Modeling and Simulation of Concurrent Workload Processing in Cloud ...

Modeling and Simulation of Concurrent Workload Processing in Cloud-Distributed Enterprise Information Systems Alexandru-Florian Antonescu12 , Torsten Braun2 [email protected], [email protected] SAP (Switzerland) Inc. Products & Innovation, Research Althardstrasse 80, 8105 Regensdorf, Switzerland 1

ABSTRACT Cloud Computing enables provisioning and distribution of highly scalable services in a reliable, on-demand and sustainable manner. Distributed Enterprise Information Systems (dEIS) are a class of applications with important economic value and with strong requirements in terms of performance and reliability. In order to validate dEIS architectures, stability, scaling and SLA compliance, large testing deployments are necessary, adding complexity to the design and testing of such systems. To fill this gap, we present and validate a methodology for modeling and simulating such complex distributed systems using the CloudSim cloud computing simulator, based on measurement data from an actual distributed system. We present an approach for creating a performance-based model of a distributed cloud application using recorded service performance traces. We then show how to integrate the created model into CloudSim. We validate the CloudSim simulation model by comparing performance traces gathered during distributed concurrent experiments with simulation results using di↵erent VM configurations. We demonstrate the usefulness of using a cloud simulator for modeling properties of real cloud-distributed applications.

Categories and Subject Descriptors C.2.4 [Distributed Systems]: Distributed applications

Keywords Cloud Computing; Distributed Applications; Performance Profiling; Modelling and Simulation

1.

INTRODUCTION

Cloud Computing [9] enables provisioning and distribution of highly scalable services in a reliable, on-demand and Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from [email protected]. DCC’14 August 18 2014, Chicago, IL, USA Copyright 2014 ACM 978-1-4503-2992-7/14/08 ...$15.00. http://dx.doi.org/10.1145/2627566.2627575.

University of Bern Communication and Distributed Systems (CDS) Neubrückstrasse 10, 3012 Bern, Switzerland 2

sustainable manner. Often, the applications running in these distributed environments are designed to concurrently execute workload, making use of available multi-core computing resources. As example, distributed Enterprise Information Systems [11] often interact with large-scale distributed databases, requiring distributed processing of results coming from multiple systems. In order to validate certain properties of distributed cloud systems, it is often the case that large-scale hardware testbeds are used for observing the application behavior under conditions similar to those met in production environments. However, getting access to such large, distributed testbeds can be difficult, and reproducing results is often not possible, due to the use of shared physical computing and network resources. This leads to the need of modeling cloud applications and resources in order to perform simulations for testing ”cloud-aware” software management algorithms. Examples of such complex interactions to be modeled and simulated are cloud infrastructure management [4], or Service Level Agreement-driven benchmarking of cloud services scaling[2]. We extend the work in [2] and [4] by building a simulation model for a distributed enterprise information system (dEIS) based on performance measurements gathered in a small-scale distributed testbed. We show how to analyze and model the distributed interactions between the dEIS entities in a state-of-the-art cloud simulator. We then perform multiple distributed experiments with varying concurrent workload. Finally, we compare the experimental results to the simulation, by running the same workload scenario. Our main contributions can be summarized as follows. We present an approach for analyzing and structuring monitoring information belonging to distributed cloud services in order to build a model of the distributed application. We show how to integrate the application’s performance model in a cloud simulator, by converting from time-bound service performance indicators to CPU and network resource utilization. We describe an approach for increasing the scale of cloud simulation and for testing the distributed application’s behavior under di↵erent concurrent workloads. The rest of our paper is organized as follows. Section 2 presents the related work in the field of cloud simulators as well as modeling and simulation of distributed cloud applications. Section 3 presents the algorithms used for building the simulation models and for modeling a distributed application in CloudSim. Section 4 presents the evaluation

0 10:26

10:28

10:30

10:32

10:34

10:36

milliseconds

100

(2) Roundtrip to send request to WK

60 50 40 30 20

milliseconds

(1) Roundtrip to create LB session 200

15

10:26

10:28

10:30

10:32

10:34

10:36

Schedule Request

Get datasource Schedule Request Execution Obtain Database Connection Execute Query Extract Result Send Result Schedule Result Processing Format Results Send Results Schedule Result Processing Validate Response

Bytes

10:28

10:30

10:32

10:34

10:36

(5) VM CPU Load 60 40 20 10:26

Close Session Send ACK

(3) Scheduling of Resp. Processing 150 100 50 0 10:26

KBytes

Submit Request Send ACK

10:28

50000 30000 10000 10:26

10:30

10:32

10:34

10:28

10:30

10:32

10:34

(4) Response Processing

10 5 0 10:26

715000 712500 710000 707500

10:36

10:36

10:28

10:30

10:32

10:34

10:36

(6) Memory Usage

10:26 10:28 10:30 10:32 10:34 10:36

(7) Network bytes read Bytes

Send ACK

ST

milliseconds

WK

milliseconds

LB Create Session Select available WK service Create new session Return Session Submit Request

Percentage

CS

8000

(8) Network bytes written

4500 1000 10:26

10:28

10:30

Time

10:32

10:34

10:36

Time

Figure 1: EIS Services Interactions results, and finally, Section 5 draws conclusions and gives future research directions.

2.

CLOUD SIMULATOR

CloudSim [8] positions itself as a cloud simulator for both applications and infrastructure, as it allows modeling of hardware and software cloud resources. Among the modeled physical entities there are hosts, network links and datacenters, while the modeled software entities are virtual machines (VMs), brokers and cloudlets (tasks). The mentioned entities are manipulated using a Java API. The simulator is implemented using discrete event communication. CloudSim provides a wide selection of resource allocation policies for VM-to-host and task-to-VM mapping. Buyya et al. [5] presented an approach for simulating large scale cloud environments by using CloudSim. While they described the steps required for simulating a large number of hosts and network connections, they did not show how to also model applications using the CloudSim simulator. Our work fills this gap by presenting a methodology for simulating a large-scale distributed application. Garg et al. [7] extend CloudSim with more advanced network simulation capabilities. They model the datacenter network landscape by introducing switch and network packet simulation entities. The authors simulate a network application using four VMs and compare the results to the simulator, in a similar manner to our approach.

