Successfully reported this slideshow.
Your SlideShare is downloading. ×

The Computing Continuum.pdf

The Computing Continuum.pdf

Download to read offline

The advent of fog and edge computing has prompted predictions that they will take over the traditional cloud for information processing and knowledge extraction in Internet of Things (IoT) systems. Notwithstanding the fact that fog and edge computing have undoubtedly large potential, these predictions are probably oversimplified and wrongly portray the relations between cloud, fog and edge computing. 

Concretely, fog and edge computing have been introduced as an extension of the cloud services towards the data sources, thus forming the computing continuum. The computing continuum enables the creation of a new type of services, spanning across distributed infrastructures, supporting various IoT applications. These applications have a large spectrum of requirements, burdensome to meet with "distant'' cloud data centers. However, the introduction of the computing continuum raises multiple challenges for management, deployment and orchestration of complex distributed applications, such as: increased network heterogeneity, limited resource capacity of edge devices, fragmented storage management, high mobility of edge devices and limited support of native monolithic applications. These challenges primarily concern the complexity and the large diversity of the devices, managed by different entities (cloud providers, universities, private institutions), which range from single-board computers such as Raspberry Pis to powerful multi-processor servers.

Therefore, in this talk, we will discuss novel algorithms for low latency, scalable, and sustainable computing over heterogeneous resources for information processing and reasoning, thus enabling transparent integration of IoT applications. We will tackle the heterogeneity challenge of dynamically changing topologies of the computing infrastructure and present a novel concept for sustainable processing at scale.

The advent of fog and edge computing has prompted predictions that they will take over the traditional cloud for information processing and knowledge extraction in Internet of Things (IoT) systems. Notwithstanding the fact that fog and edge computing have undoubtedly large potential, these predictions are probably oversimplified and wrongly portray the relations between cloud, fog and edge computing. 

Concretely, fog and edge computing have been introduced as an extension of the cloud services towards the data sources, thus forming the computing continuum. The computing continuum enables the creation of a new type of services, spanning across distributed infrastructures, supporting various IoT applications. These applications have a large spectrum of requirements, burdensome to meet with "distant'' cloud data centers. However, the introduction of the computing continuum raises multiple challenges for management, deployment and orchestration of complex distributed applications, such as: increased network heterogeneity, limited resource capacity of edge devices, fragmented storage management, high mobility of edge devices and limited support of native monolithic applications. These challenges primarily concern the complexity and the large diversity of the devices, managed by different entities (cloud providers, universities, private institutions), which range from single-board computers such as Raspberry Pis to powerful multi-processor servers.

Therefore, in this talk, we will discuss novel algorithms for low latency, scalable, and sustainable computing over heterogeneous resources for information processing and reasoning, thus enabling transparent integration of IoT applications. We will tackle the heterogeneity challenge of dynamically changing topologies of the computing infrastructure and present a novel concept for sustainable processing at scale.

