Перспективы будущих вычисленийОблака, как они есть

Editor's Notes

• #3: Every day new services and devices enter the market making the internet more accessible and more valuable in terms of information, productivity and reach. We’re seeing more users, more devices and more content causing an explosion of growth and along with it an avalanche of complexity for IT that is trying to manage it. More users Today ¼ of the world is connected to the internet – about 1.5 billion people. Over the next 5 years we’ll add another billion users by extending access more deeply into emerging countries, lowering the cost of network access and client devices and through the creation of new services that appeal to the young and old incentivizing them to join the internet revolution More Devices Many people today in develop countries have a computer and home or at work and even have cell phones capable of internet access and email. But new devices are entering the market that will put not only new types or devices but more devices into each home in developed and undeveloped countries. For example – we’ll see in-car entertainment, home appliances, and TVs all connecting to the internet. More people, using more devices will drive to over 10 billion connected devices by 2015 – 2.5x today’s rate. More Content It’s not just more content, it’s more complex content. The rise of 3D content and video content will also accelerate the growth of content on the web requiring 60 Exabytes (a billion billion)/per second data stored, 800 Exabytes/sec of peak IP traffic and 20x the compute capacity we have today. Today’s infrastructure will rapidly be exceeded if we don’t find a more energy efficient, scalable and cost effective way to handle the growth. The answer – the Cloud which is poised to be the next fundamental shift in IT.
• #4: Rapid growth in virtualization has delivered many benefits to the enterprise; Server consolidation is helping reduce capital expenditures Rapid deployment of new services and virtual servers New capabilities for disaster recovery and high-availability At the same time, however, virtualization is creating new complexity challenges; VM sprawl causes numerous VMs much be patched and managed Network complexity is increases with many virtual networks and VM migration Despite advances in consolidation, data center real estate continues to be a premium—current assessment is ~$1000 / square foot to build a new data center At the same time new public cloud services are emerging promising cost effective alternatives to data center deployment, but with unclear SLAs, regulatory compliance, and impact on corporate intellectual property—exciting, but no panacea Finally IT organizations continue to operate under continued scrutiny to reduce OPEX Key Message: the data center needs to evolve to meet these growing challenges
• #5: Cloud computing holds the promise of addressing several IT pain points and challenges. Cloud computing is an evolution of IT that delivers IT resources in a flexible, pay as you go model. From an end user perspective, cloud-based services need to provide the illusion of infinite resources, sense of ownership and security and dictate a pay-as-you model. Both internal (behind corporate firewall) and external (via public internet) clouds offer the potential for highly flexible computing and storage resources, provisioned on demand, at theoretically lower cost than buying, provisioning, and maintaining more fixed equivalent capacity. The core element that makes something a “cloud” is the architecture of the underlying compute, network and storage in the datacenter. Cloud architecture - cloud services & data reside in shared, dynamically scalable resource pools, often times being virtualized (eg Amazon EC2), and/or based on scale out applications (eg Google search). Cloud computing provides a services delivery framework which is a combination of business processes and associated data that when linked in a certain way, provide a LOB function. The basic concept is no more complex than Lego building blocks. If one Lego block represents a business process and related data, there are any number of ways the blocks can be assembled in order to provide some kind of shape. Fundamentally, this is what a service delivery framework is. You can assemble the blocks (services) in any number of ways to provide the shape (business process) you are seeking. What is nice about the Lego example, is that, for the most part, all Lego blocks are a standard size and framework….this helps support the Open Data Center discussion that comes later. Different enterprise stakeholders have different expectations of what cloud should do. For the enterprise, Cloud Computing is primarily a means of providing IT various avenues to enhance business growth, flexibility and value back to the larger enterprise.