3.

PERFORMANCE ANALYSIS AND MODELING OF A DISTRIBUTED SYSTEM

As representative dEIS distributed application we used the one described in [1, 2, 3]. The targeted dEIS system is composed of four core services: one or more Thin Clients (CS), a Load Balancer (LB), one or more Worker services (WK), and one or more Database Storage services (ST). Each service runs in its own VM and communicates asynchronously with the other services using a distributed service messaging bus. Fig. 1 presents the interactions between various dEIS services when executing a CS-initiated request. This also shows the breakdown of operations executed at each service, which serves as the foundation for building a simulation model of dEIS.

Figure 2: dEIS Consumer Service Performance Indicators In order to simulate the dEIS application, we first recorded key performance indicators of dEIS services under constant workload. We then used these monitoring traces to create the dEIS’s simulation model composed of a large number of single requests’ execution measurements. For a given request we measured the round-trip times for the service calls displayed in Fig. 1, as well as the durations of the servicelocal calls. For building the application performance profile we applied the following algorithm. We instantiated one VM for each (Consumer) CS, (Load Balancer) LB, (Worker) WK and (Storage) ST services. We initialized the concurrent load with 1 and the benchmark time duration with 10 minutes. Every 10 milliseconds the number of active requests ar executed by the system was compared to the target load, and if ar was below load, then a number of load ar CS requests would be generated. For each end-to-end request, the following parameters were recorded: round trip durations, service-local operation durations and VM-level performance metrics (CPU, memory, network, disk). By recording all these parameters for each distributed request, a performance benchmark profile was generated under the constant workload load. After running the benchmark described above, the dEIS performance profiles for constant concurrent load between 1 and max.load will be known and the performance profile will be generated with all service call round-trip times and service-local operation durations. This profile will then be used during the simulation, as described later in this section. In order to build the application’s performance profile the following assumptions were made: (1) load balancing is applied at the LB service and (2) CS requests are independently executed by the dEIS services. This allows to extend the performance profile recorded with one VM per dEIS service to any number of VMs, as the performance of individual services will be influenced only by the level of concurrent workload and not by the number of running VMs.

3.1

dEIS Performance Profiling

We analyze performance profiles generated for each dEIS service while the system was sequentially executing requests

(4) memory usage

30 20 10:28

10:30

10:32

10:34

Bytes

Bytes

10:30

10:32

10:34

(3) Scheduling of Resp. Processing

5 10:28

10:30

10:32

10:34

milliseconds

(2) Roundtrip to send request to ST

10:26

10:28

10:30

10:32

10:34

(4) Response Processing

10:26

10:28

10:30

10:32

10:34

10:30

10:32

10:34

Bytes 10:34

10:36

Time

10:32

10:34

10:36

(4) Netw. bytes written send resp. to CS

30000 20000 10000 10:26

10:28

10:30

10:32

10:34

10:36

(6) Memory Usage

596000 594000 592000

10:36

10:28 10:30 10:32 10:34 10:36

Time

10:28

10:30

10:32

10:34

10:36

(3) Database query execution time 200 100 10:28

10:30

10:32

10:34

30000 20000 10000 0 10:26

10:28

10:30

10:32

Time

10:34

10:36

(2) Obtain Database Connection 12 8 4 10:26

10:36

(5) Database query result size

(5) Roundtrip to send response to CS

10:28

10:32

10:30

40000

10:36

(1) Scheduling of Request Processing

10.0 7.5 5.0 2.5 0.0 10:26

10:26

10:36

30 27 24 21 10:26

10:30

10:28

Figure 5: dEIS Worker Service - VM Performance Indicators

10:36

20 15 10 5

10:36

10:34

1000

Time

50 40 30 20 10

10:36

10

0 10:26

10:28

bytes

milliseconds milliseconds

10:28

milliseconds

milliseconds

40

10:32

70 60 50 40

Time

60

10:30

2000 0 10:26

(5) VM CPU Load

Figure 3: dEIS Load Balancer Service Performance Indicators

10:26

10:28

3000

10:36

1000 10:26

4000 2000 0 10:26 10:28 10:30 10:32 10:34 10:36

Time

(1) Request Scheduling Duration

10:34

(3) Netw. bytes read send resp. to CS

(6) network write

80

10:32

2000

10:26 10:28 10:30 10:32 10:34 10:36

(5) network read 4000 3000 2000 1000 0 10:26 10:28 10:30 10:32 10:34 10:36

10:30

3000

430000 429000 428000 427000

10:36

10:28

KBytes

10:26

10:26

Bytes

(3) CPU utilization 40

1000

milliseconds

10:36

(2) Netw. bytes written send req. to ST

4000

milliseconds

10:34

1500

milliseconds

10:32

Bytes

10:30

milliseconds

10:28

KBytes

Percentage

0 10:26

(1) Netw. bytes read send req. to ST Bytes

1

(2) concurrent sessions

2.0 1.5 1.0 0.5 0.0 10:26 10:28 10:30 10:32 10:34 10:36

milliseconds

count

2

milliseconds

(1) session−create

3

10:28

10:30

10:32

10:34

10:36

(4) Process SQL Resultset 9 6 3 0 10:26

10:28

10:30

10:32

10:34

10:36

(6) Roundtrip send response to WK 60 50 40 30 20 10:26

10:28

10:30

10:32

10:34

10:36

Figure 4: dEIS Worker Service - Application Performance Indicators

Figure 6: dEIS Storage Service - Application Performance Indicators (Service Related)

