive and a

technology expert. They expressed an interest in us-

ing the tool within our proposed process to acquire ad-

ditional insights and expertise. This expertise would

allow them to decide on offering cloud solutions to

their own customers if such makes sense in the partic-

ular business environment.

As the feedback from the stakeholders of com-

pany B, it was noted that the use of an automated

solution within the proposed process provides valu-

able insights and information for the management-

level discussions. It was noted that the ﬁnal approved

results of the executed process are detailed and cor-

respond to the expectations of the stakeholders. The

transparency of the analysis done using the proposed

process was also noted positively. The interaction

between the stakeholders and the placement experts

within the meetings is also positively noted as an op-

portunity for the experts to provide detailed explana-

tions for the results and the decision process provided

by AOS.

The last point is speciﬁcally notable as a positive

aspect of relying on the proposed process instead of

simply providing the AOS to the stakeholders as a

self-service. The value of the data and information

provided by the owners of the IT landscape can be

truly put to use by experts in the ﬁeld of IT work-

load placement. As we initially hypothesized, this

experience is complemented by the speciﬁc business-

domain knowledge of the stakeholders within a con-

trolled process of gathering, processing, storing the

relevant data and information, and ﬁnally providing a

relevant, useful result.

In the summary of the feedback, it can be said that

an application of an automation tool for the assess-

ment of real-world IT landscapes and evaluation of

the cloud adoption possibilities is useful when done

in a systematic manner. The systematic collection of

the functional and non-functional requirements and

the use of data-driven automation can result in pro-

viding the stakeholders with the expected information

that can further serve as a basis for management-level

decision-making.

Delivery of the information relevant for the spe-

ciﬁc stakeholders, instead of simply relying on the

ability of the automation solution to deliver some

data, is the main advantage of relying on a well-

deﬁned expert and automation-supported process of

collecting and reﬁning the requirements, as well as

evaluating the results. This is the main advantage of

relying on a systematic process for collecting relevant

information from the customer as part of the consul-

tancy service, instead of simply relying on the data

and the automation as a self-service.

4.2 Lessons Learned

During the ﬁeld testing of the proposed process, a

few key takeaways can be formulated based on the

feedback received from the owners of the IT land-

scape (decision-makers) and the experts. The take-

aways concern both the functionality of the prototype

OAS, as well as the formulation and presentation of

the placement recommendation results.

4.2.1 Intelligent Results Presentation

Within our ﬁeld testing, we relied on evolutionary

combinatorial optimization to achieve the optimal

recommendation within the OAS. This type of opti-

mization evaluates a vast number of possible place-

ment combinations for the IT landscape and, in the

end, presents the best-selected combinations accord-

ing to the given requirements. While it might be

tempting to simply present only the best solution to

the customer, this turned out to be an insufﬁcient

approach as the decision-makers displayed interest

in comparisons against suboptimal solutions to bet-

ter understand the inﬂuence of their requirements and

constraints on the recommendation.

However, providing too many such options for ex-

amination via the visualization and data generated by

OAS was noted to be overwhelming for the decision-

makers. At the same time, a simple application of the

so-called Hall of Fame (HoF) strategy, where only a

certain number of the top best solutions are selected,

is displayed, resulting in the critique of the lack of

sufﬁcient diversity of the solutions. The solutions that

approach the optimum tend to have only slight differ-

ences between them.

Therefore, it was noted that a more intelligent ap-

proach to selecting the ﬁnal recommendations and

comparison solutions is needed. In our speciﬁc ﬁeld

test, such an approach, that was deemed successful by

the decision-makers, was to select the best possible

solution, and, as a comparison, data points provide

the evaluated solutions that place the entire IT land-

scape within the same location or provider. Such pre-

sentation provides an opportunity to clearly demon-

strate various trade-offs (e.g., degree of requirement

satisfaction, costs, constraints, etc.) between simpler

strategies, where the entire landscape is concentrated

within the same location, and more complex hybrid-

placement solutions. Further sorting and ﬁltering ca-

pabilities for navigating discovered solutions are also

deemed beneﬁcial for presentation and exploration,

including for future requirements or constraints tun-

ing.

Towards IT Workload Hybrid-Cloud Placement Advisory in Enterprise

837

4.2.2 Complexity of Requirements Processing

Within our ﬁeld test, we have concentrated on the

speciﬁc use case of SAP systems placement within a

hybrid-cloud environment. While there are a number

of well-understood best practices and speciﬁc rules

presented by the SAP developers, direct conversion

of the business-speciﬁc requirements into technical

details of cloud deployment architectures is far from

trivial and requires a sufﬁcient amount of initial in-

vestment from the experts to construct such rules for

translating requirements and constraints into the se-

lection of speciﬁc combinations of cloud products.

While this initial investment pays off in reducing

the workload of the experts working with OAS, the

participating domain experts noted that a degree of

ﬂexibility is required in the encoded rules. The reason

is that public cloud offerings do not stay static. New

cloud products are introduced, while older ones might

get deprecated. Therefore, changes in the placement

recommendation selection rules might require fre-

quent updates.