• #6: Intel believes the cloud of 2015 will look dramatically different from today’s computing infrastructure. The Cloud will be Automated: Dynamically allocate resources to agreed upon services levels and automatically optimize for maximum resource utilization and power efficiency. This includes automation of provisioning, resource monitoring, reporting of consumption for bill back and workload balancing. Federated: data and services will seamlessly and securely scale beyond borders to span public and private clouds when desired. Federation allows you to move your workload from one service provider to another, burst between your internal private cloud and public cloud providers if additional capacity is needed and more easily share information across vendors, partners or clients. Client aware: the data center will provide secure access and optimal experience across a range of devices to take advantage of the capabilities of the device in hand. Today many internet services are “dumbed down” to the lowest common denominator – even if the user is accessing the service via a powerful desk top computer. Or the opposite can be true where services are difficult to read/use if you access them from a device that isn’t a personal computer (e.g. have you ever tried to track a flight on your blackberry?). In a client aware environment, the capabilities of both the server and the client will be exposed to the cloud service. The cloud service will then be able to adapt based upon that information and deliver optimal service based on the device at hand – fully utilizing the power of the client/server when available. Some of these attribute exist today, but many are not fully developed, complete software capabilities do not exist and standards are missing that will enable businesses to achieve the full realization and potential of the cloud
• #7: To support the evolution to cloud to be on a path to achieve Intel’s cloud vision, there is a need to evolve cloud infrastructure to be more secure, efficient, and simplified, built on a foundation of openness. Efficient: We will deliver products that lead in energy efficiency and look not just at the component level, but board, system, rack and data center as well to drive efficiency, power savings and cost savings for enterprise and service providers. By increasing the efficiency of our products, we can help the industry avoid the build out of 45 new coal plants, thus saving enough power to support 30 million homes. (Power savings calculated based on projected performance improvements from Intel roadmap while keeping power / system flat. Moore’s Law drives ~2x perf / 18 months. At 5 years, that equals 10X. We assume that compared to 2010, we’re saving 9X (i.e, the 10x less the 1X for what you’ll need). It assumes we keep power per system constant at 200W. We assume we’ll need 16M servers in 2015 (based on market model) – that means we save 16M x 9X x 200W (average system power) x 1.6 PUE = ~45GW. The estimated power/coal plant is 1Gw 45GW = ~ 45 coal plants needed which equates to approximately 30 million homes according to the EPA website: http://www.epa.gov/RDEE/energy-resources/refs.html#coalplant Secure: We will build security into our products and technologies to assist with protecting data when it is in the data center and when it is on the move. VM’s will be protected when they launch and as they move from server to server and data center to data center. By enhancing the secure processing features of our platforms, Intel expects to help reduce already-large impact from the rapidly growing cybersecurity threats. For example, the Privacy Rights Clearinghouse chronicles and reports that over 345 million records containing sensitive personal information were involved in security breaches in the U.S. since January 2005.1. By enabling more efficient and robust encryption, users will be better protected should they be among this vast and growing tide of lost records. And Intel technologies that help reduce the malware that is key to many platform breaches stand to reduce the growing cost impact to cybercrime, which a recent PwC study2 quantified as having doubled from 2008 to 2010 to £10B in the UK alone—but is a worldwide challenge for businesses of all sizes. Intel estimating $10’s of billions in costs annually.   1 Privacy Rights Clearinghouse, A Chronology of Data Breaches, at http://www.privacyrights.org/ar/ChronDataBreaches.htm . 2 http://www.computerweekly.com/Articles/2010/04/28/241083/Cost-of-cybercrime-to-businesses-doubles.htm Simplified: Simplified: Through the power and flexibility of Xeon, we will enable solutions that will have flexibility across compute, storage and networking. We will also drive the convergence of network and storage traffic to a unified network based on Ethernet. Through a unified network alone, we estimate we could save 1 billion miles of network cabling. (Average of 8+ cables / virtualized server today. Assume 54M servers in installed base (saving 7 cables / server). Assume average length = 4m. 1604 meters / mile) There is an underlying foundation to each of these 3 principles of openness built on standards that enables interoperability, flexibility and choice. . No single company or companies should control, limit or own the cloud. Few standards exist today to make this happen, but we need the industry to align on a core set of standards and to advocate for products/services that are interoperable to keep the cloud open as it evolves.
• #8: Slide Purpose: Highlight the Energy Efficiency enhancements that Xeon® 5600 processors are bringing to the leadership Xeon® 5500-based platforms that were launched just one year ago. Four main points to how Intel is continuing to drive lower overall power consumption: Intel’s new 32nm Nehalem micro-architecture, which allows us to deliver 50% more cores and cache, and enables us to produce even lower power CPU SKUs – down to 40W for a 4C SKU. The proof point is 40% greater perf/watt. Xeon® 5600 still features Intel® Intelligent Power Technology (IPT), which automatically shifts the CPU and memory into the lowest available power state, while delivering the performance you need We’ve optimized the Power Control Unit (PCU) algorithms within the processor, specifically the ones that affect Turbo Boost and memory power management, to further reduce power consumption throughout the entire range of system performance levels. Memory power mgmt improvement:  In memory controller in WSM-EP, there is now CKEs (clock enable slow) in addition to the already existing CKEf (clock enable fast).  CKEs allows the DIMM to go into a lower power state.  Note:  CKEs only works with 1 DPC.  With > 1DPC, CKEf is used.  In addition, there have been improvement in the memory power management algorithms in WSM-EP. The Intel Xeon processor 5600 series also supports lower power memory, which could save up to 2W per DIMM, or almost 25W if 12 DIMMs are populated. Refresher on IPT features: Integrated Power Gates: Allows idling cores to be reduced to near-zero power independent of other cores, reducing server idle power consumption. Automated Low-Power States: Automatically puts processor, memory, and I/O controller into the lowest available power states that will meet the current workload while minimizing performance impact. Formerly known as DBS.