for 10 minutes, at a fixed concurrent workload. The physical setup consisted of three servers with Intel(R) Core(TM)2 Duo CPU E8400 CPUs, 4GB of memory and 1 Gbps network interface. CS and LB VMs were co-located on the same server, while WK and ST were each located on a separate server. A fourth server was used for running validation experiments, by using two additional VMs, one VM for the WK service and one VM for the ST service. In Fig. 2 we show the performance profile of the dEIS CS service (thin client). Four application metrics were recorded for the CS service: (1) round-trip time for creating a LB session and receiving a confirmation acknowledgment, (2) round-trip time for sending a request to the WK service, (3) the time during which the response was queued before being selected for processing, and (4) time for validating the response. For the VM performance there were also four metrics gathered: (5) VM CPU utilization percentage, (6) memory usage, (7) network bytes read, and (8) network bytes written. The network bytes read correspond to the messages received from LB (session creation confirmation, around 4KB) and WK (processing result, 10-30KB), and have a large variance. The CPU utilization has low values and is almost constant at around 20%, as expected for a thin client application. In Fig. 3 we present the performance profile of the dEIS

LB service. The application metrics recorded are: (1) session creation time, and (2) number of concurrent sessions. The LB VM metrics are the same as the ones gathered for CS. CPU utilization is also low at around 20%, justifying co-location of CS and LB VMs. Both network read and written bytes have a median of around 2KB and are caused by exchanging small sized monitoring and session creation/ confirmation messages. In Fig. 4 we present the application performance profile of the dEIS WK service. We gather the following monitoring information: (1) request scheduling duration - time a CS-request is queued until it is selected for processing, (2) round-trip duration of sending the WK-request to ST service, (3) ST-response queuing, (4) processing of ST-response, and (5) round-trip time for sending the response back to CS service and receiving the acknowledgment. In Fig. 5 we display the WK’s VM performance metrics. CPU utilization has high values because of the additional computing time required for processing the result from ST. Also, the size of the response sent to CS has a large variance (5-30KB), explaining the variance of the result processing time. Also, this rather large result size variance will influence the variance of the simulation results. In Fig. 6 we present the ST’s application performance metrics. The recorded metrics are: (1) time a WK-request is

30000

2500

20000 10000

2000

10:26

10:28

10:30

10:32

10:34

0 10:26

10:36

● ● ●●●● ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ●● ●●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ●● ●●●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ●● ●●● ●●● ● ● ● ● ●● ● ● ●● ● ●●●● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ●● ● ● ●● ● ●● ● ● ●● ● ● ● ●● ● ●● ●●● ● ● ● ● ●● ● ●● ● ● ●● ● ●● ● ● ● ●● ● ●● ● ● ● ●●● ● ● ● ●● ● ●●●● ●● ●● ● ● ●● ●●● ●● ● ● ● ● ● ●● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ●● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ●● ● ●● ●● ● ● ● ●●● ●● ● ●● ● ●● ● ●●●● ● ●● ● ● ●● ● ● ● ● ●● ● ● ● ● ●● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ●● ● ● ●●●● ● ●● ●●● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ●● ● ● ● ●● ●● ● ● ● ●● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ●●● ●●●● ● ● ● ● ● ●● ● ●● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ●● ● ●● ● ●● ●● ●● ●● ● ● ● ●● ● ● ● ● ● ●● ●● ● ● ●●● ● ●● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ●● ●● ● ● ● ●● ● ●● ● ●● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ●● ● ● ● ● ●● ●● ● ●● ●● ● ● ● ●● ● ● ●● ●● ● ●●● ● ● ●●●● ●● ● ● ●●●●● ● ●● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ●●●●● ● ●● ●● ● ● ●● ● ● ● ●● ●●●● ● ●● ●● ● ●● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ●●● ●●● ● ● ● ● ● ●● ● ●● ● ●● ●● ● ● ● ●● ● ● ● ● ●● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ●●● ●● ● ● ● ● ● ●● ● ●● ● ● ●● ● ●● ●● ● ● ●●● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ●● ● ● ●● ● ● ●● ●● ● ●● ● ●● ● ●● ●● ● ● ● ● ● ● ● ●● ● ●● ● ●● ● ●● ● ● ● ● ● ● ● ● ●● ●● ● ● ●● ● ● ● ● ●●● ● ●● ● ● ● ● ● ●●● ● ●● ● ● ●● ●● ● ● ●● ● ●● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ●●●● ● ● ● ● ● ●●● ● ● ● ● ●●● ● ● ● ● ● ●● ● ● ● ● ●●● ● ●●● ●● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●●●●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ●● ● ● ● ● ●● ●● ● ●● ● ● ●● ● ● ●●● ●● ●● ● ● ● ● ●● ● ● ●●● ●● ● ●●● ●● ●● ●●● ●● ● ● ● ● ● ●●● ● ● ●● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ●● ●● ●●● ● ● ●● ●● ● ● ● ●● ● ● ●●● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ●● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ●●●● ●● ● ● ●●● ●● ●● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●●● ● ● ● ● ● ●●● ●● ● ●● ●● ●● ● ● ●● ● ●● ● ●● ●●● ● ●●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ●● ●● ●● ● ● ● ●● ●●● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ●●● ● ●●● ● ● ●● ●●●● ●● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ●●● ● ● ● ●● ● ● ● ●●● ● ● ● ●● ● ● ● ● ● ●●●● ●● ● ●●●● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ●● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●●● ● ● ● ●● ● ● ●● ●● ●● ● ●● ● ● ● ● ●● ● ● ● ●● ● ●● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●●●● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ●●● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●●●● ●●●● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ●● ●●● ● ● ●● ● ● ● ●● ● ●●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●● ● ●● ●● ● ●● ● ● ●● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ●● ●● ●●● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ●● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ●●● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●●● ●● ● ● ● ● ● ● ● ●●● ● ●● ● ● ● ● ● ● ● ● ●●● ●● ● ● ● ●● ● ● ● ●●●● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ●● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ●●● ●● ●● ● ● ● ●●● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ●●● ● ●● ●● ● ● ● ● ●● ●● ●●● ●● ● ●● ● ●●● ● ● ● ●● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ●● ●● ● ●● ● ● ●●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ●● ● ●● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ●● ● ●● ● ● ● ● ● ● ● ● ●●●●● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ●●●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●●● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●●● ● ● ●● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●●● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ●●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ●● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ●● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ●●● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ●● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Distribution of input values