More Related Content

More from Förderverein Technische Fakultät

Related Books

Free with a 30 day trial from Scribd

See all

Related Audiobooks

Free with a 30 day trial from Scribd

See all

The Computing Continuum.pdf

  1. 1. The Computing Continuum: Beyond the Cloud Data Centers Dragi Kimovski Pre-submission Habilitation Talk Klagenfurt University 30.06.2022
  2. 2. About me Dragi Kimovski Postdoc assistant with “Zielvereinbarung” Previous positions: • 2016-2018 Senior Researcher and Lecturer, University of Innsbruck, Austria • 2013-2019 Assistant Professor, University of Information Science and Technology, Macedonia • 2009-2013 Teaching Assistant, University of Information Science and Technology, Macedonia Visiting positions: • University of Utrecht, Netherlands (December 2022) • University of Michigan - Ann Arbor, US • University of Bologna, Italy • University of Granada, Spain Project experience: • H2020 ASPIDE (2018-2021), Scientific coordinator • H2020 Entice project (2016-2018), WP leader and integration manager • H2020 DataCloud (2021-2023), WP leader • FFG KärntnerFog (2022-2024), WP Leader • OeAD AtomicFog (2018-2020), Coordinator Publications: Over 50 publications in Journals and Conferences 2
  3. 3. Dragi Kimovski Postdoc assistant with “Zielvereinbarung” Teaching experience: • Distributed systems, Distributed computing infrastructures, Cloud computing and IoT, Advanced programming in C++ @ University of Klagenfurt, Austria • Parallel Systems, Distributed Systems@ University of Innsbruck, Austria • Parallel computing, High performance computing, Computing organization, Computer networks, Computing systems configuration, Programming @ University of Information Science and Technology, Macedonia • Parallel processing @ Technical University of Sofia, Bulgaria Supervision: • 13 Bachelor thesis (2 in progress) • 3 Master thesis (5 thesis in progress) • 2 PhD thesis (Co-supervision) 3 About me
  4. 4. Introduction and overview
  5. 5. Cloud computing and Internet of Things 5
  6. 6. The computing continuum and Internet of Things Jetson Nano 6
  7. 7. Added value of the computing continuum Reduction of the amount of data being sent to the Cloud Faster response times Support for new applications (Smart cities, e-health, …) 7 Improved data security Reduced CO2 emissions
  8. 8. Is the computing continuum “magical”? 8 • Can the computing continuum solve the computational issues related to the emerging Internet of Things application? • The answer is: Na ja.
  9. 9. Issues (At least some of them) Heterogeneity of devices Devices can move Application distribution 9 Various control domains Data management and distribution
  10. 10. 10 Open problems and questions How to predict the mobility and the behavior of the resources? How to optimize storage repositories? How to parallelize and distribute monolithic applications? How to characterize the resources and the computing continuum architecture? How to adaptively place and schedule IoT applications? Resources characterization Placement and scheduling Mobility patterns prediction Application distribution Storage optimization
  11. 11. 11 02 Resources characterization
  12. 12. The heterogenity of the computing continuum • The heterogeneity of the computing continuum raises multiple application management challenges: – where to offload an application from the cloud to the fog or to the edge. • Large diversity of the devices: – single-board computers such as Raspberry Pis; – powerful multi-processor servers. 12
  13. 13. Performance characterisation • To answer this question it is essential to characterize the performance of the resources across the computing continuum. • We present an analysis of: • computational and network performance; • carbon emissions. • The main goal is to support the decision process for offloading an application.
  14. 14. The Carinthian Computing Continuum – C3 Table 1: Description of the resources available in the C3 testbed. Conceptual layer Device / Instance type Architecture (v)CPU Memory [GiB] Storage [GiB] Network Physical processor Clock [GHz] Operating system Cloud layer AWS t2.micro 64-bit x86 1 1 32 Moderate Intel Xeon  3.1 Ubuntu 18.04 AWS c5.large 2 4  10 Gbps Intel Xeon Platinum 8000 series  3.6 AWS m5a.xlarge 4 16 AMD EPYC 7000 series  2.5 Fog layer Exoscale Tiny 64-bit x86 1 1 32  10 Gbps Intel Xeon  3.6 Ubuntu 18.04 Exoscale Medium 2 4 Exoscale Large 4 8 ITEC Cloud Instance 4 8 Intel Xeon Platinum 8000  3.1 Edge layer Edge Gateway System 64-bit x86 12 32 32  10 Gbps AMD Ryzen Threadripper 2920X  3.5 Ubuntu 18.04 Raspberry Pi 3B 64-bit ARM 4 1 64  1 Gbps Cortex - A53  1.4 Pi OS Buster Raspberry Pi 4 4 Cortex - A72  1.5 Jetson Nano 4 Tegra X1 and Cortex - A57  1.43 Linux for Tegra R28.2.1 2.2 Fog layer 112 The fog layer comprises computing infras- 113 tructures consolidated in small data-centers 114 in close vicinity to the data sources. This layer 115 comprises resources from two providers in 116 the C3 testbed [4]: Exoscale and University of 117 Klagenfurt. We allocate these providers in the 118 fog layer as a result of the low round-trip com- 119 munication latency ( 7 ms) and high band- 120 width ( 10 Gbps). The Exoscale cloud com- 121 prises data centers in Vienna and Klagenfurt 122 (Austria). We selected three computing opti- 123 mized x86-64 instances from the Exoscale 124 cloud offering: Tiny, Medium and Large. 