Emerging Technology for the Next Decade Cloud Computing Keynote Presented by Charles Liang, President & CEO Super Micro Computer, Inc.
What is Cloud Computing? Cloud computing is Internet-based computing, whereby users shared software, information, servers, storage and other infrastructure ondemand like electricity. In cloud computing, users order (rent) virtual servers, storage and bandwidth in much lower cost, flexible and dynamic way instead of buying physical HW, software, bandwidth and other resources. Source: Wiki
The Big Picture Servers Software The Cloud Storage Internet & Bandwidth Clients 4B users by 2020
Cloud Ecosystem Cloud (Resources) Server, Storage, Software X aas Bandwidth (Delivery) Networks, Switching Clients (Growth) Apps, Mobile, $
Technology Trend From physical to virtual Technology transitions into web-based computing Grid computing in 2000 is the direct parent of cloud Xen and Vmware in 2007 enabled virtual Mosaic Graphics Source: Intel
Public Cloud vs. Private Cloud Public Cloud (Service Provider Datacenter) A canvas for cloud applications/web services Server resources are dynamically provisioned on a self-service basis over the Internet Typically pay-per-use services Private Cloud (VPN cloud) Cloud computing on private networks (for certain users instead of everyone) Better security and control Higher overhead cost Internal IT is necessary
Overall IT Landscape
Cloud Product/Service Growth
Cloud Market s Growing Fast Global Cloud Service Market (Gartner) $68.3B (2010) to $148.8B (2014) Public Cloud Server Revenue (IDC) 518M (2009) to $718M (2014) Private Cloud Server Revenue (IDC) $7.3B (2009) to $11.8B (2014) ~62% growth
Cloud Layers Software as a Service (SaaS) Platform as a Service (PaaS) Infrastructure as a Service (IaaS) Hardware Infrastructure & Domain Abstraction Compute Network Storage
SaaS Solutions Appian Anywhere & AppNexus BPM (Business Process Management) application done in cloud Hadoop A scalable distributed data processing system GoGrid and other cloud hosting services Pay-per-use computing space Google, Yahoo, etc. Search engine services Facebook, Twitter, etc. Social networking and gaming services More
PaaS Solutions Amazon Elastic Computing Cloud & S3 Scalable web-computing for developers Google AppEngine Host web apps on Google infrastructure Microsoft Windows Azure Cloud services operating system SalesForce.com Give developers a platform to cloud apps
IaaS Expectations Build and Maintain Register (Rent) and Use IT Designs Applications Build Hardware (Knowledge) Services Ready App already developed (User Choices) Application Support Compatibility and Test (Time) Deployment done virtually (Less Time) Licenses and Compliance (Money) Pay per Use (Less Wasted Capital)
Cloud Businesses Cloud is mainly composed of 3 business models today Cloud Computing Processes, Data & Services Cloud Bandwidth Switching Transcoding* Transrating** Cloud Storage Capacity Security Management *Changing format **Changing Scale
Virtualization for Cloud Workload consolidation Increase server utilization Reduce capital (20%), hardware management (15%), power (20%), cooling costs (15%) Software flexibility & cost New and Legacy OS support Lower software cost Instant provisioning Easily create new virtual machines Easily reallocate resources (processor cores, memory, IO) between running virtual machines Manageability Predicted hardware downtime Workload balancing Cloud Migration Private Cloud Public Cloud Cloud Bursting Source: VMware
Evolution of HPC Adoption Scope of sharing Enterprise HPC / Internal Cloud Cluster-to-cluster sharing management Reliable file transfer & staging Enterprise Utility Grid / Cloud Virtualization of services Dynamic service provisioning On Demand, Utility SaaS, SOA Internet Data Centers Powered by xsps Distributed Clusters 1990 2015 Today Time
HPC Cloud Advantages On-demand access to highly scalable, high-performance clusters optimized for parallel computing (at 20% infrastructure cost---save money) Services and HPC expertise available Access from anywhere in the world with high-speed data connectivity HPC cloud workflow Again pay on demand..
