3. AWS is setting the standard…
as measured by capacity…
Total Number of Objects
Stored in Amazon S3
Every day through 2011, 762B
AWS added the same Peak Requests:
amount of server 500,000+
per second
processing capacity, on
average, that it took to run 262B
the Amazon online retailing 102B
operation in 2000, when it 2.9B 14B 40B
was a $2.76bn company.
Q4 2006 Q4 2007 Q4 2008 Q4 2009 Q4 2010 Q4 2011
Prickett-Morgan. “AWS Cloud Double Fluffs in 2011.” The Source: UBS
Register, 6 Jan 2012.
4. …data center footprint and geographic
distribution…
…the company said that
with the opening of its AWS
data center in São
Paulo, Brazil in mid-
December, the company
has doubled its AWS data-
center footprint. AWS Regions
Amazon Edge Locations (CloudFront & Route 53)
Prickett-Morgan. “AWS Cloud Double Fluffs in 2011.” The
Register, 6 Jan 2012.
5. …and, most importantly, revenue…
$M
Amazon Web Services Revenue Model
$1,400
It has been estimated $1,200 All Other
that AWS could be a $1 $1,000
billion business for the $800
online retailer come $600
next year…could hit $400
$2.5B in 2014. $200
$0
Hickey, Andrew. “Amazon Q3 Cloud Revenue 2006e 2007e 2008e 2009e 2010e 2011e
Skyrockets” CRN. 26 Oct 2011.
Source: UBS
6. How did Amazon build its Cloud?
Amazon eCommerce Platform
AWS API (EC2, S3, …)
Amazon Proprietary Orchestration Software
Open Source Xen Hypervisor
Networking Servers Storage
7. How can we build a cloud using CloudStack?
Amazon eCommerce Platform
User Portal
AWS API (EC2, S3, …)
Cloud API “Cloud OS or
Data Center
Amazon Proprietary CloudStack Software
Apache Orchestration OS”
VMware KVM OVM XenServer Hyper-V Bare-Metal
Open Source Xen Hypervisor
Networking Servers Storage
8. The Virtual Datacenter OS allows businesses to efficiently
pool all types of hardware resources - servers, storage and
network – into an aggregated on-premise cloud
VMware press release Sept 15, 2008
9. Eucalyptus is the only cloud architecture to support the
same application programming interfaces (APIs) as public
clouds, and today Eucalyptus is fully compatible with the
Amazon AWS public cloud infrastructure.
Eucalyptus Systems Press Release April 2009
12. 2.0 AWS 2.2 3.0 Quality 4.X
Prototype 1.0 GA
Refactor Compatibility Refactor Improvements Refactor
2008 2009 2010 2011 2012
Sept 2008: Nov 2009: May 2010: July 2011: April 2012:
VMOps CloudStack Cloud.com Citrix Apache
Founded 1.0 GA Launch & Acquires CloudStack
CloudStack Cloud.com
2.0 GA
13. 2.0 AWS 2.2 3.0 Quality 4.X
Prototype 1.0 GA
Refactor Compatibility Refactor Improvements Refactor
2008 2009 2010 2011 2012
• Initial target: hosting companies like Rackspace and Savvis
• 3 engineers built a fully functional prototype in 5 months
• Use the demo to sell to early customers
(ReliaCloud, CloudCentral, 1800hosting.com, Go Daddy, etc.)