125 University of Klagenfurt provides a private 126 cloud infrastructure operated by OpenStack 127 3 Benchmark applications 150 We selected three representative appli- 151 cation classes with complementary require- 152 ments to evaluate the computational perfor- 153 mance and the CO2 emissions of the com- 154 puting continuum. 155 3.1 Video encoding 156 Video encoding allows transmission of 157 video content with different qualities over 158 limited and heterogeneous communication 159 channels. It compresses an original raw video 160 to reduce its effective bandwidth consump- 161 tion, while maintaining a subjective high qual- 162 ity for viewers. Video encoding has wide fields 163 of applications, including content delivery (live 164 and on-demand video streams), traffic control 165 14
  15. 15. Video encoding • Ffmpeg version 3.4.6 with the most popular H.264/MPEG-4 video encoder. • Raw video segment with length of 4 second and size of 514 MB. rge tion nce Fm- ular by We eg- MB, deo HD- ates ding urce achieved the lowest encoding and transfer 233 time due to the low utilization rate and its high 234 computing and networking capabilities. R P i 3 B R P i 4 J e t s o n E G S E S T i n y E S M e d i u m E S L a r g e A W S t 2 . m i c r o A W S c 5 . l a r g e A W S m 5 a . x l a r g e 0 10 20 30 40 30.6 4.8 3.6 0.4 1 0.68 0.75 4.7 0.7 0.6 32.3 5.5 4 0.5 1.4 0.89 0.93 4.7 0.8 0.7 36.3 8.2 6.4 0.9 3.1 1.4 1.2 5.2 1.4 1.3 Execution time [s] HD-Ready (720p) Full HD (1080p) Quad HD (1440p) (a) Average encoding time. 200 7.5 170.6 63.8 4 Performance evaluation 191 4.1 Video encoding 192 We evaluate the encoding performance 193 of the computing continuum using FFm- 194 peg version 3.4.6 with the most popular 195 H.264/MPEG-4 video encoder4 deployed by 196 more than 90% of the video industry5 . We 197 perform the encoding on a raw video seg- 198 ment with length of 4 s and size of 514 MB, 199 available in the Sintel6 video-set. The video 200 segment is encoded in three resolutions (HD- 201 ready, Full HD and Quad HD) with data rates 202 of 1500, 3000, and 6500 kbps. 203 Figure 2 depicts the average encoding 204 time and transfer time, from the video source 205 (located at the University of Klagenfurt) to 206 the encoding device or instance, for a single 207 raw video segment in the three resolutions. 208 The standard deviation ranges from 1.3% 209 for the AWS m5a.xlarge instance to 3.6% 210 for the Raspberry Pi 3B devices. We ob- 211 serve that the older generation single-board 212 computers (Raspberry Pi 3B) have a signif- 213 icantly higher encoding time than the other 214 resources. However, the Raspberry Pi 3B 215 devices provide lower transfer times than the 216 cloud instances and are suitable for video-on- 217 demand services employing offline encoding. 218 The Raspberry Pi 4 and the Jetson Nano de- 219 R P i 3 B R P i 4 J e t s o n E G S E S T i n y E S M e d i u m E S L a r g e A W S t 2 . m i c r o A W S c 5 . l a r g e A W S m 5 a . x l a r g e 0 10 20 30 40 30.6 4.8 3.6 0.4 1 0.68 0.75 4.7 0.7 0.6 32.3 5.5 4 0.5 1.4 0.89 0.93 4.7 0.8 0.7 36.3 8.2 6.4 0.9 3.1 1.4 1.2 5.2 1.4 1.3 Execution time [s] HD-Ready (720p) Full HD (1080p) Quad HD (1440p) (a) Average encoding time. R P i 3 B R P i 4 J e t s o n E G S E S T i n y E S M e d i u m E S L a r g e A W S t 2 . m i c r o A W S c 5 . l a r g e A W S m 5 a . x l a r g e 0 50 100 150 200 12.5 5.1 9.2 5.1 74.5 63.1 67.1 157.5 170.6 163.8 Transfer time [s] (b) Average raw video segment transfer time. Figure 2: Average encoding performance of a 4 s long video segment with the x264 codec and FFmpeg 3.4.6. 15
  16. 16. Machine learning • Tensor Flow training: – A quantum neural network using the MNIST data-set limited to 20000 samples with a size of 3.3MB with 90% accuracy; – A convolutional neural network using the Kaggle data-set with a size of 218MB with 80% accuracy. enarios ages: ng the amples rio cre- ers and r to the each a 0%. ing the 18 MB. s 80%. e layers er uses e range we use tization nsions. execu- network e train- the de- raining. m 1.2% 4% for aluation neural is significantly higher for the convolutional 296 neural network, the cloud and fog resources 297 outperform the edge devices, except EGS. 298 Recommendation. We recommend the 299 model training with large data-sets and 300 multiple layers in the cloud or on dedicated 301 systems (such as EGS), whenever possible. 