• #9: Intel Intelligent Power Node Manager was designed with the needs of hyper-scale customers in mind - particularly Public Cloud or Internet Portal Data Center (IPDC) segments as well as High Performance Computing Environments. Reducing power and cooling costs in the datacenter is a key challenge and requirement for these large scale datacenters. We estimate that as much as 25% of the total cost of operating an internet datacenter is the cost of power and cooling. We are delivering industry leading products and technologies to reduce power consumption and drive significant costs savings. Intel is leading in the SPECpower benchmark with 10 of top 10 SPECpower results using Intel based systems (refer to spec.org website for latest results). In addition, we have delivered a suite of technologies and software capabilities to enable industry leading power management and instrumentation in internet datacenters. Hyper-scale customers want the ability to be able to monitor power, reduce power and also cap power consumption to improve density and utilization of their datacenters. We have enabled the industry with Intel Intelligent Node Manager which is a system level technology, implemented via a combination of platform power supplies, Intel’s manageability engine (ME) embedded in the chipset and firmware, that delivers power reporting and power capping functionality for individual servers. We also offer Intel Data Center Manager (DCM) which enables management consoles to aggregate and manage logical and virtual groups of servers. DCM is a software SDK to allow OEM consoles and cloud IVS/OE providers a license to aggregate data from MEs on multiple servers and do policy based power management at the server, rack and data center level.
• #10: There is broadening customer and ecosystem support for Intel datacenter power instrumentation, with >30 partners acknowledging momentum for Intel Power Management tools including Intel® Intelligent Power Node Manager and Intel® Datacenter Manager which support Intel Xeon 5500 and 5600 series Nehalem generation of servers. The combination of server level and aggregated capabilities enables intelligent and directed management of power within the data center, and can be used to increase performance density, shed load during power & thermal events and provide the power data required for power based dynamic resource balancing. The industry likes this approach because it’s standards based and enables interoperability across multiple Intel Node Manager compliant systems. The OEM and ODMs listed above have already or plan to release Intel Node Manager compliant systems this year. The following ISVs/systems providers have licensed Intel Data Center Manager: Supermicro, ZZNode, Dragon Flow, Power Leader and Inspur. As you can see here from the >30 partners on this slide, there is growing momentum in the industry with a broad range of industry partners and customers acknowledging support and benefits for these technologies.
• #11: Slide Purpose: The vision and role Intel plays with datacenter security. We want to become the foundation of data center security by providing the VMM independent platform Si, firmware, and architecture enhancements to deliver confidentiality and security policy enforcement for mobile VMs. The near term is focused on basic platform hardening and efficiency. Our focus is on making platforms more resistant to attacks from things such as malware, and more efficient for security workloads such as encryption. AES-NI and Intel TXT are the first steps in accomplishing this goal. Introduce AES-NI and Intel TXT, making the point that AES-NI can provide clear IT benefits today, while Intel TXT is a way to help future-proof your datacenter, as Intel is working closely with the ecosystem to deliver robust TXT solutions starting later in 2010 and into 2011. AES-NI: Data encryption is not new, but new AES instructions from Intel are. In the past, data encryption either required custom hardware, such as security appliances and HDDs, or took away from CPU performance to manage. But with Intel® AES (Advanced Encryption Standard) Technology, you get robust encryption without needing additional appliances or increased performance overhead. So what is AES-NI: AESNI is a set of new instructions for enhancing the performance for cryptography using the widely-accepted Advanced Encryption Standard (AES) algorithm. There are 7 new instructions in the Westmere processor that target some of the more complex and compute-expensive encryption, decryption, key expansion and multiplication steps (and there are multiple steps in every instance of working with encrypted data) that in case the performance and efficiency of these operations. But note that the instructions do not implement the entire AES algorithm in silicon—only the most processor intensive elements have been targeted. This provides more flexibility and balance between HW performance and SW extensibility. Another benefit of the new instructions is that actually helps protect the data better as well. The use of the more efficient steps enabled in AES-NI makes the use of “side channel” snooping attacks. These attacks use SW agents to analyse how a system processes data and searches for cache and memory access patterns to try to