How To Build A Stable And Robust Base For a “Cloud”
- 1. How To Build A Stable And Robust Base For a “Cloud” Doctor Cheng Liang Pengyun Network Technology
- 2. Historical Background and Cloud Trends • Overturn of traditional business models • The transactional businesses based on the Internet or mobile terminals have grown exponentially • The requirements for user experience are higher – Available at anytime – Smooth accesses – Comprehensive functions – Safe and Reliable
- 3. Imperative Cloud Transition • The challenges and opportunities brought by "Internet+" force the transition form traditional data centers to cloud computing • Objects hosted by the cloud platform: – New business system – Incremental IT – The transition of inventory IT • Hot spots of construction: private clouds and industrial clouds
- 4. Imperative Storage Platform Transition Cloud Computing Data Center Traditional Enterprise IT Cloud Storage Distributed Storage Software Universal PC Server and Built-in Storage + Traditional Storage
- 5. Traditional StorageVSCloud Storage Procurement Costs High-cost Dedicated Hardware Low-cost Universal Hardware The higher requirements, the more expensive Non-universal, poor resource utilization Bound to vendors lower cost, but higher capacity Universal hardware, high utilization, on-demand flow Not bound to hardware vendors, pro- competitive Capacity Expansion Costs High Input, Periodic Capacity Expansion Low Input, On-demand Capacity Expansion One-time high input, locked funds Expansion of old equipment, weak input-output ratio Bound to vendors, seller's market on-demand input, optimized allocation of funds Using mainstream hardware, reasonable input- output ratio Not hardware bound to vendors, buyer's market Operation and Maintenance Costs Dependent on Mechanical Operation and Maintenance Dependent on System's Intelligent Operation and MaintenanceUniversal pare parts, high cost Dedicated teams and technologies, high staffing cost Complex operation and maintenance, high input in professional services Universal spare parts, low cost No specific requirements, low staffing cost Simple operation and maintenance, no need of professional services Time Costs Sudden Construction, Pursue Opportunities Ready to Use, Meet Opportunities Establishment or expansion takes a very long period Complex resource allocation, slow Fast commissioning Fast resource allocation Traditional Storage Environments Cloud Storage
- 6. Cloud Storage is the Cloud Infrastructure • Data storage of video surveillance system
- 7. Cloud Storage is the Cloud Infrastructure • It stores the pictures and video in billions of mobile devices (mobile phones)
- 8. Cloud Storage Definition of Cloud Storage 应用实例 Software Defined Storage X86Storage Servers Cloud Storage External interfaces Distributed Block Storage Distributed Database Storage Object Storage
- 9. Cloud Storage is the Cloud Infrastructure • The construction of cloud software stacks needs to be based on cloud storage IaaS Cloud StorageCloud computing Software definition network PaaS Cloud Resource Scheduling Cloud Software Installation Cloud Resource Management SaaS Virtual Desktop Web Application Big Data Analysis Economical Hardware X86Servers, Network Switches, etc.
- 10. Cloud Storage is the Cloud Infrastructure • The construction of cloud software stacks needs to be based on cloud storage IaaS PaaSand SaaS Workflow Cloud Services Message Cloud Services Cloud Resource Management Virtual Desktop Web Application Big Data Analysis Software Defined Computing Distributed Block Storage Distributed Database Storage Object Storage
- 11. Characteristics of Cloud Storage Products • "Three Highs" – High Scalability – High Availability – High Performance • "One Intelligent" – Intelligent Self-Maintaining
- 12. Pengyun Reliable Intelligent Storage Distributed Block Storage Distributed File Storage Distributed Object Storage Distributed Database We are not porters We don't take "shortcuts" Based on Surpassing Starting from the Bottom We are engineers Project Formation
- 13. Pengyun ZettaStor Technology-leading Storage Products High Scalability Horizontal Scalability It can support millions of nodes ZB-level mass storage High Performance Parallel IO Processing Hard Disk IO Strategy Optimization Optimization of Network Paths High Availability Share-Nothing Architecture No single-point failure Failure self-healing, which doesn't affect the application
- 14. Cloud Storage How to Achieve Higher Scalability Fully Distributed Architecture Share-Nothing Architecture No Central Control Node Million-level Scalability Control Node Node Node Node Node Limited Scalability Difficult to Support Large-scale Environments Unified Central Control Nodes??
- 15. Cloud Storage How to Achieve Higher Performance User Block Data Object Layer Transition File System Layer Transition Physical Storage Location Multiple Transitions High Cost, High Latency Non-native Block Storage?? User Block Data Physical Storage Location Real Native Block Storage Direct Bare Disc Operation High Performance, Low Latency
- 16. Cloud Storage How to Achieve Higher Performance Multiple Data Forwarding High Latency, Low Performance Too Long Network Data Path? Optimized Network Data Path Direct Data Distribution Low Latency, High Performance
- 17. Cloud Storage How to Achieve Higher Availability Metadata User Data Copy 1 Copy 2 Copy 3 Metadata and User Data Distributed and stored in the system together With multiple copies at the same time After a certain copy is invalid, the system will generate a new copy on the other nodes
- 18. Cloud Storage How to Achieve Intelligent Maintenance-free Each Storage Node with Intelligence Perceive the health status of any other nodes Automatically work together with other nodes
- 19. Best Practices of Cloud Storage Architecture Share Nothing(purely distributed) architecture, no central control node High Scalability Native block storage, the bottom layer doesn't use file system or object storage Optimize network data path Disk IO optimization and others High Performance Eliminate single-point failure Intelligent failure self-healing, automatic data reconstruction High Availability. Each node is intelligent Work together with other nodes Intelligent Maintenanc e-free
- 20. 应用实例 Cloud Virtual Machines Database Applications NASSystems ZettaStor DBS Volume Generated ZettaStor DBS and the Managed Hardware ZettaStor DBS Application Scenario Support the Hyper-Converged Architecture Pengyun ZettaStor DBS Technology-leading ServerSAN Products
- 21. Pengyun ZettaStor DBS Technology-leading ServerSAN Products NAS Data Access Interface Integrated Interface RESTful API OpenStack Cinder Driver DFSiSCSIPYD
- 22. Pengyun ZettaStor's Customers and Partners B2B: Construct Private Clouds for Enterprises Customers' Industries: Operation, finance, government, military, medical, IT, scientific research, culture and entertainment, manufacturing, transportation, media, energy IaaS Cloud Storage Cloud computing Cloud Network PaaS Cloud Resource Scheduling Cloud Software Installation Cloud Resource Management SaaS Web Application Virtual Desktop Big Data Analysis Economical Hardware X86Servers, Network Switches, etc. · Our Products Block Device Storage OR ServerSAN Object Storage
- 24. Thank you for your supports! Please pay attention to more coming contents Software-defined Data Center Special Keynote Speech