302 We recommend offloading to the edge only 303 when the training data is of limited size, or 304 when the neural network has few layers. R P i 3 B R P i 4 J e t s o n E G S E S T i n y E S M e d i u m E S L a r g e A W S t 2 . m i c r o A W S c 5 . l a r g e A W S m 5 a . x l a r g e 0 100 200 300 400 500 54.1 35.6 40.1 6.2 8.4 8.7 8.4 5.3 6.2 8.6 1,630 1,050 432 65.2 210.2 192.1 134.26 220.1 178.83 180.3 Execution time [s] Quantum neural network Convolutional neural network (a) Average training time. 80 .3 1 The convolutional network has three layers 262 with a kernel size of three. Each layer uses 263 increasingly higher filter sizes in the range 264 [32, 64, 128]. After each layer, we use 265 a max-pooling sample-based discretization 266 process to reduce the spatial dimensions. 267 We repeat the training five times. 268 Figure 3 analyzes the average execu- 269 tion time for training the two neural network 270 types and the transfer times of the train- 271 ing data from centralized storage to the de- 272 vice or instance that performs the training. 273 The standard deviation ranges from 1.2% 274 for the Raspberry Pi 4 devices to 5.4% for 275 the AWS t2.micro instance. The evaluation 276 shows that the less complex quantum neural 277 network requires a relatively lower training 278 time across all resources. The old generation 279 single-board computers show again a lower 280 performance, and their suitability for training 281 heavily depends on the size of the training 282 data and the model. The other fog and edge 283 devices provide similar performance to the 284 cloud resources. The single-board computers 285 provide lower training performance for the 286 convolutional network. The only exception are 287 the Jetson Nano devices able to train the 288 convolutional network up to four times faster 289 than the Raspberry Pi devices. In general, 290 the EGS provides the lowest training time 291 R P i 3 B R P i 4 J e t s o n E G S E S T i n y E S M e d i u m E S L a r g e A W S t 2 . m i c r o A W S c 5 . l a r g e A W S m 5 a . x l a r g e 0 100 200 300 400 500 54.1 35.6 40.1 6.2 8.4 8.7 8.4 5.3 6.2 8.6 1,6 1,0 432 65.2 210.2 192.1 134.26 220.1 178.83 180.3 Execution time [s] Quantum neural network Convolutional neural network (a) Average training time. R P i 3 B R P i 4 J e t s o n E G S E S T i n y E S M e d i u m E S L a r g e A W S t 2 . m i c r o A W S c 5 . l a r g e A W S m 5 a . x l a r g e 0 20 40 60 80 0.1 0.1 0.1 0.1 0.4 0.4 0.4 1 1 1 5.4 2.2 3.9 2.1 32.2 27.3 29.1 68.3 65.1 68 Transfer time [s] Data-set (Quantum neural network) Data-set (Convolutional neural network) (b) Average training data transfer time. Figure 3: Average training and data transfer 16
  17. 17. Carbon emissions analysis • We evaluate the power consumption of the physical devices used for the convolutional neural network training in TensorFlow. – We use a digital multimeter to physically measure the average electrical current; – We rely on an AWS research report to approximate the power consumption of the fog devices and cloud instances. S E S T i n y E S M e d i u m E S L a r g e A W S t 2 . m i c r o A W S c 5 . l a r g e A W S m 5 a . x l a r g e 55.1 65.3 61.2 26.9 24.8 25.2 0.94 7.1 7 7 22.1 20.7 20.1 R P i 3 B R P i 4 J e t s o n E G S E S T i n y E S M e d i u m E S L a r g e A W S t 2 . m i c r o A W S c 5 . l a r g e A W S m 5 a . x l a r g e 0 1 2 3 4 5 6 1 0.7 0.71 1.5 2.3 4.1 3.9 2.4 3.8 5.4 Carbon emission [g] Carbon emissions Figure 6: Carbon footprint for training a neural 17
  18. 18. What we learned • We compiled a set of recommendations for practitioners on where to offload their applications across the computing continuum. • However, is this enough for proper application management? Recommendations for application offloading across the computing co Application Requirement Low network load Low execution time Low CO2 emissions Video encoding Edge/Fog Cloud Edge Machine learning Edge Cloud/Fog Edge In-memory analytic Cloud/Fog Cloud Edge ional Conference on the Internet of Things Prodan, R., Barrionuevo, J. J. D., Fahringer, ay). A multi-objective approach for workflow n heterogeneous environments. In 2012 12th International Symposium on Cluster, Cloud puting, and communication in UAV swa Radu Prodan is professor in distribute ITEC, Klagenfurt University. He receive gree in 2004 from the Vienna University and was Associate Professor until 201 versity of Innsbruck (Austria). His rese Research papers: 1. Dragi Kimovski, Josef Hammer, Narges Mehran, Hermann Hellwagner and Radu Prodan, “Cloud, Fog or Edge: Where to Compute”, IEEE Computing Magazine, 2021. 2. Roland Matha, Dragi Kimovski, Anatoliy Zabrovsky, Christian Timmerer and Radu Prodan, “Where to Encode: A Performance Analysis of x86and Arm-based Amazon EC2 Instances”, eScience 2021. 18
  19. 19. 19 Scheduling and mobility prediction
  20. 20. Placement and scheduling • To efficiently place and schedule complex distributed applications’ workflows in heterogeneous environment, with limited computing and storage capacity. 20
  21. 21. Approach • To cope with the complexity of the problem we apply multi-objective approach. • Searching for a set of non- dominated scheduling/placements of applications on the continuum. • An automated decision-making module to select an appropriate solution. 21