4.2.3 Customer-Speciﬁc Pricing

When selecting a public cloud location for place-

ment of the IT landscape, relying solely on the list

prices for all cloud products might not be sufﬁcient.

Speciﬁcally, corporate customers of the public cloud

providers can negotiate discounts on certain cloud of-

ferings (Deb and Choudhury, 2021).

However, details of such discounted contracts

with the public cloud providers must be provided by

the organization requiring the placement optimization

recommendation and shared with the experts. This

data can be sensitive and must be guaranteed to be

protected from disclosure. Access to this information

should be strictly controlled, even among the experts

operating OAS. There must also be a mechanism to

guarantee the removal of this information from OAS

after the recommendation process is ﬁnished.

Furthermore, encoding the details of such dis-

counted pricing can be tedious and require care. Any

errors will result in incorrect price models and, conse-

quently, placement recommendations. Therefore, this

stage directly beneﬁts from a specially made frontend

GUI supporting the experts in this task.

The complexity of the customer-speciﬁc cost es-

timation extends to the on-premise costs as well.

Within our ﬁeld tests, we discovered that an accurate

estimation of the private data center’s running costs

was not possible within the timeframe of the evalu-

ation ﬁeld test in one of the cases mentioned earlier.

This led to potentially inaccurate placement recom-

mendations generated by AOS. In such cases, the re-

sults have to be carefully processed by the placement

experts before ﬁnally being sent to the stakeholders.

The complexity of gathering customer-speciﬁc costs

and estimating them should not be underestimated.

5 FUTURE WORK

Relying on SAP-based solutions in our ﬁeld test al-

lowed us to beneﬁt from a signiﬁcant degree of stan-

dardization and support from cloud providers (e.g.,

Azure (B

ogelsack et al., 2022)), as well as an abun-

dance of domain experts, who are vital for our pro-

posed approach. That is typical for large enterprise

solutions. However, that brings to question the gener-

alization of such an approach. We intend to validate

this approach with a similar solution aimed at smaller

organizations (e.g., Odoo (Wu and Chen, 2020)).

Furthermore, we intend to investigate the feasibil-

ity and how beneﬁcial it is to apply the proposed ap-

proach with non-IaaS placement models (e.g., RISE

with SAP (Subrahmanyam, 2022)). This investigation

might prompt a further adaptation of the proposed for-

malized process.

6 CONCLUSION

In this work, we propose a formalization based on

BPMN of a systematic process for applying automa-

tion solutions to decision-making advisory in hybrid-

cloud placement selection for enterprise IT workloads

in a data-driven manner. The proposed process is

business-centric and includes a systematic collection

of the data, as well as the required information needed

for the decision-making of the IT workload placement

by experts collaborating with stakeholders. We aim to

demonstrate how algorithmic-based automation can

ﬁt into real-world business processes.

The preliminary feasibility testing, in the form of

an overall proof-of-concept evaluation, of the pro-

posed approach is performed within controlled, real-

world enterprise environments and involves stake-

holders from two distinct companies. It is done as a

ﬁeld test with a speciﬁc use case of SAP enterprise ap-

plications placement in a hybrid-cloud infrastructure.

For this purpose, an automated solution for placement

optimization was used within the constraints of the

proposed process. The evaluation results include the

feedback received from the stakeholders. The feed-

back conﬁrms the validity of the proposed approach

for an enterprise environment, where trained experts

rely on the automated solution and communicate with

the stakeholders who own the speciﬁc IT landscape.

ICEIS 2025 - 27th International Conference on Enterprise Information Systems

838

Several important takeaways from the ﬁeld test are

noted. These include the importance of presenting the

results generated by the automation solution, which

should be done without overwhelming the stakehold-

ers. We also note the complexity of the initial formal-

ization of use-case-speciﬁc requirements, such that

these requirements could be supplied to the automa-

tion system. Additionally we note the particular com-

plexity of estimating and accounting for the costs that

are speciﬁc to the speciﬁc companies (e.g., cloud dis-

count contracts, running costs of the private data cen-

ter placements). This complexity can be detrimental

to the validity of the ﬁnal results. Finally, we note

the importance of the appropriate visualization used

in the ﬁnal results presentations.

REFERENCES

Aloysius, G. C., Bhatia, A., and Tiwari, K. (2023). Stabil-

ity and Availability Optimization of Distributed ERP

Systems During Cloud Migration. Lecture Notes in

Networks and Systems, 654 LNNS:343–354.

Altmann, J. and Kashef, M. M. (2014). Cost model based

service placement in federated hybrid clouds. Future

Generation Computer Systems, 41:79–90.

Badger, M., Grance, T., Patt-Corner, R., and Voas, J. (2012).

Cloud Computing Synopsis and Recommendations.

Technical Report NIST Special Publication (SP) 800-

146, National Institute of Standards and Technology.