Optimizing Cloud Infrastructure Optimized hardware for virtual systems for use in the cloud Multicore architecture Larger Memory/storage Capacity High performance I/O Modularized computing Cost Effective Servers Avg. 3 VMs per computing core 2-4 GB Memory / VM Instance Networking with Virtual Support Performance uptime guarantee Full redundancy on system level (replication) Failover servers on different power feeds power subsystems cooling subsystems, etc. Blade System
Optimizing Cloud Infrastructure Faster connectivity with load balancing Enable applications to load seamlessly regardless of location (InfiniBand, 10/40GbE, FCoE, etc.) Low Latency 10GbE Switch Networking with virtual support watch Open Flow architecture for switching! System Scalability enable cloud providers to increase performance and storage with less cost Easy to replicate servers and storage Enable standard platform: X86 to reduce the cost of adapting applications into cloud environment High Density Storage
Optimizing Power Efficiency 100% Power Supply Efficiency Digital PSU - 95% peak 94% 93% Efficiency 83% 67% 50% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loading Industry Average 1U Twin SuperBlade / 2U Twin² 90% efficiency at 20% load, 94% efficiency at 50% load, and 91% efficiency at 100% load. Save hundreds of dollars per node Software Power Management NMView (Node Management View) allows users to monitor and control the power consumption of managed servers in a datacenter
Cloud Energy Costs Energy costs are the fastest-rising cost element in a datacenter A single x86 server from a data center is responsible for > $600 a year in energy costs (100K systems cost up to $60M/yr.) Insufficient attention to the process of measuring, monitoring and modeling energy use in datacenters. High efficiency servers and storage are necessary to keep the overall operating cost down Source: IDC (March 2009) Overall IT Costs Demands to make systems operate at higher temperatures (water or free-air cooling methodology become possible)
The Green Cloud Savings $$$ High efficiency systems in cloud computing could help alleviate the world s biggest environmental challenges while driving profitable growth for users Each 93%+ high efficiency server can save up to ~$100 USD per node per year in energy Worldwide adoption of high-efficiency solutions could save the world ~$2.4 billion USD per year in electricity or ~4.8 billion trees for the sequestration of CO2* Grow a virtual forest through green cloud *Calculation Source: CIA
Cloud Advantages Save Costs No Initial setup costs HW, SW, licensing, real estate, etc. Pay only for what/when you use Dynamic ownership Easy & fast deployment Requires less in-house IT staff Ease of maintenance Economy of Scale
Cloud Advantages Highly Scalable and Flexible Anytime/anywhere access to data Enhanced reliability and security Inherent Failover redundancy Optimized IT resource usage Multi-tenancy IT Centralization Matching Supply with customer Demand Peak load capacity improvement No loss business opportunities Enables new SaaS and business models Green for the Environment Computing utilization improvement (up to 5X) Hardware power consumption optimized Example of matching computing supply and demand Example of higher server resource usage Source: Intel IT
Other Cloud Examples Notable Government Public Clouds
Healthcare Cloud Other Cloud Example
ISP Cloud Other Cloud Examples
Privacy & Security Cloud Concerns Cloud providers need to establish trust Many still believed that data is more secure when managed internally Compliance & Standard Incomplete compliance guidelines for cloud computing No open standards established for cloud computing yet Availability and Performance Varied levels of availability and performance of applications hosted in the cloud Cloud provider shutting down (applications run in multiple clouds beyond cloud-burst to thunderstorm)
Next Steps Centralization of IT resources through cloud Demand-based IT resource via cloud solution providers Open Standard to enable interoperability Move freely between hosted clouds Address cloud security concerns Cloud providers to address different customer security requirements Flexibility & portable Encryption Performance and manageability for cloud environment Hardware optimization (efficient computing) tailored for cloud
Summary Opportunities for everyone HW, SW, Service Servers, Storage and Infrastructure Need collaboration to further enhance cloud & capture the opportunities The limit is the SKY Thank You!