  22. 22. Approach • Objectives: • Time à lower completion time; • Energy à reduce energy consumption; • Cost à reduce monetary cost. • Constrained on devices in the computing continuum that tend to move (roam) less. 22
  23. 23. Objectives • Time objective: • The computation time; • The time required for transferring data among components. Computation time rk rj mi pred (mi) 23
  24. 24. Objectives • Energy objective: • Energy for executing the component on a device; • Energy for receiving data; • Static energy consumption for maintaining the device active. 24
  25. 25. Objectives • Cost objective: • Processing the application on CPUj • Storing the data on STORj • Communication cost between resources 25
  26. 26. The scheduling and placement problem i | i 2 N, 0  i   }, where |M|= . Each decision vari- ble i is the placement of one component mi onto a source: i = plc (mi). The goal is to find a placement c(A) for an application A that assigns all its components the set R of resources that minimizes the three objectives: 8 > > > < > > > : f1(T) = min plc(A)=R T(A,R); f2(E) = min plc(A)=R E(A,R); f3(C) = min plc(A)=R C(A,R). (11) Searching an optimal placement plc(A) results in a set solutions, which must satisfy the processing, memory nd storage constraints on device rj = (CPUj, MEMj, STORj) signed to each component mj, and the movement proba- lity of a device rj within a given time window P W rj (s): to the set R of resources that minimizes the three objectives: 8 > > > < > > > : f1(T) = min plc(A)=R T(A,R); f2(E) = min plc(A)=R E(A,R); f3(C) = min plc(A)=R C(A,R). (11) Searching an optimal placement plc(A) results in a set of solutions, which must satisfy the processing, memory and storage constraints on device rj = (CPUj, MEMj, STORj) assigned to each component mj, and the movement proba- bility of a device rj within a given time window P W rj (s): 8 > > < > > : CPU (mi) < CPUj; MEM (mi) < MEMj; STOR (mi) < STORj; P W rj (s) < km. (12) where k is a mobility constraint coefficient defined by • We utilize NSGA-II to tackle this NP-complete optimization problem 26
  27. 27. Mobility constraint • Improve application scheduling by addressing the mobility characteristics of the Fog/Edge devices in the computing continuum. • Prediction of the mobility allows to optimize: • Application placement: select the device having lowest mobility; • Priority scheduling: assign „harder“ tasks to less mobile devices and vice versa. 27
  28. 28. Mobility characterization • By modeling the devices in the continuum 𝑟 ! through (first order) discrete Markov chains (MC). • Calculates the probability of a system to reach a particular state in future. • Consists of (already specialized): • Finite set of states, i.e. {𝑆", 𝑆#, 𝑆$}. • 𝑆" − 𝐷𝑖𝑠𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑒𝑑; • 𝑆# − 𝐶𝑜𝑛𝑛𝑒𝑐𝑡𝑒𝑑; • 𝑆$ − 𝑅𝑜𝑎𝑚𝑒𝑑. 𝑆! 𝑆" 𝑆# " 1 3 0 " 2 3 0 1 " 3 4 0 0 " 1 4 28
  29. 29. Markov chain mobility prediction • Given a MC for a (𝑤, 𝑑) – tuple and the functions: – 𝑖𝑛𝑑𝑒𝑥 𝑟 ≔ 𝑅𝑒𝑡𝑢𝑟𝑛𝑠 𝑡ℎ𝑒 𝑙𝑜𝑤𝑒𝑠𝑡 𝑖𝑛𝑑𝑒𝑥 𝑐𝑜𝑛𝑡𝑎𝑖𝑛𝑖𝑛𝑔 𝑡ℎ𝑒 𝑚𝑎𝑥𝑖𝑚𝑢𝑚 𝑜𝑓 𝑡ℎ𝑒 𝑣𝑒𝑐𝑡𝑜𝑟 𝑟 ∈ 0, 1 ! × # – 𝑠𝑡𝑎𝑡𝑒 𝑖𝑛𝑑𝑒𝑥 ≔ ; 𝑖𝑛𝑑𝑒𝑥 = 1 → 𝑆$ 𝑖𝑛𝑑𝑒𝑥 = 2 → 𝑆% 𝑖𝑛𝑑𝑒𝑥 = 3 → 𝑆& • We can derive the next state that is most likely to occur by: 𝑠$%& = 𝑠𝑡𝑎𝑡𝑒 𝑖𝑛𝑑𝑒𝑥 𝜋 ⋅ (𝑃!! '( )$ / 𝑖 ∈ ℕ 𝑖𝑠 𝑡ℎ𝑒 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑎𝑛𝑠𝑖𝑡𝑖𝑜𝑛𝑠 = 𝜋 ⋅ 𝑃!) "# ⋅ … ⋅ 𝑃!) "# = 𝑖 𝑡𝑖𝑚𝑒𝑠 29
  30. 30. Case studies Mental health care Insulin pump 7: end if 8: end for 9: return True . Constraints fulfilled 10: end function Algorithm 3 mMAPO automated decision making 1: function DECISION_MAKING(Y ) 2: Input: ! OPV . Objective priority vector 3: IC initiate_centroids( ~ OPV ) 4: CP cluster_pareto(Y, IC) 5: returnselect_pareto_solution(CP) 6: end function IoT micro-sensors embedded in the patient body [33]. The sensors send the blood sugar information to the resources executing machine learning algorithms to create a model that identifies the patient state variation and computes the proper level of insulin upon abnormal state detection. Afterwards, the pump controller receives the information through eight components interacting as in Fig. 4: • Compute blood sugar level of the patient; • Compute insulin level and store it in a remote database; • Retrieve patient records from the database; • Review values of a patient for taking proper decisions; • Send to doctor the blood sugar for review of insulin intake; • Send review results back with the proper insulin dose based on the patient history. • Compute pump command and adjust miniaturized pump pressure to avoid the risk of falling into coma; • Control insulin pump needle for delivering the correct dose. 6.2 Mental healthcare This application manages near real-time information of pa- tients who suffer from mental disorders [33] in a number of UK hospitals (Fig. 5). Due to privacy concerns, the patients may not always attend the same clinic and need support through appointments and emergency services: • Determine mental state for a specific patient’s record; • Decompose the safety concern for the patient to prevent Compute insulin level doctor Compute pump command insulin pump Log insulin dose patient records  sensor pump Review values blood   sugar Figure 4. Insulin pump application. Decom- pose the safety concern Mental health act Find the closest clinic Generate record for medical staff  Smart wearable device  Emergency services Determine  mental states Summarize View patient history Figure 5. Mental healthcare application. 