Advanced Cloud Infrastructures

Transcription

1 Advanced Cloud Infrastructures From Data Centers to Fog Computing (part 1) Guillaume Pierre Master 2 CCS & SIF, 2017 Advanced Cloud Infrastructures 1 / 35

2 Advanced Cloud Infrastructures 2 / 35

3 Advanced Cloud Infrastructures 3 / 35

4 Table of Contents 1 Cloud Data Centers 2 Cloud-based applications are changing 3 Edge Computing 4 Fog Computing 5 Research question: what is the smallest, cheapest machine which can still be used as a fog computing node? 6 Virtualizing a Raspberry Pi Advanced Cloud Infrastructures Cloud Data Centers 4 / 35

5 Traditional system architectures Application Middleware OS Machine Traditional architecture Advanced Cloud Infrastructures Cloud Data Centers 5 / 35

6 Traditional system architectures Application Middleware OS Machine Traditional architecture Advanced Cloud Infrastructures Cloud Data Centers 5 / 35

7 The varying capacity problem Demand Compute capacity Time Advanced Cloud Infrastructures Cloud Data Centers 6 / 35

8 The varying capacity problem Demand Necessary capacity Compute capacity Time Advanced Cloud Infrastructures Cloud Data Centers 6 / 35

9 The varying capacity problem Demand Necessary capacity Compute capacity Unused capacity Time Advanced Cloud Infrastructures Cloud Data Centers 6 / 35

10 The varying capacity problem Demand Necessary capacity Compute capacity Unused capacity Time What if demand increases beyond the capacity? Advanced Cloud Infrastructures Cloud Data Centers 6 / 35

11 Cloud Computing Di cult to vary capacity! Resources available on demand Manual resource Resource management is fully management automated Pay only for what you use Application Middleware Infrastructure as a Service OS Virtualization Machine Machine Machine++OS OS Machine + OS Traditional architecture Advanced Cloud Infrastructures Cloud architecture Cloud Data Centers 7 / 35

12 Cloud Computing Applications & Services Application PaaS Map Reduce Spark Storm... Middleware OS Machine Traditional architecture Infrastructure as a Service Virtualization Machine + Machine OS + OS OS Cloud architecture Advanced Cloud Infrastructures Cloud Data Centers 7 / 35

13 Table of Contents 1 Cloud Data Centers 2 Cloud-based applications are changing 3 Edge Computing 4 Fog Computing 5 Research question: what is the smallest, cheapest machine which can still be used as a fog computing node? 6 Virtualizing a Raspberry Pi Advanced Cloud Infrastructures Cloud-based applications are changing 8 / 35

14 The new mobile computing landscape Since 2016: mobile network trac > xed trac Advanced Cloud Infrastructures Cloud-based applications are changing 9 / 35

15 New types of mobile applications Interactive applications require ultra-low network latencies E.g., augmented reality require end-to-end delays under 20 ms But latencies to the closest data center are ms using wired networks, up to ms on 4G mobile networks!!! Advanced Cloud Infrastructures Cloud-based applications are changing 10 / 35

16 The new IoT landscape Fact: IoT-generated trac grows faster than the Internet backbone capacity Advanced Cloud Infrastructures Cloud-based applications are changing 11 / 35

17 New types of mobile applications Throughput-oriented applications require local computations E.g., distributed videosurveillance is relevant only close to the cameras Why waste long-distance network resources? Advanced Cloud Infrastructures Cloud-based applications are changing 12 / 35

18 Question Who owns computing resources located closest to the mobile end users and the IoT devices? Advanced Cloud Infrastructures Cloud-based applications are changing 13 / 35

19 Question Who owns computing resources located closest to the mobile end users and the IoT devices? Answer: Mobile network operators Advanced Cloud Infrastructures Cloud-based applications are changing 13 / 35