●

milliseconds

Bytes

Bytes

Simulation input values

(2) Net. bytes written send resp. to WK

10:28

10:30

10:32

10:34

10:36

1000

500

0

0

(3) VM CPU Load

2500

5000

0.004

density

(1) Net. bytes read send resp. to WK 3000

0.002

0.000

7500

0

1008000

25 1006000 0 10:30

10:32

10:34

10:36

● ●● ● ● ●● ● ● ● ● ●● ●●● ● ● ● ● ● ● ● ●● ● ●● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ●● ●●●●● ● ● ●● ● ● ●● ● ● ●● ● ●●● ● ●●● ● ● ● ● ●● ● ● ●●●●●● ●● ● ● ● ● ● ●● ● ●● ●●●●● ● ●● ● ● ● ● ●● ● ● ● ● ●● ● ● ●● ● ● ● ●●● ● ● ●● ●● ● ●● ● ●●● ●● ● ●● ● ● ● ● ● ●● ●● ●●● ●● ● ● ● ● ● ● ● ●● ● ● ●● ●●●● ● ●●● ● ● ●●●● ●● ● ● ● ●● ● ●●●● ● ● ● ●● ● ●●● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●● ●●● ● ● ● ● ●● ●● ● ●●● ● ●●●●● ● ●● ● ● ●● ● ● ●●●● ● ● ●● ● ●● ● ● ●● ● ●●●● ●● ● ●● ●● ● ● ●●● ●● ● ● ●● ● ● ● ● ●● ● ● ●● ●● ● ●●● ●● ●● ● ● ●● ● ●●● ● ●●●●● ● ● ● ●●●●● ● ●● ● ● ●● ● ● ● ●● ● ●● ●● ● ●●● ● ● ●●● ● ● ●●●● ● ● ●●● ● ●● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ●● ●● ●● ● ● ● ● ● ●● ● ● ●● ●●●●● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ●● ● ●●● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ●● ● ● ● ●●●● ● ● ●● ● ● ●●●●● ● ●● ● ● ● ● ●●● ● ●●● ●● ● ● ● ●● ● ● ●● ●● ● ● ● ●●● ● ● ●● ●●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ●● ● ●●● ● ● ●● ● ● ● ● ● ● ● ● ●● ●●● ● ● ● ● ● ●●●●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ●●●●●●●● ●● ● ● ● ● ●● ● ●● ● ● ● ● ●● ● ● ●●●● ● ● ●●● ●● ●● ●● ● ●● ●● ● ● ● ●●●● ● ● ● ●● ● ●●● ● ● ●● ● ●● ● ● ● ●● ● ● ●●● ●● ● ●● ● ● ●● ●● ●●● ● ●● ●● ●● ● ● ●● ● ● ● ● ● ● ● ●●●●● ●●●● ●● ●● ● ● ●●●● ● ● ●● ●●●●● ●●● ●● ● ● ● ●● ● ●● ●●●●● ●● ● ●● ● ●● ● ●●● ● ● ●● ● ●● ● ●● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●●● ●● ● ● ●● ● ● ● ● ● ● ●● ●● ● ● ● ● ●● ●●●● ● ●● ● ● ● ●● ● ● ● ●●●● ● ●●● ● ● ●● ● ● ●● ●● ●● ●● ● ● ●●●● ● ●● ●● ● ●● ●● ●●●● ● ● ●●● ●●●● ● ● ● ● ●● ●●● ● ● ● ● ● ● ●● ●● ● ● ● ●● ● ●● ●● ● ●●● ● ●● ● ●●●● ● ● ● ● ●● ●●● ● ●● ●● ● ● ● ●● ● ●● ● ● ●● ● ● ● ●●●●● ● ● ● ●● ● ●● ● ● ●● ● ●●● ● ● ● ● ●●● ● ● ● ● ● ●● ● ●● ●●● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●● ● ● ●● ● ● ● ●● ● ●●●● ● ● ● ●● ●● ● ● ● ●● ●●●●● ● ●● ●● ● ●● ●● ● ● ● ● ● ● ●● ●● ● ● ●● ● ● ● ●●● ●● ●● ● ● ● ● ●● ●● ● ● ● ●● ●●● ●●● ●●●●● ●● ● ●● ● ● ●●● ● ●● ● ● ● ● ● ● ●● ●● ●●●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ●● ● ●● ● ● ●● ●● ●● ● ● ● ●● ● ● ● ● ●● ● ●● ● ● ●●● ●● ●● ● ● ● ●●●● ●● ● ●● ●● ● ●● ● ● ●● ●● ● ● ●●● ●● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●●●●● ● ● ●● ●● ● ● ●● ● ●●●●● ● ● ● ● ●●●● ●●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ●● ● ●● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ●● ● ●● ● ●● ● ● ●●● ● ● ● ● ● ●●● ●● ● ●● ● ●●●●● ● ● ●● ●● ●● ● ● ● ●●● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ●● ● ● ● ● ● ●● ● ●● ● ● ●● ● ● ●● ● ●●● ● ●●● ● ● ●●● ● ●●●●● ●● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ●● ● ●● ●● ● ●● ● ● ● ●● ● ●● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ●●● ●●● ●●●● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ●● ● ●●● ● ●● ●● ●● ● ●● ●● ● ●● ●●● ● ● ●● ●●● ●● ● ●● ● ● ●● ● ●●● ● ● ●● ● ●● ●● ● ● ● ●● ● ● ● ● ●● ● ●● ● ●● ● ● ● ●● ●●● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●●● ● ● ●● ● ● ●●●●● ● ●●●● ●● ●● ● ●● ●● ● ● ● ● ●●● ● ● ●● ●● ● ● ● ● ●● ●●● ● ● ● ● ●●● ●● ● ●● ● ● ●●● ●● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ●● ● ●● ● ● ● ● ● ● ●● ●●●● ●●●●●● ● ●● ● ● ● ● ● ●● ● ●●● ● ● ● ●● ● ●● ● ● ● ● ●● ● ● ●● ● ●● ●●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ●● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ●● ●● ● ● ● ● ● ●●● ● ● ● ●● ● ●● ● ● ● ●● ● ●● ●● ● ●●● ● ● ● ● ● ● ● ●●● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ●● ●●● ● ●● ● ●● ●● ● ●●●● ● ● ● ● ● ● ●●● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ●● ●● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ●●●● ● ● ● ● ●● ● ● ● ● ● ●● ● ●●● ●●● ● ● ●●●● ● ●● ● ● ●● ●● ● ●● ● ● ● ●● ● ● ●●● ● ● ● ● ● ●●● ●● ●● ●● ● ● ● ●● ● ●●●● ● ● ●●● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ●● ●● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ●● ●● ● ● ● ● ●● ●● ● ● ●● ● ●● ● ●● ●●● ● ●● ●● ● ● ●● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ●●● ●● ● ● ●● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ●● ● ●● ● ●●●● ●● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ●● ● ● ●● ● ●●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ●● ●● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ●●● ● ● ●● ●● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ●● ● ● ● ● ●● ● ● ●● ●● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●●● ●● ● ● ● ● ● ●●●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ●● ● ● ●● ● ● ● ● ●●● ● ●● ● ● ● ● ●● ● ● ●● ● ● ●●● ● ● ● ● ● ● ● ●●● ● ● ●● ●●● ● ● ● ●● ● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ● ●● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●●● ● ● ● ●● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ●●● ●● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ●● ● ●●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ●● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●●●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ●●● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ●●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ●●●● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●●● ● ● ● ● ●● ●● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ●● ●● ● ● ● ● ● ● ●●●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ●● ● ● ●● ●●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ●● ●● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

