Cloud essential

479 views

Published on

Cloud essential

Published in: Technology
0 Comments
2 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
479
On SlideShare
0
From Embeds
0
Number of Embeds
4
Actions
Shares
0
Downloads
0
Comments
0
Likes
2
Embeds 0
No embeds

No notes for slide

Cloud essential

  1. 1. Cloud Essential Tasuka at Gmail.ComTuesday, December 04, 2012 1
  2. 2. Cloud ComputingTuesday, December 04, 2012 2
  3. 3. Cloud Computing What is Cloud Computing from Youtube Cloud Computing Explained from Youtube The NIST Definition of Cloud Computing in PDF 什麼不是雲端運算 翟本喬 in PDFTuesday, December 04, 2012 3
  4. 4. NIST definition Cloud computing is a model for enabling convenient, on-demand network access to a shared pool configurable computing resources. On demand self service Rapid elasticity Broad network access Measured service Resource poolingTuesday, December 04, 2012 4
  5. 5. Cloud Computing From WikipediaTuesday, December 04, 2012 5
  6. 6. Cloud Computing Device Device Device Device Device Device Apps Apps Apps User Interface Service Service Service Service Service Service Software Software Virtual Platform Platform Platform Infrastructure Data CenterTuesday, December 04, 2012 6
  7. 7. Cloud Computing IaaS: Infrastructure as a Service PaaS: Platform as a Service SaaS: Software as a Service XaaS: X as a ServiceTuesday, December 04, 2012 7
  8. 8. Cloud Computing Private Cloud Community Cloud Public Cloud Hybrid CloudTuesday, December 04, 2012 8
  9. 9. Cloud Services Amazon AWS/EC2 Google Services/Chrome OS Dropbox Apple iCloud CHT HiCloudTuesday, December 04, 2012 9
  10. 10. Data CenterTuesday, December 04, 2012 10
  11. 11. Data Center Google Container Data Center Tour Google Data Center in Street View Dirty Cloud of Amazon, Apple, Microsoft Dirty CloudTuesday, December 04, 2012 11
  12. 12. Traditional Data Center • Different Hardware • Mixed Hot/Cold Aisle • Different Operation System • Simple Policy of Networking • Different Application Software • Install and Boot from Local Disk Driver • Single Point of Failure or RedundancyTuesday, December 04, 2012 12
  13. 13. Cloud Data Center • All Same Hardware • All Same Operation System • May run Software with Virtualization • Complex Policy of Netowkring • Boot from Network without Install • Easy to Move And Quickly Service When Fail • Easy to Extend for ScaleTuesday, December 04, 2012 13
  14. 14. Data Center Power Efficiency Networking Cooling Efficiency Facility Storage Space Space Efficiency Easy Management Computing Power Easy Maintenance Data Storage Space Easy to Recovery BandwidthTuesday, December 04, 2012 14
  15. 15. Power efficiency Generator loss Travel Line Loss Transformer Loss Circuit Loss Green power, Solar, Wind, or Others Power Recycle from Heat Water/AirTuesday, December 04, 2012 15
  16. 16. Cooling Efficiency Temperature Humidity Particle Cooling Water Room Temperature Water Free AirTuesday, December 04, 2012 16
  17. 17. Space Arrange More Host in a Rack Hot/Cool Aisle Separate Weather Consideration Location Power/Pipeline/Network managementTuesday, December 04, 2012 17
  18. 18. Computing Super Fast Single Computer Parallel Computing in Single Computer Distributed Computing in a Group Virtualization in Single ComputerTuesday, December 04, 2012 18
  19. 19. NodeTuesday, December 04, 2012 19