20 Table of Contents 1 Cloud Data Centers 2 Cloud-based applications are changing 3 Edge Computing 4 Fog Computing 5 Research question: what is the smallest, cheapest machine which can still be used as a fog computing node? 6 Virtualizing a Raspberry Pi Advanced Cloud Infrastructures Edge Computing 14 / 35

21 Mobile network operator's revenues Where does mobile network operator's revenues come from? I I I I 1990's: 2000's: 2010's: What's Voice (not any more) Text/SMS (not any more) Data (not for very long... ) next? Advanced Cloud Infrastructures Edge Computing 15 / 35

22 Mobile network operator's revenues Where does mobile network operator's revenues come from? I I I I 1990's: 2000's: 2010's: What's Voice (not any more) Text/SMS (not any more) Data (not for very long... ) next? Services! Let's steal part of the cloud computing business... I No cloud data center can be closer to the users than us! Advanced Cloud Infrastructures Edge Computing 15 / 35

23 Edge computing Before: After: Advanced Cloud Infrastructures Edge Computing 16 / 35

24 Standardization eorts European Telecommunications Standards Institute: Mobile Multi-Access Edge Computing Fujitsu, Hewlett-Packard, Huawei, Intel, Juniper, Motorola, NEC,Nokia, Orange, Samsung, Sony, Vodafone,... Focus: integration within mobile phone networks Advanced Cloud Infrastructures Edge Computing 17 / 35

25 Standardization eorts European Telecommunications Standards Institute: Mobile Multi-Access Edge Computing Fujitsu, Hewlett-Packard, Huawei, Intel, Juniper, Motorola, NEC,Nokia, Orange, Samsung, Sony, Vodafone,... Focus: integration within mobile phone networks Open Edge Computing Initiative Focus: openness Intel, Huawei, Vodafone, Carnegie-Mellon university,... Advanced Cloud Infrastructures Edge Computing 17 / 35

26 Standardization eorts European Telecommunications Standards Institute: Mobile Multi-Access Edge Computing Fujitsu, Hewlett-Packard, Huawei, Intel, Juniper, Motorola, NEC,Nokia, Orange, Samsung, Sony, Vodafone,... Focus: integration within mobile phone networks Open Edge Computing Initiative Focus: openness Intel, Huawei, Vodafone, Carnegie-Mellon university,... Open Fog Consortium ARM, Cisco, Dell, Intel, Microsoft,... Advanced Cloud Infrastructures Edge Computing 17 / 35

27 Standardization eorts European Telecommunications Standards Institute: Mobile Multi-Access Edge Computing Fujitsu, Hewlett-Packard, Huawei, Intel, Juniper, Motorola, NEC,Nokia, Orange, Samsung, Sony, Vodafone,... Focus: integration within mobile phone networks Open Edge Computing Initiative Focus: openness Intel, Huawei, Vodafone, Carnegie-Mellon university,... Open Fog Consortium ARM, Cisco, Dell, Intel, Microsoft,... Fog/Edge/Massively Distributed Clouds WG at OpenStack Advanced Cloud Infrastructures Edge Computing 17 / 35

28 Table of Contents 1 Cloud Data Centers 2 Cloud-based applications are changing 3 Edge Computing 4 Fog Computing 5 Research question: what is the smallest, cheapest machine which can still be used as a fog computing node? 6 Virtualizing a Raspberry Pi Advanced Cloud Infrastructures Fog Computing 18 / 35

29 What is fog computing? Fog computing = cloud + edge + end-user devices as a single execution platform Low latency Localized trac Less global trac Better reliability Privacy... Advanced Cloud Infrastructures Fog Computing 19 / 35

30 Fog Cloud We need cloud servers close to the users, but the users are everywhere (and they are mobile) Let's place one cloud server within short range of any end-user device Fog computing resources will need to be distributed in thousands of locations Advanced Cloud Infrastructures Fog Computing 20 / 35