10:26 10:28 10:30 10:32 10:34 10:36

Time

Time

750 500 250 0

1500

Distribution of output values

●

0

2500

5000

7500

0.004

0.002

0.000 0

Cloutlet

Figure 7: dEIS Storage Service - VM Performance Indicators (Network, CPU, Memory) queued until it is selected for processing, (2) time it takes to obtain a database connection for processing the data queries, (3) database query execution time, (4) SQL result set processing time, (5) size of the result in bytes, and (6) roundtrip time for sending the result back to WK. In Fig. 7 we show the VM performance metrics associated with the ST service. Both CPU and memory utilization are high due to the large amounts of processing required for retrieving and processing the data from the database. Finally, in Fig. 8 we present how the various durations and round-trip times stack up to compose the complete endto-end time of each chain of requests (CS-WK-ST) and responses (ST-WK-CS). This plot represents the foundation of the simulation model. During the simulation, vertical ”slices” (service-local durations and round-trip times) will be used for simulating CloudSim cloudlets.

3.2

1000

0.006

● ●

1000

density

50

1010000

milliseconds

KBytes

Percentage

Simulation output values 1250

75

10:28

500

(4) Memory Usage

Modeling and Simulation

The performance-profiling traces previously gathered can be represented as shown in Equation 1. P rof ileload is a matrix with all measurable time values of applications’ operations under concurrent load. RT (Si , Sj ) is the round-trip time of a remote service call on service Sj from service Si .

500

1000

Duration

Figure 9: Simulation input vs. output comparison Di (Opk ) is the time of performing service-local operation k on service Si and (cpu|mem|net)CS|LB|W K|ST is the utilization level of CPU, memory and network on the VMs corresponding to CS, LB, WK and ST services. Each line of this matrix corresponds to a single chain of requests from CS, to WK, to ST, and then back to CS. P rof ileload =

⌧

RT (Si , Sj ), Di (Opk ), (cpu|mem|net)CS|LB|W K|ST

(1)

By combining performance profiles for concurrent workload between 1 and max.load end-to-end requests, we form the dEIS performance model P rof ile as dictionary with key load and corresponding values P rof ileload . In order to convert the created performance profile to CloudSim entities we transform the durations in milliseconds to their equivalent MIPS rating. We also map the VM resource consumption for CPU, memory and network to CloudSim characteristics of the created cloudlets - CPU and memory utilization model s and network input and output payloads. The conversion of milliseconds to MIPS is done according to Equation 2. ms · CP UM IP S (2) 1000.0 instr is the length in instructions of an operation with a duration of ms milliseconds when executed on CPU with a MIPS rating of CP UM IP S . This equation is valid only when there is only one task being executed on the CPU. If there are multiple tasks in execution on the same CPU, the tasks’ length in number of CPU instructions needs to be reduced by the CPU concurrency level in order to keep the same execution deadline. As our performance profile contains operations’ durations measured under constant workload, we apply Equation 3 in order to convert the operation’s duration in milliseconds to MIPS. instr =

400

probe A B C D E F G H I J K L M N

Duration (ms)