gather patterns or other system data to help deduce elements of the cryptographic processing—and therefore make it easier to “crack”. AES-NI helps hide critical elements such as table lookups, making it harder to determine what elements of crypto processing are happening. Intel TXT:   Intel® Trusted Execution Technology (Intel® TXT) addresses a critical security need for all server deployments, especially virtualized and cloud-based use models by helping protect your server prior to OS launch or hypervisor launch. Intel TXT complements other malware protections such as anti-virus and intrusion detection to help ensure that only trusted software is on the platform. VMs on trusted platforms are also protected, so you can easily migrate them onto other trusted platforms or create pools of platforms with trusted hypervisors. We’ll talk more about Intel TXT in a minute…
• #12: Speaker notes Datacenters continue to experience new demands from their customers: scalability on demand, pressure to deliver “IT as a service”, & lowest possible TCO This creates challenges in manageability, security, flexibility, & affordability. So IT managers are looking for simplification. Fortunately a major technology transition is underway which will simplify IT managers’ lives: the convergence of Compute, Networking, & Storage. Compute: Intel’s heritage is compute performance, but with the Nehalem uArch, we brought intelligence to adapt to workloads w/ dynamic features like Turbo. Storage: Standard building blocks are transforming the Storage market – this puts more emphasis on responsiveness of the storage compute engine. In storage, we expect Intel architecture to drive 7 out of every 10 external storage systems shipped by end of 2010 Industry leaders like EMC have chosen the Xeon architecture of choice for storage solutions w/ their recent announcement of EMC Symmetrix solutions. Networking: The industry continues to drive bandwidth up and latency down via ethernet & the world is moving to converged fabrics. Many people don’t know that Intel is the world’s leading supplier of LAN connections – we shipped over a half billion connections over past 10 years. Intel is committed to lead the transition to converged networks w/ our leadership LAN products & technologies. Close: So Xeon is the cornerstone of next-generation Intelligent Data Centers.
• #13: Intel has led the creation of the Intel Cloud Builder Program with key eco-system partners to enable the industry and end users to more easily build cloud solutions in their data center. There are 3 key areas of focus - Deployment Best Practices: When a company has a choice to build a private cloud or to create a cloud within a service provider, the hardware and software infrastructure trade-offs for cloud management can be daunting. Provides public cloud service providers, hosters, telco service providers and private enterprises with reference architecture white papers to simplify the design and operation of a cloud. The reference architectures combine the latest Intel-based servers with leading ISV cloud operating environment software. These reference architectures provide a starting point from which to build clouds and over time will highlight the use of software by Intel and other industry players to address needs such as trusted multi-tenancy. The majority of major cloud software management ISVs are collaborating with Intel in this program. More details on Intel Cloud Builder and references architectures - http://www.intel.com/cloudbuilder   Building Optimized Clouds: When a company wants to assure that their cloud infrastructure is the most efficient and performant, a wide variety of technologies and solutions can be applied to achieve that end. Intel is working with large scale service providers to help build out optimized clouds via technology proof of concept tests using Intel’s latest products and technologies, such as Intel Xeon processor 5500 series and Intel Node Manager for power management. Intel is producing white papers, case studies, and related content derived from these POCs that are made available as part of the Intel Cloud Builder Program to help guide end users in building their own optimized clouds and take advantage of Intel technologies. In addition, IT@Intel continues to deliver best practice documents based on Intel IT’s own cloud strategy and implementation. Learn more at http://www.intel.com/it   Advanced Cloud Research: Open Cirrus is an open cloud-computing research testbed, jointly formed by Intel, HP and Yahoo with more than 10 test beds worldwide, and designed to support research into the design, provisioning, and management of services at a global, multi-datacenter scale. The open nature of the testbed is designed to encourage research into all aspects of service and datacenter management. In addition, we hope to foster a collaborative community around the testbed, providing ways to share tools, lessons and best practices, and ways to benchmark and compare alternative approaches to service management at datacenter scale. To learn more about Open Cirrus – http://opencirrus.org Intel is also an affiliate member of UC Berkeley’s RAD Lab, a research group focused on leveraging powerful techniques from statistical machine learning (SML), as well as