Berhorst, R., D

urk, M.-O., Hauzeneder, C., Sperzel, S.,

Wagner, P., and Zimmermann, R. (2021). DSAG-

Leitfaden: Hybrid Operations. Technical report,

Deutschsprachige SAP Anwendergruppe e.V.

ogelsack, A., Chakraborty, U., Kumar, D., Rank, J., Tis-

chbierek, J., and Wolz, E. (2022). SAP S/4HANA

on Microsoft Azure – Concepts and Architecture. In

ogelsack, A., Chakraborty, U., Kumar, D., Rank, J.,

Tischbierek, J., and Wolz, E., editors, SAP S/4HANA

Systems in Hyperscaler Clouds: Deploying SAP

S/4HANA in AWS, Google Cloud, and Azure, pages

267–323. Apress, Berkeley, CA.

Brogi, A., Corradini, A., and Soldani, J. (2019). Estimat-

ing costs of multi-component enterprise applications.

Formal Aspects of Computing, 31(4):421–451.

Deb, M. and Choudhury, A. (2021). Hybrid Cloud: A New

Paradigm in Cloud Computing. In Machine Learning

Techniques and Analytics for Cloud Security, chap-

ter 1, pages 1–23. John Wiley & Sons, Ltd.

Dimara, E., Zhang, H., Tory, M., and Franconeri, S. (2022).

The Unmet Data Visualization Needs of Decision

Makers Within Organizations. IEEE Transactions on

Visualization and Computer Graphics, 28(12):4101–

4112.

Greenberg, A., Hamilton, J., Maltz, D. A., and Patel, P.

(2009). The cost of a cloud: Research problems in

data center networks. SIGCOMM Comput. Commun.

Rev., 39(1):68–73.

Jung, G., Mukherjee, T., Kunde, S., Kim, H., Sharma,

N., and Goetz, F. (2013). CloudAdvisor: A

recommendation-as-a-service platform for cloud con-

ﬁguration and pricing. Proceedings - 2013 IEEE 9th

World Congress on Services, SERVICES 2013, pages

456–463.

Kashef, M. M. and Altmann, J. (2012). A Cost Model for

Hybrid Clouds. In Vanmechelen, K., Altmann, J., and

Rana, O. F., editors, Economics of Grids, Clouds, Sys-

tems, and Services, pages 46–60, Berlin, Heidelberg.

Springer.

Kharitonov, A., Nahhas, A., M

uller, H., and Turowski, K.

(2023). Data driven meta-heuristic-assisted approach

for placement of standard IT enterprise systems in

hybrid-cloud. In Proceedings of the 13th International

Conference on Cloud Computing and Services Science

- CLOSER, pages 139–146. SciTePress / INSTICC.

Marquard, U. and G

otz, C. (2008). SAP Standard Appli-

cation Benchmarks - IT Benchmarks with a Business

Focus. In Kounev, S., Gorton, I., and Sachs, K., ed-

itors, Performance Evaluation: Metrics, Models and

Benchmarks, pages 4–8, Berlin, Heidelberg. Springer.

Menzel, M., Sch

onherr, M., and Tai, S. (2013). (MC2)2:

Criteria, requirements and a software prototype for

Cloud infrastructure decisions. Software - Practice

and Experience, 43(11):1283–1297.

Mettouris, C., Vanezi, E., Zampas, G., Argyrou, S.,

Achilleos, A., Constantinides, A., and Papadopoulos,

G. A. (2022). CloudRecoMan: Cloud adoption made

easy. ACM International Conference Proceeding Se-

ries, pages 110–117.

OMG (2010). Business process model and notation

(BPMN) version 2.0. Standard, Object Management

Group (OMG).

Sahu, P., Roy, S., and Gharote, M. (2024). CloudAdvisor

for Sustainable and Data Residency Compliant Data

Placement in Multi-Cloud. In 2024 16th International

Conference on COMmunication Systems & NETworkS

(COMSNETS), pages 285–287.

Salapura, V. and Mahindru, R. (2016). Enabling enterprise-

level workloads in the enterprise-class cloud. IBM

Journal of Research and Development, 60(2-3):3:1–

3:8.

Subrahmanyam, R. S. (2022). An Overview of SAP

S/4HANA, pages 1–23. Apress, Berkeley, CA.

Venkatraman, A. and Arend, C. (2022). A Resilient, Ef-

ﬁcient, and Adaptive Hybrid Cloud Fit for a Dy-

namic Digital Business. International Data Corpora-

tion (IDC). Report No.: IDC #EUR149741222.

Wu, C., Buyya, R., and Ramamohanarao, K. (2019).

Cloud Pricing Models: Taxonomy, Survey, and In-

terdisciplinary Challenges. ACM Comput. Surv.,

52(6):108:1–108:36.

Wu, J. Y. and Chen, L. T. (2020). Odoo erp with busi-

ness intelligence tool for a small-medium enterprise:

A scenario case study. In Proceedings of the 2020

11th International Conference on E-Education, E-

Business, E-Management, and E-Learning, IC4E ’20,

page 323–327, New York, NY, USA. Association for

Computing Machinery.

Towards IT Workload Hybrid-Cloud Placement Advisory in Enterprise

839