7 EXPERIMENTAL SIMULATION We implemented the mMAPO Pareto analysis algorithm in the jMetal [34] framework and integrated within the Sched- uler of the ASKALON environment. We created elaborate simulation scenarios using iFogSim [14], which considers the computational and storage characteristics of both the Edge devices and the Cloud virtual machine instances. 7.1 Experimental design We evaluated the benefits of mMAPO for application place- ment compared to three state-of-the-art methods: 1) Fog Service Placement Problem (FSPP) [11] based on linear inte- ger programming model focused on reducing the economic cost and improving resources utilization; 2) Edge-ward de- lay-priority (EW-DP) [8] that implements a hierarchical best fit algorithm to cope with users mobility; and 3) Best-fit Queue (BQ) [12] as a queuing algorithm that reduces the completion time by primarily using Edge devices. We con- sidered completion time, energy consumption and economic cost for executing a request from the IoT sensors until the final data collection at another device or end-user. 6 nents ri) _ filled filled ector The urces model putes ction. ation ase; ns; sulin Table 1 Resource requirements per component. Application CPU [MI] MEM [MB] Storage [MB] Insulin pump 200 – 2000 10 – 60 256 – 1024 Mental healthcare 200 – 2000 10 – 50 256 – 512 Compute insulin level Send to doctor Compute pump command Control insulin pump Log insulin dose Retrieve patient records Blood  sensor Insulin pump Review values Compute blood   sugar Figure 4. Insulin pump application. Decom- pose the safety concern Mental health act Find the closest clinic Generate record for medical staff  Smart wearable device  Emergency services Determine  mental states Summarize View patient history Figure 5. Mental healthcare application. 7 EXPERIMENTAL SIMULATION We implemented the mMAPO Pareto analysis algorithm in the jMetal [34] framework and integrated within the Sched- uler of the ASKALON environment. We created elaborate simulation scenarios using iFogSim [14], which considers the computational and storage characteristics of both the 6 Table 1 Resource requirements per component. Application CPU [MI] MEM [MB] Storage [MB] Insulin pump 200 – 2000 10 – 60 256 – 1024 Mental healthcare 200 – 2000 10 – 50 256 – 512 Send to doctor Control insulin pump Retrieve patient records Blood  sensor Insulin pump Compute blood   sugar 30
  31. 31. Evaluation (placement and scheduling) Figure 3: Insulin pump application completion time, energy consumption, and economic cost for different data size Figure 4: Mental health care application completion time, energy consumption, and economic cost for different CPU load 31
  32. 32. Evaluation (mobility prediction) Figure 5: Experimental evaluation of the Markov chain single-transition mobility prediction model Figure 6: Average request probability failure 32 1,000 2,000 4,000 0 50 INSTR [MI] Co 1,000 2,000 4,000 0 1,000 INSTR [MI] Energ 1,000 2,000 4,000 0.5 1 INSTR [MI] Ec Figure 11. Mental healthcare application time, energy, and cost for different CPU workloads. 0 25 50 100 200 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Number of prediction samples per device Execution time [ s ] 25 50 100 200 80 90 100 Number of prediction samples per device Accuracy [%] 0 25 50 100 80 90 100 Number of single-step predictions per device Accuracy [%] DPM IDV Figure 12. Experimental evaluation of the Markov chain single-transition mobility prediction model. 9.4 Request failure probability This section evaluates the effect of the mobility on the request failure probability for executing the mental health application. We performed a series-based reliability analysis [39] that identifies the average time spent by the Edge devices in a connected Sc or roamed Sr state in the four ime windows W defined in Section 9.2: Morning (MO), Forenoon (FN), Afternoon (AN) and Night (NI). The empir- cal analysis shows that the average connected or roamed ime T for each time window W is 2790 s, 2888 s, 3322 s and 2442 s, respectively. We therefore utilize this informa- ion to create probability function for a request failure f in failure rate. In contrast, FSPP primarily relies on the Cloud infrastructures with limited use of Edge devices, which results in a relatively low failure probability of around 35%. The introduction of the mobility prediction reduced the failure probability of mMAPO to less than 10%. The high probability of a request failure in EW-DP, BQ and FSPP indirectly increases the energy and execution costs by a factor of two to six compared to mMAPO. 9.5 Complexity and quality analysis We investigate the ability of mMAPO to provide optimized MO FN AN NI 0 20 40 60 80 100 Window W Requests probability failure [%] FSPP mMAPO EW-DP BQ Figure 13. Average request probability failure for mental healthcare application. Cloud. Moreover, the mMA Edge devices can reduce th up to 80%. The results also s is more energy efficient for require high performance r munication latency has a l time than the data size. L devices can have large impa proper consideration durin REFERENCES [1] What edge computing me