recent insights from networking and distributed systems, to enable an individual developer to rapidly turn a prototype or idea into a robust distributed service. Findings and insights from some of these projects are published. Q&A Q: Who are the target users of the Intel Cloud Builder Program? A: The program will help telecommunications companies, hosters, Internet services providers and enterprise IT professionals who wish to deploy clouds. The program also aids ISVs that want to offer their applications through the cloud. Q: Which ISVs are participating in the Intel Cloud Builder Program to deliver reference architectures? A: The initial participants include seven ISVs; Canonical, Citrix, Microsoft, Parallels, Redhat, Univa and VMware. The Xen.org community is also involved. More are joining the program over time. Q: When will the first reference architectures be publicly available from Intel Cloud Builder Program? A: Several are already posted. More reference architecture papers are being posted in 2H and beyond. Go to http://www.intel.com/cloudbuilder to learn more. Q: What content is included in the reference architecture white papers? A: The reference architecture papers provide an in depth description of the Intel test bed blueprint, detailed discussion on the ISV software solution used in the test bed, technical review of the specific cloud use cases, setup and configuration information, and key things to consider in implementing the specific reference architecture. The purpose of these papers is provide a starting point for end users to be able to deploy a cloud solution on Intel hardware using the ISV software stack.
• #14: To support the evolution to cloud to be on a path to achieve Intel’s cloud vision, there is a need to evolve cloud infrastructure to be more secure, efficient, and simplified, built on a foundation of openness. Efficient: We will deliver products that lead in energy efficiency and look not just at the component level, but board, system, rack and data center as well to drive efficiency, power savings and cost savings for enterprise and service providers. By increasing the efficiency of our products, we can help the industry avoid the build out of 45 new coal plants, thus saving enough power to support 30 million homes. (Power savings calculated based on projected performance improvements from Intel roadmap while keeping power / system flat. Moore’s Law drives ~2x perf / 18 months. At 5 years, that equals 10X. We assume that compared to 2010, we’re saving 9X (i.e, the 10x less the 1X for what you’ll need). It assumes we keep power per system constant at 200W. We assume we’ll need 16M servers in 2015 (based on market model) – that means we save 16M x 9X x 200W (average system power) x 1.6 PUE = ~45GW. The estimated power/coal plant is 1Gw 45GW = ~ 45 coal plants needed which equates to approximately 30 million homes according to the EPA website: http://www.epa.gov/RDEE/energy-resources/refs.html#coalplant Secure: We will build security into our products and technologies to assist with protecting data when it is in the data center and when it is on the move. VM’s will be protected when they launch and as they move from server to server and data center to data center. By enhancing the secure processing features of our platforms, Intel expects to help reduce already-large impact from the rapidly growing cybersecurity threats. For example, the Privacy Rights Clearinghouse chronicles and reports that over 345 million records containing sensitive personal information were involved in security breaches in the U.S. since January 2005.1. By enabling more efficient and robust encryption, users will be better protected should they be among this vast and growing tide of lost records. And Intel technologies that help reduce the malware that is key to many platform breaches stand to reduce the growing cost impact to cybercrime, which a recent PwC study2 quantified as having doubled from 2008 to 2010 to £10B in the UK alone—but is a worldwide challenge for businesses of all sizes. Intel estimating $10’s of billions in costs annually.   1 Privacy Rights Clearinghouse, A Chronology of Data Breaches, at http://www.privacyrights.org/ar/ChronDataBreaches.htm . 2 http://www.computerweekly.com/Articles/2010/04/28/241083/Cost-of-cybercrime-to-businesses-doubles.htm Simplified: Simplified: Through the power and flexibility of Xeon, we will enable solutions that will have flexibility across compute, storage and networking. We will also drive the convergence of network and storage traffic to a unified network based on Ethernet. Through a unified network alone, we estimate we could save 1 billion miles of network cabling. (Average of 8+ cables / virtualized server today. Assume 54M servers in installed base (saving 7 cables / server). Assume average length = 4m. 1604 meters / mile) There is an underlying foundation to each of these 3 principles of openness built on standards that enables interoperability, flexibility and choice. . No single company or companies should control, limit or own the cloud. Few standards exist today to make this happen, but we need the industry to align on a core set of standards and to advocate for products/services that are interoperable to keep the cloud open as it evolves.