Hyper-Scale Cloud Infrastructure Rick Bakken Senior Director, Datacenter Evangelism Cloud Infrastructure & Operations
Hyper-scale infrastructure strategy principles
Software-Defined
Availability Risk Management Fault assessment
Commodity Infrastructure Converge and standardize common resource usage
across workloads
Hyper Scalable
Dynamic Supply Chain
Resilient Services
Automation Optimize Infrastructure utilization and efficiency via standardization
Cost Model & Demand Forecasting
Global Operations
Generation 2 (SLA 99.999)
Generation 4 (SLA 99.9)
Generation 5 (SLA 99.9)
Traditional, physical redundancy N+2, Tier 3/4
Service geo-redundancy Active/active nodes – geo-
distributed
Service geo-resiliency
Lower Capex/Opex Increased efficiency Hyper-scale
Microsoft Cloud Infrastructure and Operations (MCIO)
Platform as a Service (PaaS)
Infrastructure as a Service (IaaS)
Consumer and Small Business
Services
Enterprise and Public Sector
Services Third-party
Hosted Services
Network Data Centers
Software as a Service (SaaS)
MCIO “engineering first” Deliver op5mal capacity, lowest cost, at any global loca5on
Drive opera5onal simplifica5on, eliminate waste, increase agility Con5nuous improvement and innova5on
ISO 27001
HIPAA / HITECH
European Union
Model Clauses
FedRAMP JAB P-ATO
PCI DSS
Level 1
SOC 1 Type 2
SOC 2 Type 2
Criminal Justice Information System
China Multi Layer Protection Scheme
FIPS 140-2
United Kingdom G-Cloud
21 CFR Part 11 Family Educational Rights and Privacy Act
Singapore Multi-Tier Cloud
Security
China CCCPP
F
ISO 27018
Australian Signals
Directorate
Cloud Controls Matrix
Content Delivery and Security Association
DISA Level 2
New Zealand Government Chief Information Office
United States
Regional
Enterprise / Industry
Center for Financial Industry Information
Systems
200+ Cloud Services 1+ billion customers · 20+ million businesses · 90+ markets worldwide
50+ billion minutes of connections
handled each month
18+ billion Active Directory authentications
per week
85% Fortune 500
users
250+ million
active users
2.4+ million emails per day
7+ billion worldwide queries each month
30+ trillion objects stored
48+ million users in
41 markets
50+ million active users
3.5+ million subscribers
400+ million active users
FY09 FY10 FY11 FY12 FY13 FY14 FY15
Network Device Count Growth
Geo-Redundant Service/Application Design • All nodes active
Top 2 Most Connected Networks in the World • Peer with over 3000 ISP’s globally
DC-to-Internet Backbone
• Multiple Terabits • Over 50 Points of Presence globally • Global backbone connecting MS Datacenter to
the Internet
DC-to-DC Backbone • Multiple Terabits of Capacity • Dark fiber based DC-DC backbone to enable
high bandwidth between Datacenters
Dark Fiber
• Tens of thousands of Route Miles of owned Dark Fiber Backbone
• Million+ 100G DWDM Route Miles of capacity deployed
Cache Node • Hosting Services collocated at User location (metro)
Edge Nodes • Multiple Terabits of Edge Interconnect capacity • Directly connected to more than 2000 networks
with over 4,000 connections
Decoupled DCs • Separation of Compute, Storage, DB Services
IT Capacity Unit = STAMP • DC Capacity Unit or Workload Appliance
Large L2 Domains
HW-based service modules
Simple Tree Design
L3 at all layers
Services in software
Clos-based design
L3
L2
LB/FW LB/FW LB/FW LB/FW
Low due to diversity and manual provisioning process
Low due to complex hardware and lack of automated operations
Low due to high complexity and human error
Resilient design, automated monitoring and remediation, minimum human involvement