300

200

100

0 0

1000

2000

Request

Figure 8: dEIS Decomposition of System Response Time

instr ⇤ =

instr · (1 clnew + concurrencyV M

CP UOS )

(3)

instr ⇤ is the task’s length in MIPS considering VM concurrent load, instr is the value calculated by Equation 2, clnew is the number of cloudlets newly submitted to the VM

Experiment: End−To−End Response Time

milliseconds

Conccurrent LB sessnions Conccurrent LB sessnions

Experiment 1

3000 2000 1000 0

15 10

16:16

16:18

16:20

16:22

16:24

16:26

Time milliseconds

5

Simulation: End−To−End Response Time

3000

0 16:16

16:18

16:20

16:22

16:24

●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ●● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ●● ● ●●● ● ●●● ●● ● ●● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ●● ● ● ● ●● ●● ●● ● ●●● ● ●● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ●● ● ●●● ●● ●● ●●● ● ● ●●● ● ●●●●● ● ● ●● ● ●● ● ●● ● ● ●● ●● ● ●● ● ● ● ●● ● ● ● ●● ● ● ●● ● ●● ●● ● ●●● ● ●● ●●● ● ● ● ●●●● ●● ● ●● ● ● ● ● ● ● ●● ● ●●● ● ● ●● ● ● ● ●●● ● ● ●●● ● ● ●●● ● ● ● ● ●● ● ●● ● ● ●● ●● ● ● ● ● ● ●● ●●●● ●● ●● ●● ●● ● ●● ●●● ● ● ●● ● ● ● ● ●●● ● ● ● ● ●● ● ● ● ●● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●●● ● ● ●●● ●● ● ●● ●● ●● ●● ● ●● ●● ●● ●● ● ●● ● ●● ● ●● ● ● ● ● ● ●●● ● ●● ● ● ●● ● ● ● ● ● ●● ● ●● ● ● ●● ● ● ● ●● ● ● ●●●● ●● ● ● ● ●● ●●●●● ●● ●● ●●● ●● ● ●● ● ● ●●● ● ● ● ● ●● ●●● ●● ●●●●● ● ●●● ●● ●●● ●●● ●● ● ● ● ●● ● ● ● ●● ● ●● ●● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●●●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●● ●● ●● ●● ● ● ●●● ●● ●●● ●●● ● ● ● ●● ●● ●● ● ●● ● ●● ●● ●●●● ● ● ●● ●●● ● ● ● ●● ●● ●● ● ● ● ● ● ●● ●●● ●● ●● ● ● ● ● ● ● ●●● ● ●● ● ●● ● ●●● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ●● ● ● ● ● ● ● ●● ●● ● ● ● ● ●● ● ●●●● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ●●●● ● ● ● ● ● ●● ●●● ● ●● ● ● ● ● ● ● ● ● ●●● ●●● ●● ● ● ● ●● ●● ●●● ● ● ●● ● ● ● ●● ● ●● ●● ●● ●● ● ●● ● ● ●● ● ●● ●● ● ● ● ● ●● ●● ● ●● ● ●●● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ●● ●● ● ●● ● ●● ● ● ● ●● ● ● ● ● ●● ● ●● ●● ● ● ● ●●● ● ● ● ● ●● ● ● ● ●● ●●● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ●● ●●●● ● ● ● ●● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ●● ● ● ●●● ● ● ● ● ● ● ● ●● ● ●● ● ●● ●● ● ●● ● ● ●● ● ● ●●● ●● ●● ● ● ● ● ● ● ●● ●● ● ●●● ●● ●● ● ● ● ● ● ● ●● ● ●● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ●●● ● ● ● ● ●● ●●● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ●●●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ●● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ●● ● ● ● ● ● ● ●● ●● ●●● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ●●● ● ●● ●●● ● ● ● ● ● ● ● ●●● ● ● ●● ● ● ●● ●● ● ●● ● ● ●● ● ● ● ●●●● ● ● ● ●● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ●●● ●●● ●● ● ● ●● ● ● ● ● ● ● ●● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ●●● ● ●● ● ● ● ● ● ● ●●●●● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ●●● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ●● ●● ● ●● ●● ● ● ●● ● ● ●●● ● ● ● ●● ● ●● ● ● ●● ● ●●●● ●● ●● ● ●●●● ● ●● ●●●●● ● ● ●● ● ●●● ● ●●● ●●● ● ● ●●● ●● ● ● ●● ●● ●● ● ●● ●● ● ●●●● ● ●●● ● ● ● ● ●● ● ● ● ●●● ● ●●●● ● ● ● ●● ● ●● ● ●● ● ● ●● ●● ● ●●● ●● ● ●● ● ● ●● ●● ● ● ● ● ● ● ● ●●● ● ● ●●● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ●● ●● ●● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ●● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ●● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ●●● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ●●● ● ● ● ● ● ●●● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ●● ●● ● ● ●●● ●● ● ● ●●●● ● ● ● ●●● ● ● ●● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ●● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ●● ● ● ● ● ●●● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ●● ●●●● ● ● ●● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●●● ● ●●●● ● ●● ● ●●●● ● ● ● ● ● ●● ● ●● ● ● ● ●●●●● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ●●●●● ● ●● ● ● ● ●●● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ●● ●● ● ● ● ● ●● ● ● ●●● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ●● ● ● ●●● ● ● ● ●●●● ● ● ● ●● ● ●●● ● ● ●●● ●● ●● ●● ●● ● ● ● ● ● ● ●● ● ●● ● ● ●● ●● ● ●●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ●●●● ● ● ●●● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ●● ●●● ● ● ● ● ● ●●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ●●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●●● ●● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ●●● ●● ● ● ●● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ●● ●●●● ● ● ● ● ● ● ●●● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ●● ●●● ● ● ● ● ● ●●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ●● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ●