  33. 33. 04 Storage optimization and distribution
  34. 34. IoT application storage repositories • The modern methods for IoT applications delivery utilize virtualization as an efficient tool for software packaging. • This is usually done using virtual machine images (VMI) or containers. • The VMIs and containers are stored in either centralized or distributed storage repositories. 34
  35. 35. IoT application storage repositories • Unfortunately, these repositories are not optimized and are not able to transparently distribute VMIs based on the IoT application requirements and usage history. • Therefore, we present a novel multi-objective middleware for the management of VMIs in federated computing continuum repositories with replication support. 35
  36. 36. The ENTICE optimization middleware • The ENTICE middleware provides easy to use interface capable of receiving unmodified and functionally complete VM images from its users. • It can transparently distribute the VMIs to a specific infrastructure in a federation concerning their size, configuration, and geographical distribution, such that they are loaded, delivered, and executed faster and with improved QoS compared to their current behavior. 36 Multi-objective Middleware for Distributed VMI Repositories in Federated Cloud Environment 305 Fig. 5.1. Top level view of the Multi-objective Optimization Framework for VM Image distribution
  37. 37. The ENTICE middleware data flow 37 .1 VMI cost objective model the cost objective, we use the notion of the financial resources required to store particular VMI in a given repository site Cstorage and the t for transferring the data from the initial to the new repository in the federation Ctransfer. For each VMI, we calculate the cost objective by addin two components C = Cstorage + Ctransfer. URE 5 Virtual machine image redistribution framework flowchart
  38. 38. Evaluation (Repository optimization) 38 13 of 16 1 0.8 0.6 0.4 0.2 0 0 200 400 600 800 1000 1200 1400 1600 Value of replication coeficient − K Execution time (ms) hm execution time for different values of k hm execution time for different numbers of individual evaluations n framework with replication support has been evaluated on the bases of the quality of the Pareto optimal solutions 1 0.8 0.6 0.4 0.2 0 0 200 Value of replication coeficient − K FIGURE 8 Optimization algorithm execution time for different values of k FIGURE 9 Optimization algorithm execution time for different numbers of individual evaluations Furthermore, the optimization framework with replication support has been evaluated on the bases of the quality of the Pareto optimal s fordifferentvaluesofthegeneticparameters,suchasnumberofindividualsandnumberofevaluations/generations.Theassessmentwasco for a specific scenario enclosing the redistribution of 100 VMIs with replication coefficient k = 1. The hypervolume indicator24 has been compare the quality of the separate solutions provided during the experiments. Figure 10 presents a comparison of the mean value and deviation of the quality of the solutions in contrast to the number of utilized individuals and performed evaluations per experiment. By a the results, it can be concluded that in the cases when low computational time is required, it is essential to reduce both the number of in and performed evolutions, thus guaranteeing best computational performance to quality ratio. Contrary, when the computational time issue, like in the case of off-line VMI redistribution, higher populations sizes with increased number of evolutions should be used. For ex the case when 100 individuals are being utilized, we can expect best quality of the solutions when less then 5000 separate individual eva are performed. Above this threshold, the low number of individuals can easily lead to low diversity and reduced search capabilities, thus p solutions with limited hypervolume values. In the case of ENTICE, the redistribution and replication are performed off-line, thus requi population size and increased number of evolutions. T o better illustrate the difference in quality among the solutions, on Figure 11, we p comparison between the best solutions gathered from the experiments with 100 and 500 individuals. Lastly, in T able 3, we provide a comprehensive review of the exact objective