  20. 20. Node Hardware CPU Video Multi CPU, Multi Core, Multi Threads None for save power consumption RAM Network Interface More RAM for More Performance More Bandwidth and Weak on LAN Support Disk Cooling Fan More for Aggregate Space None is better or at least need to be control Serial Power Supply For Easier Maintenance High Efficiency with Higher Voltage BIOS Easier Setting and Quickly BootTuesday, December 04, 2012 20
  21. 21. Node Software Open Source is the First Choice Cheap and easier to use No Central Controller Prevent complex HA design for single point of failure No Specifically Hardware Easier Management and Maintenance No Dedicated Network Address Connect More Scaleability Services from Profile and Template No manual inactive Keep Simple of Software Let Operating System take care as it canTuesday, December 04, 2012 21
  22. 22. Monitoring CPU/Memory Load Networking Load HDD System TemperatureTuesday, December 04, 2012 22
  23. 23. Node Simple Hardware Complex Software Auto Negotiation and Share Information Between Nodes Self Management Easier MaintenanceTuesday, December 04, 2012 23
  24. 24. OS Boot Sequence • BIOS POST • Device Drive • Boot Loader • Kernel • Load Device Driver • Shell • Application ServiceTuesday, December 04, 2012 24
  25. 25. NetBoot With PXE Power On BIOS Hardware • PXELinux Initialize NetBoot? • DHCP With Option 82 N Y PXE NetBoot Local Boot Requiire • TFTP Load IP Address Require DHCP Server • RootFS In RAM Disk Boot Kernel TFTP Server Linux Kernel RAM Require RootFS Image RAM DISK TFTP Server RootFS Image Require • Data Center Services System Management DNS Server Maintenance NTP Server Init local HDD • OS Level Initialize scripts Join Distributed System Virtualization FileSystem Configuration Distributed FileSystem Bring Up Cloud Cloud Apps OS Cloud OS upTuesday, December 04, 2012 25
  26. 26. ComputingTuesday, December 04, 2012 26
  27. 27. Computing Desktop Computing Grid Computing Cluster Computing Parallel Computing Distributed Computing Cloud ComputingTuesday, December 04, 2012 27
  28. 28. Computing PVM: Parallel Virtual Machine MPI/MPICH: Message Passing Libraries ZeroMQ: Network Message Passing Library Real Time OS Tiny Computing Kernel Beowulf ClusterTuesday, December 04, 2012 28
  29. 29. StorageTuesday, December 04, 2012 29
  30. 30. Storage Access, Backup and Recovery Network Attached Storage (NAS/SAN) iSCSI/ATA over Ethernet Network Share Storage NFS, SMB, P2P Distributed Storage GlusterFS, MooseFS Object Storage OpenStack Swift, HDFS List of File Systems NFS/GlusterFS/MooseFS testTuesday, December 04, 2012 30
  31. 31. Gluster FS File Based Distributed File System No Single Point of Failure Distributed Meta-Data Required Fully-Mash Connection with TCP Still not Stable for Large Scale AccessTuesday, December 04, 2012 31
  32. 32. MooseFS File based Distributed File System High Loading in Mata-Data Server Has Single Point of Failure Seems more Stable Not Easy to ExpendTuesday, December 04, 2012 32
  33. 33. OpenStack Swift Object Storage Slow with Python Large Object Read/Write too Slow IssueTuesday, December 04, 2012 33
  34. 34. HDFS Object Storage Slow with JavaTuesday, December 04, 2012 34
  35. 35. RDRB/NFS/iSCSI Single Point of Failure Connection Limitation Bandwidth Limitation Not Real Distributed File SystemTuesday, December 04, 2012 35
