Cloud computing offers compelling business benefits, yet achieving scalability, performance and ROI is not slamdunk. You must think differently about your architecture, usage monitoring, scaling strategy, performance best practices, and high availability to maximize the ROI of cloud computing. In this Webinar, Forrester’s analyst Mike Gualtieri and Compuware CTO Imad Mouline will describe how to optimize cloud applications for your business, your users, and your checkbook.
Join Mike Gualtieri, Senior Analyst, Forrester Research, and Imad Mouline, CTO APM Solutions, Compuware to learn:
-The real ROI for cloud – scalability, reduce costs, increase revenue – and how ROI isn’t guaranteed if you ignore the new reality
-Four key qualities of cloud deployed applications to obtain ROI and quality end user experience
-Best practices to measure and ensure success with your cloud applications
19. Availability
• Monitor application and infrastructure
Monitored for expected and unexpected events
that can detect or foretell an outage.
• Mitigate the effect of known
Fault vulnerabilities by implementing
redundant components and the ability
tolerant to operate in a crippled mode.
• Identify and rehearse failure scenarios
and put in place a support regime to
Fixable quickly investigate, escalate, and
resolve problems.
24. Performance
• Application response time
meets or beats user
Fast expectations and business
requirements.
• Application response time is
consistent under normal and
Consistent load conditions of increasing
users, workload, and data.
36. “ROI in the Cloud – Myth, Magic, or Measureable?”
February 24, 2011
Imad Mouline - Compuware CTO, APM Solutions
37. Agenda
• ROI goals of moving to the cloud
– Business and technical objectives
– Compuware Gomez case study
• What we found
– Performance impacts ROI
– Cloud isn’t as simple as it seems
• How we solved those problems
– Optimization techniques
– Learned best practices
38. Web Applications Are Moving To The Cloud –
February 2011
Percentage of web app transactions that include at least one
object hosted on Amazon EC2 or Azure
Amazon EC2 Region Percentage
EC2 Asia Pacific - Singapore 0.185
EC2 US West - Northern California 3.460
EC2 EU - Ireland 5.286
EC2 US East - Northern Virginia 35.289
Azure AppFabric Region Percentage
Azure US South Central 8.669
Azure US North Central 4.389
Azure EU North 3.313
Azure EU West 0.062
Azure Asia Southeast 0.483
Azure Asia East 0.020
40. Best Practice: Define your goals and build a plan
Align goals across your organization If only there was a
button to push!
Why are we moving to the cloud?
Common goals include:
1. Trade CapEx for OpEx – can we just pay for just what we use?
2. Additional Capacity – How much capacity
do we need during normal and peak times?
3. Improved End-User Experience – What performance goals
are we trying to deliver against?
4. Greater Elasticity – How quickly can the provider we select
ramp up to meet our needs?
5. Flexible Bursting – How fast do we need to be able to access
additional capacity?
41. Case Study: Gomez Load Testing Network
• Our Goals
– Provide as much load as possible to test infrastructure/app
– Reduce overall costs to us
– Scalability (handle as many customers -- big & small -- as possible)
– Response time consistency
• Our challenge: Handling huge, unpredictable variations in demand
2 enterprises Options:
100K+ users
1M+ pages
5 enterprises
10K+ users
100K+ pages
42. Case Study: Gomez Load Testing Network
• Our Goals
– Provide as much load as possible to test infrastructure/app
– Reduce overall costs to us
– Scalability (handle as many customers -- big & small -- as possible)
– Response time consistency
• Our challenge: Handling huge, unpredictable variations in demand
Plan for peak for 100% of time
2 enterprises Options:
100K+ users
1M+ pages 1) Build massive data centers with
frequently under-used capacity
5 enterprises
10K+ users
100K+ pages
43. Case Study: Gomez Load Testing Network
• Our Goals
– Provide as much load as possible to test infrastructure/app
– Reduce overall costs to us
– Scalability (handle as many customers -- big & small -- as possible)
– Response time consistency
• Our challenge: Handling huge, unpredictable variations in demand
Plan for peak for 100% of time
2 enterprises Options:
100K+ users
1M+ pages 1) Build massive data centers with
frequently under-used capacity
2) Use “co-lo” facilities with
5 enterprises frequently under-used capacity
10K+ users
100K+ pages
44. Case Study: Gomez Load Testing Network
• Our Goals
– Provide as much load as possible to test infrastructure/app
– Reduce overall costs to us
– Scalability (handle as many customers -- big & small -- as possible)
– Response time consistency
• Our challenge: Handling huge, unpredictable variations in demand