values for the optimal trade-off distribution solutions calcu the optimization framework. Moreover, a comparison has been presented with a set of mapping solutions determined by using nonoptimize robin” mapping model for storing VMIs, and a multi-objective redistribution algorithm without replications support.17 The cost objecti been calculated based on the publicly provided price list for storing data in the Cloud by public companies, such as Amazon S3 and Microso The delivery time (performance) objective has been modeled based on the reported communication performance measures for 10 Gbit an KIMOVSKI ET AL. imal solutions in contrast with the number of individuals 14 of 16 KIMOVSKI ET AL. FIGURE 10 Quality of the optimal solutions in contrast with the number of individuals FIGURE 11 Visualized comparison of the best and worst solutions provided by the optimization framework TABLE 3 Comparison between different virtual machine image redistribution strategies Figure 7: Optimization algorithm scalability for different replication coefficients Figure 8: Optimization algorithm scalability with and without replication Figure 9: Quality of the repository optimization solutions Figure 10: Analysis of the optimization solutions
  39. 39. 39 Summary and future directions
  40. 40. Summary and future work • We proposed a new resource characterisation approach with pollutants emissions analysis. • We researched and developed a multi-objective scheduling approach with a Markov-chain based model for predicting when an edge device might move (roam) within the computing continuum. • We introduced a novel middleware for optimisation of the storage distribution of the IoT applications across the computing continuum. • In future we plan to merge the developed software in a single toolset for management of IoT applications across the computing continuum. 40
  41. 41. References 41 Research papers: 1. Dragi Kimovski, Sasko Ristov and Radu Prodan, “Decentralized Machine Learning for Intelligent Health Care Systems on the Computing Continuum”, IEEE Computer Magazine, 2022. 2. Dragi Kimovski, Josef Hammer, Narges Mehran, Hermann Hellwagner and Radu Prodan, “Cloud, Fog or Edge: Where to Compute”, IEEE Internet Computing Magazine, 2021. 3. Roland Matha, Dragi Kimovski, Anatoliy Zabrovsky, Christian Timmerer and Radu Prodan, “Where to Encode: A Performance Analysis of x86and Arm-based Amazon EC2 Instances”, eScience 2021. 4. Vincenzo De Maio and Dragi Kimovski, “Multi-objective scheduling of extreme data scientific workflows in Fog”, Future Generation Computer Systems, Volume 106, 2020, pp. 171-184. 5. Narges Mehran, Dragi Kimovski and Radu Prodan, “MAPO: A Multi-Objective Model for IoT Application Placement in a Fog Environment”, Proceedings of the 9th International Conference on the Internet of Things, 2019, pp. 1-8. 6. Dragi Kimovski, Narges Mehran, Christopher Kerth and Radu Prodan, , “Mobility-Aware IoT Application Placement in the Cloud -- Edge Continuum”, IEEE Transactions on Services Computing, 2021. 7. Dragi Kimovski, Julio Ortega, Andres Ortiz and Raul Banos, “Parallel alternatives for evolutionary multi-objective optimization in unsupervised feature selection”, Elsevier Expert Systems with Applications No.9/Vol.42, 2015 8. Dragi Kimovski, Julio Ortega, Andres Ortiz and Raul Banos, “Leveraging cooperation for parallel multi-objective feature selection in high-dimensional EEG data”, Concurrency and Computation: Practice and Experience No.18/Vol.27, 2015 9. Dragi Kimovski, Julio Ortega, Andres Ortiz and Raul Banos “Feature selection in high-dimensional EEG data by parallel multi- objective optimization”, 2014 IEEE International Conference on Cluster Computing, 2014 10. Nishant Saurabh, Dragi Kimovski, Francesco Gaetano and Radu Prodan, ”A Two-Stage Multi-Objective Optimization of Erasure Coding in Overlay Networks”, CCGrid 2017, Madrid, Spain 2017 11. Dragi Kimovski, Attila Marosi, Sandi Gec, Nishant Saurabh, Atilla Kertezs, Gabor Kecskemeti, Vlado Stankovski and Radu Prodan “Distributed Environment for Efficient Virtual Machine Image Management”, Concurrency and Computation: Practice and Experience 12. Sandi Gec, Dragi Kimovski, Uros Pascinski, Radu Prodan and Vlado Stankovski “Semantic approach for multi-objective optimisation of the ENTICE distributed Virtual Machine and container images repository”, Concurrency and Computation: Practice and Experience 13. Dragi Kimovski, Nishant Saurabh, Vlado Stankovski, Radu Prodan, “Multi-Objective Middleware for Distributed VMI Repositories in Federated Cloud Environment”, Scalable Computing: Practice and Experience, No.4/Vol.17, 2016 14. Dragi Kimovski, Nishant Saurabh, Sandi Gec, Polona Stefancic, Gabor Kecskemeti, Vlado Stankovski, Radu Prodan, Thomas Fahringer, “Towards Decentralized Repository Services for Efficient and Transparent Virtual Machine Operations: The ENTICE Approach”, IEEE International Conference on Cloud Networking, CloudNet 2016, Pisa, Italy, October 2016
  42. 42. Questions?

×