Simplify requirements, optimize design, and unify infrastructure
Automated network provisioning, integrated process Agility
Efficiency
Availability
More than just server racks…
Traditional datacenter infrastructure
Source: EYP Mission Critical Facilities Inc., New York
Offline UPS technologies can drive Electrical losses substantially down
Widening temperature range can remove chillers and drive cooling to zero
Virtualization, active power management increase IT return on investment
Traditional Generation 4
PUE=1.8 PUE=1.15
Server Capacity 20 year Technology
2.0+ PUE
Colocation
Generation 1
Density Rack Density & Deployment Minimized Resource Impact
1.4 – 1.6 PUE
Generation 2
2012 2009 1989-2005 2007
Containment
1.2 – 1.5 PUE
Containers, PODs Scalability & Sustainability Air & Water Economization Differentiated SLAs
Generation 3
Modular
1.12 – 1.20 PUE
ITPACs & Colocations Reduced Carbon Right-Sized Faster Time-to-Market Outside Air Cooled
Generation 4
SW Defined
1.07 – 1.19 PUE
Fully Integrated Resilient Software Common Infrastructure Operational Simplicity Flexible & Scalable
Generation 5
2015
Internet users ■ 500,000,000+ ■ 100,000,000 – 499,999,999
■ 50,000,000 – 99,999,999 ■ 25,000,000 – 49,999,999
■ 5,000,000 – 24,999,999 ■ 100,000 – 4,999,999
■ 50,000 – 999,999 ■ 0 – 49,999
*Operated by 21Vianet
Microsoft global datacenter footprint Microsoft Azure datacenter regions
1 million+ servers • 100+ Datacenters in over 40 countries
Microsoft’s network is one of the two largest in the world Microsoft’s global datacenter footprint
In-rack fuel cell research • Natural gas converted directly to electricity to power servers • Wastewater treatment methane recovery pilot
Dramatic improvement in holistic efficiency • Beyond PUE – removes losses inherent in energy production and delivery • Efficient energy supply chain from source to motherboard
Increased datacenter reliability • Fewer moving parts, fewer potential points of failure. Increased global
commonality
Lower infrastructure costs • Elimination of electrical distribution, power conditioning, and back-up
infrastructure
Substa5on Substa5on Server
Datacenter
CPU
Server Datacenter
CPU
Radically simplified supply chain delivers more data with less resources
Energy innovation
Application Portfolio Analysis 1. Application
assessment 2. Retire Duplicate
workloads 3. Move SaaS to Cloud
• O365 • CRM • Power BI
4. Decipher Private/Public Mix
Hybrid Datacenter
Azure Stack Azure
Security +
Management Enterprise Mobility Suite
+ Operations Management Suite
Utility IT
Bill Presentment Cloud Appliances
IT Aligned to Business Optimized Supply Chain
Application Migration & BI SQL Server + Azure
Analytics
Developer Services (Visual Studio + VS
Online + Azure AppService +
Azure IoT Suite)
MicrosoB IT – Cloud migra/on strategy
Re5re it, right-‐size, eliminate environments
Use or convert to a SaaS (1st or 3rd party) solu5on
Expose func5onality in exis5ng SaaS/ PaaS solu5on
Convert to Azure PaaS solu5on
Op5mize for and move to Azure IaaS VM
No change liB ‘n shiB to IaaS
Remains on-‐prem
Azure IT Roadmap
~30%
First to move → Basic Web Apps → Advanced Portals → Any New solu5ons → Any re-‐architected solu5ons
Next to move → High I/O OLTP → Regulatory & High business
impact
Hard or costly to Move → HVA Systems → PKI Systems → Legacy Source Control
How to move from on premise to Microso< Cloud
…To → Office365 → SharePoint Online → CRM Online → VSO → ADL + PowerBI → Best 3rd Party SaaS
From…. Office Servers Portals and SPS Any rela5onship mgmt Ac5ve Source Control & WIT Data Warehouses Industry Std ver5cals
~10% ~5%
~15%
~35%
~5%