  36. 36. BigData You might Has Collected a Lot of Data First What is your Data Means and Value That is Application StuffTuesday, December 04, 2012 36
  37. 37. NetworkingTuesday, December 04, 2012 37
  38. 38. Networking Bandwidth Bridging Routing MPLS Virtualization Networking Virtual Infrastructure Link MaintenanceTuesday, December 04, 2012 38
  39. 39. ManagementTuesday, December 04, 2012 39
  40. 40. OpenStack IaaS Python Computing/Object Storage/Image ServiceTuesday, December 04, 2012 40
  41. 41. Hadoop Cluster BigTable MapReduce Slow JavaTuesday, December 04, 2012 41
  42. 42. VirtualizationTuesday, December 04, 2012 42
  43. 43. Virtualization MS Windows Large Scale Easy to Recovery Large Size of File as HDD Image Complex Network ConfigurationTuesday, December 04, 2012 43
  44. 44. Xen Specific Kernel required Not Really Open SourceTuesday, December 04, 2012 44
  45. 45. QEMU Emulate x86/x86-64/PowerPC/ARM... Emulated Hardware Slow Speed Single Thread HyperVisor Support Any Guest OS include MS WindowsTuesday, December 04, 2012 45
  46. 46. KVM Only for Linux Specific Kernel Required Can work with QEMUTuesday, December 04, 2012 46
  47. 47. LXC Linux Host and Guest only Chroot like not really isolated Host and Guest Guest Speed as HostTuesday, December 04, 2012 47
  48. 48. OperationTuesday, December 04, 2012 48
  49. 49. Operations • Facility Maintenance • Hardware Maintenance • Service Software Maintenance • Network Topology Maintenance • Daily Check and Maintenance • Disaster Recovery • Security • AutomationTuesday, December 04, 2012 49
  50. 50. User Interfaces Graphic UI Web GUI Mobile UI Command Line UI Programming API Automation from Experience without UITuesday, December 04, 2012 50
  51. 51. SummaryTuesday, December 04, 2012 51
  52. 52. Next Step We live in a connected world, and modern software has to navigate this world. So the building blocks for tomorrow’s very largest solutions are connected and massively parallel. It’s not enough for code to be “STRONG and SILENT” any more. Code has to talk to code. Code has to be chatty, sociable, well- connected. Code has to run like the human brain, trillions of individual neurons firing off messages to each other, a massively parallel network with no central control, no single point of failure, yet able to solve immensely difficult problems. And it’s no accident that the future of code looks like the human brain, because the endpoint of every network are, at some level, human brains. From ZeroMQTuesday, December 04, 2012 52
  53. 53. Next Step 如何讓一個人可以維護上萬部機器 如何讓程式像人一樣的互相配合、互相合作 自我檢視、自我維護、自我協調、故障排除 故障預警 無中央集權式的管理Tuesday, December 04, 2012 53
  54. 54. 天下武功 熱情和視界才是重點 有熱 你以為你在創新, 其實是在代工 無堅不破 情但看不遠也沒有用 看 惟快不破 但沒有熱情持續努 力下去 也只會失敗 上班打卡制 你以為你在作 App,其 實你在做網頁 你以為 下班責任制 你在作雲端,其實你在 對於只有錘子的人, 作末端 你以為你在藍 海,其實你在紅海 他能解決的問題就是 從 Impossible 釘釘子 到 Im possible 理想勝過策略 不要隨便被新技術 和人 開放勝過封閉 們為新技術勾勒出的未來 簡單勝過複雜 吸引。相反,有時間的 話,要認真瞭解計算機的 公開勝過封鎖 歷史 Eat your own dog food! ScrumTuesday, December 04, 2012 54
  55. 55. How To Let Computers Works As One Stay Hungry Stay Foolish 熱情、眼界Tuesday, December 04, 2012 55
  56. 56. 延申閱讀 Steve Jobs: 2005 Stanford Commencement Xdite:給尚未大學畢業的朋友幾個人生建議 vgod:過早最佳化是萬惡的根源 就是有人跳下去會死啊 vgod:我為什麼要念博士 Book:學徒模式:優秀軟體開發者的養成之路 Teddy:搞笑談軟工Tuesday, December 04, 2012 56
  57. 57. 夏日大作戰 at Wikipedia Summer War Part 1 at Youtube Summer War Part 2 at YoutubeTuesday, December 04, 2012 57

×