2000

16:26

1000

Experiment 2

20

0

0

1000

2000

3000

4000

5000

Request

15

Experiment

10

density

density

5 0 17:18

17:20

17:22

17:24

17:26

0

Time

for execution, concurrencyV M is the number of cloudlets already executed by the VM, and CP UOS is the average CPU utilization caused by independent OS-level kernel tasks. This calculation is performed in each CloudSim Datacenter Broker corresponding to CS, LB, WK and ST services for each batch of cloudlets submitted for execution. For running the simulation in CloudSim, a Datacenter Broker is created for each dEIS service and also at least one VM per dEIS service is added to CloudSim. A simulation scenario is given as a list of pairs of simulation time and the corresponding expected number of concurrent CS-requests at that time. The list is then linearlyinterpolated for determining the expected concurrency for each simulation time slot. For each simulation time step, the current number of running CS-requests cl is compared to the simulation plan values clsim , and if cl < clsim , then (clsim cl) CS-cloudlets are created. Kolmogorov−Smirnov 0.30

● ● ● ● ● ●

● ●

● ● ● ●

● ●

● ●

0.25

● ●

D value

● ●

● ●

● ● ● ● ● ● ● ●

● ●

● ●

0.20 ● ●

0.15

● ●

● ●

5

10

15

20

Linear Dependency

1.02

● ●

0

1000

2000

3000

Request

Figure 12: EIS End-To-End Response Time (ms) - Experiment 1 vs. simulation results comparison with 1 WK and 1 ST VMs CPU, memory and network utilization are identified for each service-local operation and remote service calls, as shown in Fig. 1. These identified parameters are then used for creating CloudSim cloudlets. The chain of cloudlets is then submitted for execution, starting with the CS create-session operation. Subsequently, each cloudlet’s completion will trigger the creation of a cloudlet for the next dEIS operation. Finally, when the last operation in the end-to-end request, the request execution time will be recorded.

4.

EVALUATION RESULTS

We aim to evaluate the CloudSim simulation’s ability to model and reproduce the properties of the targeted distributed application. For this purpose, we used KolmogorovSmirnov (KS) [6] statistical test and linear dependency regression coefficient [10] for comparing the simulation and experimental results. We also evaluated the model’s ability to simulate environments with di↵erent numbers of VMs and di↵erent concurrent workload. The simulations used the same concurrent workload values as the real-world scenario presented in Section 3. The two-sample KS test compares cumulative distribution functions (CDF), and the test statistic is the maximum di↵erence in CDF value, as defined by Equation 4, F2,n0 (x)|

(4)

● ● ● ●

● ● ● ●

● ●

● ● ● ●

● ● ● ●

● ●

● ●

● ● ● ● ● ●

● ●

0.96 ● ●

5

3000

x

● ●

0.98

2000

Dn,n0 = sup |F1,n (x)

● ●

● ●

1.00

1000

Request

Figure 10: Experiments 1 and 2: Concurrent number of LB sessions

Regression Coefficient

Simulation 8e−04 6e−04 4e−04 2e−04 0e+00

6e−04 4e−04 2e−04 0e+00

10

15

20

Concurrent Workload

Figure 11: EIS End-To-End Response Time (ms) Experiment vs. Simulation comparison at constant load Based on the expected concurrency level defined in the simulation plan, one request trace is selected from the performance profile with the corresponding concurrency level. Next, from the performance trace, the operation duration,

where F1,n and F2,n0 are the empirical distribution functions of the first and the second sample respectively. In Fig. 9 we show the comparison between the simulator input values (P rof ile performance traces) and the output produced by CloudSim, showing very similar density distributions of the execution times. The di↵erence between the distribution of values is low, as indicated by a D value of 0.0855 of the KS test. We also validated the ability of the simulation model to correctly reproduce end-to-end response times under constant concurrent workload. In Fig. 11 we show the KS test values and linear dependency between the values gathered in experiment and simulation. The graphs show a good correlation with a value of 0.3 of KS statistic (maximum

milliseconds

Experiment: End−To−End Response Time

2000 1000 17:18

17:20

17:22

17:24

17:26

milliseconds

Time

Simulation: End−To−End Response Time ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ●● ● ● ● ●● ●● ● ●● ● ● ● ●● ● ● ● ●●●● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●● ● ●●● ●● ● ● ●● ●●● ●● ●● ●● ● ● ●● ● ● ● ● ● ● ● ●● ●●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●●● ● ● ●● ● ●● ● ● ●● ● ● ● ● ● ●●● ●●● ● ● ● ●● ● ●● ● ● ●● ● ● ●● ● ● ●● ● ●●● ●●● ● ● ● ● ●● ● ●●●● ●● ● ●●● ●● ● ●●● ● ● ● ● ● ●●●● ● ●●● ● ● ●● ● ● ● ● ●● ● ● ● ●● ● ● ●●● ● ● ●● ● ● ● ● ● ● ●●●● ● ●● ● ● ● ● ● ●● ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ●●● ● ● ●● ● ●● ● ● ● ● ●●●● ●● ●●●● ● ●●● ● ● ●●● ● ● ● ● ●●●● ● ●● ● ● ●● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ●● ● ● ● ● ● ● ● ●● ●● ●● ● ●● ● ●● ●●● ●● ● ●● ● ●● ● ●●●● ●● ●●● ● ● ●● ● ●● ●●● ●● ● ● ● ●●● ● ●● ● ● ● ● ●●●● ● ●● ● ● ● ● ●● ● ● ● ● ●● ●●●● ● ●● ● ●●●● ● ● ●● ● ● ● ●● ● ● ● ● ● ●● ● ● ●● ● ●● ● ● ●● ● ● ●● ● ● ● ●● ● ● ● ● ●● ● ● ●● ● ● ● ●● ●● ● ●● ● ●●● ●● ● ● ● ● ● ● ●●● ● ●● ● ●●●● ● ● ● ● ● ● ● ●● ● ●● ●● ● ●● ● ● ●● ● ● ●● ● ●●●●● ● ●● ●●● ●● ●● ● ●● ● ● ● ● ● ● ● ●● ● ● ●●● ●● ● ● ● ● ● ● ● ● ● ●●● ● ●● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ●● ●●●● ● ● ● ●●● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ●● ●●● ●●● ● ● ● ● ●●● ● ● ● ● ● ● ●● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ●● ●● ● ● ● ● ● ●● ●●●●● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ●● ●● ●● ● ● ●● ● ● ● ● ●● ● ●● ● ● ●● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ●● ●● ● ● ●● ● ● ● ● ●● ●● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ●●● ● ● ● ●●● ● ●● ●● ● ●● ● ● ● ● ● ●● ●● ● ●●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ●● ●● ●● ● ●●●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●●● ● ● ● ●● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ●● ● ●● ●● ● ●●●●●●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ●●● ● ● ● ●●● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ●● ● ● ● ● ● ● ● ●●● ● ● ●● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ●● ● ● ● ●● ● ●● ● ● ● ●●●● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●●● ● ●●● ● ● ● ● ● ●● ●● ● ● ● ● ●● ● ● ●● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●●● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●●● ●●● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●●● ● ● ● ●● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ●● ● ● ● ● ●● ● ● ●● ●● ● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ●●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