31 Fog Cloud We need cloud servers close to the users, but the users are everywhere (and they are mobile) Let's place one cloud server within short range of any end-user device Fog computing resources will need to be distributed in thousands of locations Fog nodes will be connected with commodity networks Forget your multi-gbps data center networks! Advanced Cloud Infrastructures Fog Computing 20 / 35

32 Fog Cloud We need cloud servers close to the users, but the users are everywhere (and they are mobile) Let's place one cloud server within short range of any end-user device Fog computing resources will need to be distributed in thousands of locations Fog nodes will be connected with commodity networks Forget your multi-gbps data center networks! Fog nodes must be small, cheap, easy to deploy and to replace Advanced Cloud Infrastructures Fog Computing 20 / 35

33 Fog Cloud Typical Cloud platform Few data centers Lots of resources per data center High-performance networks Cloud resource location is (mostly) irrelevant Advanced Cloud Infrastructures Fog Computing 21 / 35

34 Fog Cloud Typical Cloud platform Few data centers Lots of resources per data center High-performance networks Cloud resource location is (mostly) irrelevant Typical Fog platform Huge number of points-of-presence Few resources per PoP Commodity networks Fog resource location is extremely important Advanced Cloud Infrastructures Fog Computing 21 / 35

35 Table of Contents 1 Cloud Data Centers 2 Cloud-based applications are changing 3 Edge Computing 4 Fog Computing 5 Research question: what is the smallest, cheapest machine which can still be used as a fog computing node? 6 Virtualizing a Raspberry Pi Advanced Cloud Infrastructures Research question: what is the smallest, cheapest machine which can still be used as a fog computing node? 22 / 35

36 Raspberry PIs are more powerful than you may think RPi3 Pine A64+ HP 820 G2 CPU (s) Memory (MB/s) Network (Mb/s) Storage (MB/s) Power when idle (W) Power under load (W) Price 92 e 74 e 1600 e We chose RPi3 as our base platform for the time being Mostly because of easy purchase options and community support... Advanced Cloud Infrastructures Research question: what is the smallest, cheapest machine which can still be used as a fog computing node? 23 / 35

37 What about a real application? Input: live video stream Processing: face recognition algorithm running inside Apache Flink Advanced Cloud Infrastructures Research question: what is the smallest, cheapest machine which can still be used as a fog computing node? 24 / 35

38 What about a real application? Input: live video stream Processing: face recognition algorithm running inside Apache Flink The RPIs are only 3-5 times slower than my laptop But they are 17 times cheaper If my applications scale horizontally I can use as many RPIs as necessary Advanced Cloud Infrastructures Research question: what is the smallest, cheapest machine which can still be used as a fog computing node? 24 / 35

39 Table of Contents 1 Cloud Data Centers 2 Cloud-based applications are changing 3 Edge Computing 4 Fog Computing 5 Research question: what is the smallest, cheapest machine which can still be used as a fog computing node? 6 Virtualizing a Raspberry Pi Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 25 / 35

40 Virtualization App 1 App 2... App 3 Micro-architecture Digital logic layer (memory, compute circuits) Operating system Assembler layer Bus Peripherals (disk, screen, network, keyboard...) "The hardware machine" Traditional machine architecture: Applications Operating system Hardware Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 26 / 35

41 Virtualization Assembler layer Micro-architecture Digital logic layer (memory, compute circuits) Operating system Assembler layer Bus Peripherals (disk, screen, network, keyboard...) "The hardware machine" Let's create a special application which behaves exactly the same as the assembler layer... Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 26 / 35

42 Virtualization Virtual machine App 1 App 2... Guest operating system Hypervisor (assembler layer) Micro-architecture Digital logic layer (memory, compute circuits) Host operating system Assembler layer Bus Peripherals (disk, screen, network, keyboard...) "The hardware machine" We can execute any operating system on top of it and any application over the guest operating system We have a virtual machine Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 26 / 35

