Future of cloud up presentation m_dawson
1 like341 views
The document discusses how exponential data growth is straining centralized cloud infrastructure and driving up costs due to lack of economies of scale. It argues that a more distributed and decentralized approach is needed to better manage and leverage the vast amount of unused capacity at the edge. This includes distributing data across devices, networks, and data centers instead of concentrating it within massive centralized data centers. A hybrid model is proposed that keeps some functions like policy centralized while pushing processing and storage out closer to where the data is created and used.
Future of cloud up presentation m_dawson
- 1. Future of Cloud: Will Dis-economies of Scale Break the Silver Lining Margaret Dawson, Vice President, Product Management
- 2. It’s all about the data 2
- 3. Exponential digital data growth . . . ➡ Every second, 58 Terabytes of digital data are being created ➡ 35 Zettabytes of digitally stored data by 2020 ➡ Storage managed by a storage administrator grew from 81TB in 2010 to 126TB in 2011 ➡ 15 out of 17 sectors in the US have more than 235 TB of data – same as the Library of Congress 33
- 4. And it’s all kinds of data Unstructured Structured 4 4
- 5. Even data about the data . . . “one person’s data is another person’s metadata” 5 5
- 6. And it’s all business or mission critical Not surprisingly, financial data comes in first, but nearly all data used by businesses is deemed critical Financial Data 70% Customer Records 65% Email, IM & other Communication 65% Productivity Files: Excel, PPT, Word, etc 57% Anything in our Database 56% Employee Personnel Information 48% Product Code or Other IP 27% Anything on Employee Laptops 24% CAD/CAM Designs or other Schematics 24% Patient Data or Medical Records 17% Photos, Images Video or Graphics 10% Other 5% Source: Symform Data Growth & Protection Research March 2012 6
- 7. But the Cloud will save us! 7 7
- 8. Well, not exactly . . . 8 8
- 9. One of the big problems Cloud is powered by massive centralized infrastructure 9
- 10. This other side of cloud is getting attention 10
- 11. Questions are starting to be asked ➡ How Green is Your Cloud? • Growing concentration of data centers in key locations is having significant impact on electricity grid & sources • Amazon, Apple & Microsoft are rapidly expanding without adequate regard to source of electricity use dirty energy • Yahoo & Google lead sector in prioritizing access to renewable energy • Facebook has now committed to power its platform with renewable energy • Akamai first company to report on its carbon intensity 11
- 12. The Cloud is driving record build out 122 1
- 13. Our current footprint is BIG 500,000+ data centers worldwide 700 million+ square feet Data Center costs of $1.44 Trillion 2% of North American energy consumption goes to servers and data centers 2% of global carbon footprint is data centers used to power cloud computing 13 13
- 14. And we are not slowing down . . . 10%+ new capacity is expected to be added every year for the next five years 14 14
- 15. So what? Isn’t that good news? Centralized infrastructure brings economy of scale 15 15
- 16. But we aren’t really seeing economy of scale Local storage and commodity servers might be cheap, but huge costs for: 2% 9% • Land • Bandwidth 7% • Construction • Maintenance • Power • Security 82% • Cooling • Environmental impact Land Core/Shell Architecture Mech/Elec 16 16
- 17. Example: the cloud storage model is broken 2TB Hard-drive for $100 but thousands for same data size in the cloud Amazon S3 Storage Storage Cost Bandwidth Cost Total Annual Cost 2TB $2820 $288 $3,108 Google Cloud Storage Storage Storage Cost Bandwidth Cost Total Annual Cost 2TB $2700 $288 $2,988 17
- 18. Cloud Security & SLA Lapses are Expensive Average cost of downtime per hour is $145,000 Forrester Research 18 18
- 19. And we can’t build data centers fast enough 35,000 Digital Data Created Data Center Capacity Available 60% gap 14,000 45% gap 8,000 35% gap 4,400 1,400 1,274 2010 2015 2020 Source: IDC Digital Universe Study, 2011 19 19
- 20. Even as we look to creative locations 20 20
- 21. Ironically, we are overbuilding for ―our‖ use We are oversizing by 5x than capacity at start-up and more than 1 ½ times ultimate actual capacity 21 Source: ―Avoiding Costs from Oversizing DC & Network Infrastructure,‖ Schneider Electric
- 22. And not utilizing what we already have Utilization of physical and power infrastructure is around 50–60% Many co-lo facilities, public cloud data centers and enterprise centers are woefully under used What is the cost of white space? 22
- 23. Same goes for the storage capacity We build capacity for peak usage Up to 80% of storage capacity is on average idle – Gartner $250 to power one server for one year - $62,500 to power idle servers if your percent idle was just 5%. 23
- 24. What’s the solution? 1. Improved Data Management 2. More Co-Lo / shared Facilities with higher density 3. Use the public / hybrid cloud when you can 4. Increase use of distributed systems 5. Move toward more decentralized approaches 24