14. 2.0 AWS 2.2 3.0 Quality 4.X
Prototype 1.0 GA
Refactor Compatibility Refactor Improvements Refactor
2008 2009 2010 2011 2012
• Took 6 more months to make 1.0 software production ready
• Deployed on 5 production customers
15. 2.0 AWS 2.2 3.0 Quality 4.X
Prototype 1.0 GA
Refactor Compatibility Refactor Improvements Refactor
2008 2009 2010 2011 2012
• Product first, architecture second
• From web hosting to enterprise workload
• Multi-hypervisor, SAN, and VLAN support
• Learn needs of enterprise workload from: Tata Communications, Korea
Telecom, Macquarie Telecom
• Competition: vCloud Express
16. 2.0 AWS 2.2 3.0 Quality 4.X
Prototype 1.0 GA
Refactor Compatibility Refactor Improvements Refactor
2008 2009 2010 2011 2012
• Private cloud demand picked up
• Zynga wanted private cloud
• Support Amazon-style flat networking and security groups
• Competition: Eucalyptus
17. 2.0 AWS 2.2 3.0 Quality 4.X
Prototype 1.0 GA
Refactor Compatibility Refactor Improvements Refactor
2008 2009 2010 2011 2012
• Second major refactoring of CloudStack code
• Network-as-a-service combing both Amazon and traditional style networking
• More flexible orchestration engine
18. 2.0 AWS 2.2 3.0 Quality 4.X
Prototype 1.0 GA
Refactor Compatibility Refactor Improvements Refactor
2008 2009 2010 2011 2012
• Citrix acquisition
• Rapid growth of CloudStack user base
• Quality is more important than new features
19. 2.0 AWS 2.2 3.0 Quality 4.X
Prototype 1.0 GA
Refactor Compatibility Refactor Improvements Refactor
2008 2009 2010 2011 2012
• Third major refactoring of CloudStack code
• Apache contribution drive rapid growth of CloudStack developer base
• Apache license compliance
• Services framework
• Hadoop integration
23. • Server failure comes from:
8%
ᵒ70% - hard disk
ᵒ - RAID controller
6%
ᵒ - memory
5%
ᵒ18% - other factors
• Application can still fail for
Annual Failure Rate of servers other reasons:
ᵒNetwork failure
ᵒSoftware bugs
Kashi Venkatesh Vishwanath and ᵒHuman admin error
Nachiappan Nagappan, Characterizing
Cloud Computing Hardware
Reliability, SoCC’10
#CitrixSynergy
23
24. Internet
Core Routers
… Access Routers
Aggregation Switches
Load Balancers
… Top of Rack Switches
Servers
25. • Bugs in failover
40%
Effectiveness of network
redundancy in reducing failures
mechanism
• Incorrect configuration
• Protocol issues such
as TCP back-
off, timeouts, and
Phillipa Gill, Navendu Jain & Nachiappan
spanning tree
Nagappan, Understanding Network Failures
in Data Centers: Measurement, Analysis
reconfiguration
and Implications, SIGCOMM 2011
#CitrixSynergy
25
26. Cloud workloads
Traditional-Style Amazon-Style
Reliable hardware, backup entire Tell users to expect failure.
cloud, and restore for users when Users to build apps that can
failure happens withstand infrastructure failure
Link Aggregation VM Backup/Snapshots
Storage Multi-pathing Ephemeral Resources
VM HA, Fault Tolerance Chaos Monkey
VM Live Migration Multi-site Redundancy
27. Designing a zone for a traditional workload
Hypervisor
Traditional-Style Availability Zone
vSphere or XenServer Enterprise
vCenter/XenCenter
Storage
Enterprise Networking (e.g., VLAN) SAN
Networking
Hypervisor Hypervisor Hypervisor L2 VLANs
Cluster Cluster Cluster
Network Services
Enterprise Storage (e.g., SAN) Load Balancing VPN
Multi-tier Apps
Multi-tier VLANs OVF
28. Designing a zone for an Amazon-style workload
Amazon-Style Availability Zone
Software Defined Networks Hypervisor
(e.g., Security Groups, EIP, ELB,...) XenServer Advanced
Server Server Server Server Storage
Racks Racks Racks Racks
Local EBS Object store
Server Server Server Server
Networking
Racks Racks Racks Racks
L3 SDN based L2 Elastic IP
Server Server Server Server
Network Services
Racks Racks Racks Racks
Security Groups ELB GSLB
Elastic Block Storage Multi-tier Apps
3rd Party Tools
(e.g., RightScale, enStratus CloudFormation
)
29. CloudStack can Support Both Styles
Apache CloudStack
Traditional Traditional
AWS-style AWS-style AWS-style Style Style
Availability Availability Availability Availability Availability
Zone Zone Zone Zone Zone
31. 146 Companies 238 Developers Global User Groups
Service Providers 100’s of Production Clouds
32,000 Community Members
Enterprises Universities
32. Apache CloudStack community projects
• SDN • Smart Storage
ᵒNicira ᵒHadoop + S3 API for object store
ᵒMidokura ᵒNetApp (FlexPod, object store)
ᵒBig Switch Networks ᵒBasho RIAK CS
ᵒStratosphere ᵒCaringo object store
ᵒCloudian S3
• Backup/DR
ᵒSungard • PaaS
ᵒCloudFoundry implementation through
• Networking IronFoundry and Stackato teams
ᵒCisco (VXLAN, Nexus) ᵒEngine Yard
ᵒBrocade (ADX) ᵒCumulogic
ᵒGigaSpaces
33. “The Apache Way”
• Collaborative software development
• Commercial-friendly standard license
• Consistently high quality software
• Respectful, honest, technical-based interaction
• Faithful implementation of standards
• Security as a mandatory feature