43 Virtualization 2 cores, 6 GB RAM 4 cores, 8 GB RAM Virtual machine 1 Virtual machine 2 App 1 App 2... Guest operating system Hypervisor Micro-architecture Digital logic layer (memory, compute circuits) Host operating system Assembler layer App 1 App 2... Guest operating system Hypervisor Bus Peripherals (disk, screen, network, keyboard...) "The hardware machine" 8 cores, 16 GB RAM We can run multiple virtual machines on the same physical machine: Each virtual machine runs in full isolation from the other VMs Each virtual machine owns a subset of the hardware resources of the physical machine Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 26 / 35

44 Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 27 / 35

45 Why is virtualization interesting for cloud providers? Isolation: I can create multiple VMs on the same machine and give each VM to a dierent user (they will not see nor interfere with each other) Customization: Each user can customize their VMs according to their own requirements. Consolidation: Few applications can really exploit a large server machine to its maximum capacity. With virtualization I can split this capacity in smaller units and thereby increase my resource utilization. Management: Virtualization simplies resource management: I can measure how many resources each user is using, migrate VMs from one host to another, etc. Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 28 / 35

46 Virtualization technologies Virtualization technologies are now totally mainstream: Commercial: VMware, Microsoft App-V,... Open-Source: KVM, VirtualBox, Xen,... Paravirtualization vs. full virtualization: Paravirtualization works on any hardware platform but it requires special support in the guest OS. Slow! Full virtualization exploits special features of modern CPUs, does not require special support in the guest OS. Faster! Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 29 / 35

47 Virtualization technologies Virtualization technologies are now totally mainstream: Commercial: VMware, Microsoft App-V,... Open-Source: KVM, VirtualBox, Xen,... Paravirtualization vs. full virtualization: Paravirtualization works on any hardware platform but it requires special support in the guest OS. Slow! Full virtualization exploits special features of modern CPUs, does not require special support in the guest OS. Faster! Virtual machines are waaaaaayyyyyy too heavyweight for a RPI Each guest OS needs lots of memory Each OS needs to execute lots of background stu Impossible to run 100+ VMs on a single machine... Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 29 / 35

48 Containers App 1 App 2... App 3 Micro-architecture Digital logic layer (memory, compute circuits) Operating system Assembler layer Bus Peripherals (disk, screen, network, keyboard...) "The hardware machine" Traditional machine architecture: Applications Operating system Hardware Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 30 / 35

49 Containers Container 1 Container 2 App 1 App 2 App 3... Micro-architecture Digital logic layer (memory, compute circuits) Operating system Assembler layer Bus Peripherals (disk, screen, network, keyboard...) "The hardware machine" Let's create groups of processes which belong together: Process groups are totally isolated from each other Each process group belongs to a single user Each process group has its own hardware resource limits (CPU, RAM,... ) Each process group has its own network access policy Etc. We have containers Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 30 / 35

50 Container technologies Containers are an operating system feature No need for special CPU support Fully supported in Linux, Windows,... We usually add an extra software layer to simplify management: Docker Containers are less customizable than VMs Container owners cannot choose their OS But was that really necessary in the rst place? Not always. Containers are much more lightweight than VMs No need to run lots of (mostly identical) operating systems next to each other Containers often start in less than 1 second We can easily run hundreds of containers on a mid-sized machine Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 31 / 35

51 LXC containers Performance Agility Performance is extremely close to bare-metal Starting 150 containers w/ Apache Total time: 36 s (240 ms/container) Consumes about 2% of CPU Memory usage: 10 MB/container Stopping 150 containers: Total time: 9 seconds Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 32 / 35

52 Docker is a shipping container system for code docker-the-future-of-virtualization-for-your-django-web-development/ Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 33 / 35

53 Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 34 / 35

54 Teaser Next week we'll see that building a real fog platform is not as easy as it looks... Advanced Cloud Infrastructures Virtualizing a Raspberry Pi 35 / 35