- 25. Data Management Know Your Data . . . • What data exists and what is created • How much, how often, how long do you need to keep it? • Where is the data? • What is business/mission critical data? • Who needs access to it? and how do they get it today? • How long can the business function without access to the data? • Make a plan for improvement management • Data classification • SOA • New polices and access control rules 25 25
- 26. More Co-Lo / Less Build Out Do you really need your own data center? SuperNAP = Sustainability, Future Proofing of Design & Massive Scale SuperNAP density means One (1) SuperNAP equals at least three (3) other Data Center Facilities of equal floor space 26 26 www.switchlv.com
- 27. Co-Lo’s are staying ahead of the curve This is all they do, so the leaders: ➡ Best practices across security, availability, burst scalability, power efficiency, cooling efficiency, bandwidth optimization, etc. Source: Eaton 27 27
- 28. Use someone else’s cloud when you can 28
- 30. Internet is inherently distributed Internet Users by Country Billions of devices sitting on the edge Data distributed across devices, networks, data centers, and 30 geos
- 31. Cloud is driving distributed models & solutions ➡ Multi-tenant architecture ➡ Distributed application ➡ Distributed database ➡ Virtualization ➡ Multi-threading ➡ Multi-core CPUs ➡ Parallel processing ➡ Distributed file system ➡ Distributed Load Balancing ➡ RAID algorithms 31
- 32. And we’re realizing benefits ➡ System or process optimization ➡ Improved performance - faster ➡ Increased reliability & fault tolerance ➡ Lower costs ➡ Increased efficiencies ➡ Easier scalability or expansion ➡ Continuous or near continuous operations 32
- 33. But. . . distributed alone is not enough . . . Distributed only will not keep up with our data growth Still heavily based on centralized models with distributed components 33
- 34. Need to go beyond distributed to decentralized Why? Unused instances Over provisioning Under use of reserved instances Orphaned services Millions of dollars invested and wasted Contributing source: Mat Ellis 34
- 35. How to think about Decentralization ➡ No central bottleneck ➡ Power of large numbers ➡ Organic, demand driven growth in capacity ➡ Leverages existing infrastructure and devices on edge ➡ Shared information ➡ Concept of ―Contribution‖ to the community ➡ Assume everyone / every node is ―untrusted‖ ➡ Geographic spread of: • Ownership & participation • Costs • Management overhead • Risks 35 35
- 36. Decentralized development Open Source Movement Programmers who support the ➡ Linux – a Unix-Based operating system open source movement philosophy contribute to the open ➡ Apache — a leading server software and scripting language on the web source community by voluntarily writing and exchanging ➡ MySQL — a database management system programming code for software ➡ PHP — a widely used open source general- development. purpose scripting language ➡ Blender — a 3D graphics and animation software ➡ OpenOffice.org – an office suite software with word processor, spreadsheet, and presentation capabilities ➡ Mozilla — a web browser and e-mail client ➡ Perl — a programming/scripting language ➡ Wikipedia — Online encyclopedia open for anyone to update and revise content 36 36
- 38. Decentralized communications Skype is the largest telephone company in the world but has almost no centralized infrastructure 38
- 39. Decentralized cloud storage What if we aggregated all the unused capacity across servers, desktops and storage devices on the edge of the Internet to build a global storage network? 39
- 40. What Stay’s Centralized? Doesn’t mean loss of centralized control or IT power ➡ IT Policy and governance ➡ Shared Services ➡ Security and compliance ➡ High volume, low latency mandates transactions ➡ Definition of ―trust‖ ➡ Data analysis aggregation (search, e-discovery & ➡ Vendor evaluation reporting) guidelines ➡ Data warehousing ➡ API management ➡ Search 40 40
- 41. We are still in early stages Issues with decentralization ➡Desire for Control ➡Geo-political differences ➡Data security and encryption ➡Need more open API’s ➡Consistent Quality of Service 41 41
- 42. Look for opportunities To Leverage Distributed & Decentralized Models ➡ Assumption of ―untrusted‖ should be your security principle today ➡ Worry less about where the data is and on how to control access and gain intelligence about it ➡ Be the source of centralized policy and governance 42
- 43. Today’s Cloud: Centralized Data Centers Apple Amazon Google 43
- 44. Tomorrow’s Cloud: Decentralized Grids 44
- 45. Thank You [email protected] @seattledawson CloudNOW Cloud Connect Advisory Council Advisory Council www.symform.com
Editor's Notes
- #31: Reduce function: Chopping something into pieces: data, processes, applicationsDistribution function: Spreading those pieces across multiple systemsResource sharingParallel executions