2000

●

●

1000 0

● ●

●

0

2500

5000

7500

10000

Request

Experiment 0.0020 0.0015 0.0010 0.0005 0.0000

Simulation density

density

0.0020 0.0015 0.0010 0.0005 0.0000

0

1000

2000

Request

0

1000

2000

Request

Figure 13: EIS End-To-End Response Time (ms) - Experiment 2 vs. simulation results comparison with 2 WK and 2 ST VMs di↵erence between CDF functions), and a linear coefficient of 1.05 between experiment and simulation results (indicating that the simulation values di↵er with only 5% from the experiment values). In order to test the simulation model under varying workload conditions and with a di↵erent number of VM instances for the dEIS services, we ran two additional experiments, as shown in Fig. 10. In the first experiment still used one VM per service. However, we varied the concurrent workload from 1 to 17 and back to 1 parallel requests. In the second experiment we used 2 VMs for each of WK and ST services, using in total 4 servers, while the concurrent workload was varied from 1 to 20 parallel requests, with 5 seconds pause before switching to a new workload level. The experiments were performed with real resources and then the results were compared with the same scenario running in the simulator. In case of experiment 1, simulation results shown in Fig. 12 validated the modeling of the dependency between service tasks durations and the corresponding number of executed CPU instructions, as the simulated end-to-end response time closely followed the measured dEIS response times. The increased variance in simulation results was caused by the simulator’s event communication overhead and by the fact that the simulator randomly chooses the requests to be simulated. Experiment 2 validated the modeling of concurrent computing workloads distributed across multiple VMs. The simulation results closely matched the ones measured during the experiment experiment, as can be seen in Fig. 13.

5.

CONCLUSIONS

We have presented an approach for building a simulation model of a distributed application for concurrent workload processing, by analyzing application’s performance traces gathered in small-scale real deployments. We then showed

how to integrate the simulation model into CloudSim and how to build a simulation with varying concurrent workload. We have also presented an approach for modeling concurrent computing workloads using the CloudSim simulator. Finally, we validated our models by running experiments with varying workload and with di↵erent VM scaling factors. Our results show that it is possible to accurately model the behavior of a distributed enterprise information system using CloudSim, although extensions for supporting concurrent task processing need to be added. Also, due to variance in the execution times of distributed requests, the simulation models cannot perfectly represent the real distributed systems, however, a good-enough representation of key application-level metrics has been shown to be obtainable.

6.

REFERENCES

[1] A.-F. Antonescu and T. Braun. Improving management of distributed services using correlations and predictions in sla-driven cloud computing systems. In Proc. IEEE/IFIP Network Operations and Management Symposium (NOMS), May 2014. [2] A.-F. Antonescu, A.-M. Oprescu, et al. Dynamic optimization of sla-based services scaling rules. In Proc. 5th IEEE Internetional Conference on Cloud Computing Technology and Science (CloudCom), December 2013. [3] A.-F. Antonescu, P. Robinson, and T. Braun. Dynamic topology orchestration for distributed cloud-based applications. In Proc. 2nd IEEE Symposium on Network Cloud Computing and Applications (NCCA), 2012. [4] A.-F. Antonescu, P. Robinson, and T. Braun. Dynamic sla management with forecasting using multi-objective optimizations. In Proc. 13th IFIP/IEEE Symposium on Integrated Network Management (IM), May 2013. [5] R. Buyya et al. Modeling and simulation of scalable cloud computing environments and the cloudsim toolkit: Challenges and opportunities. In Int. Conf. on High Performance Computing & Simulation, 2009. HPCS’09., pages 1–11. IEEE, 2009. [6] W. Conover. Practical nonparametric statistics. J. ˘ S314, 1980. Wiley., New York, page 309ˆ aA¸ [7] S. K. Garg et al. Networkcloudsim: Modelling parallel applications in cloud simulations. In Utility and Cloud Computing (UCC), 2011 Fourth IEEE International Conference on, pages 105–113. IEEE, 2011. [8] T. Goyal et al. Cloudsim: simulator for cloud computing infrastructure and modeling. Procedia Engineering, 38:3566–3572, 2012. [9] P. Mell and T. Grance. The NIST definition of cloud computing. NIST special publication, 2011. [10] C. R. Rao et al. Linear models. Springer, 1995. [11] D. Woods. Enterprise Services: Architecture. O’Reilly Media, Inc., 2003.