2 enterprises Options:
100K+ users
1M+ pages 1) Build massive data centers with
frequently under-used capacity
2) Use “co-lo” facilities with
5 enterprises frequently under-used capacity
10K+ users
100K+ pages 3) Use the cloud and just pay for
what we need, when we need it
45. Our approach: Use the cloud
Simulate hundreds of thousands of concurrent users, or
millions of page views per hour by leveraging Gomez‟
network of public Cloud data centers
Easily create load tests
leveraging a global
network of cloud
datacenters
Run load tests with real IE and FF
browsers or choose from over 500
mobile devices
47. What we found:
Cloud Performance is a business issue …
Poor Performance = Missed ROI
48. What we found:
The Cloud Creates Availability and Performance Concerns
• The Cloud is opaque, shared and elastic
– Loss of visibility and control - traditional tools don’t apply
– It’s complex, multi-source and variable
– Others can affect my performance
– Capacity issues and inconsistent bursting capabilities
49. Geographic Latency – West Coast Provider
• Response time for sample transaction of reference application hosted on
Amazon EC2 US West (California), as measured from major US cities
50. Geographic Latency – East Coast Provider
• Response time for sample transaction of reference application hosted on
Amazon EC2 US East (Virginia), as measured from major US cities
51. What we found:
Cloud performance is highly variable
• Average response time of reference app transaction across various Cloud providers,
measured from several global locations
55. Our major concerns:
• Will we have all the
capacity needed when our
customers demand it?
• Are we in the right places
to provide a consistent
service our customers?
56. Addressing Our Concerns: Cloud Vendor Criteria
• Criteria
– Additional Capacity and Diversity of High Volume Locations
– Network of Cloud based Data Centers
• Generate enough load to test any customer site
• From enough locations to meet the needs of our global customer base
• Key Requirements:
1. API Based – to support self-service scheduling of tests for our
customers
2. Elastic Capacity – spin up servers when needed with our load
agent already installed and ready to go
3. Hourly Pricing – pay only for the capacity we use
4. Location Diversity – have cloud locations in new geographies
or in areas for heavy use
5. Capacity – have enough elastic capacity for the number of
customers we have and volume of concurrent load tests
57. Our solution: Use Multiple Cloud Providers
• Multiple providers
• Multiple regions Asia/Pac
• Multiple availability zones
Europe
North America
58. Our solution: Use Multiple Cloud Providers and
our own Last Mile Network
Simulate hundreds of thousands of concurrent users, or
millions of page views per hour by leveraging Gomez‟
network of public Cloud data centers and Last Mile Peers
Easily create load tests
leveraging a global
network of cloud
datacenters
Include peers from
dozens of countries to
gain insight into end user
experience under load
Run load tests with real IE and FF
browsers or choose from over 500
mobile devices
59. Pricing in the Cloud can be confusing
And it can cost you more than necessary
• Some vendors charge by
• CPU hour or RAM hour or both
• They all charge for bandwidth / transfer
• Some are free inbound, some aren't
• Some charge by volume, some by bandwidth
• Only the developer really knows
• How the application works
• What types of resources it requires
Build your applications or select a Cloud vendor with an eye
on optimizing their performance / cost ratio.
60. The Cloud Offers Unique Opportunities For
Optimization
• Domain sharding is a common performance optimization
• In traditional hosting environment, it came at a price
• The Cloud, with proper planning, can provide sharding for free
Potential bottleneck
Bucket
1
Bucket
2
61. The Cloud Offers Unique Opportunities For
Optimization
• Better performance, for little work, at no extra cost?
49 % Response Time Improvement
$0 additional cost
No reduction in capacity
62. Best Practice: Performance Testing Cloud Capabilities
Evaluate vendors based on your goals…
Capacity
• Test vendors to 15-20% past estimated capacity goals
Elasticity
• Baseline end-user performance before & after testing
• Test during pre-deployment and in production
• Ramp elasticity testing to peak levels
Burstability
• Isolate the cloud elements from other
infrastructure to test
• Test the “failover process”
63. Summary
ROI in the Cloud is real and it’s measurable
Be aware of the unique characteristics of the cloud
Architect your applications to leverage cloud characteristics
Consider usage pricing when designing your applications
Be paranoid about availability, performance and scalability.
www.compuware.com
Thank You! www.cloudsleuth.net