SlideShare a Scribd company logo
CDN Monitoring: 

From Users to Edge to Origin
Young Xu, Product Marketing Analyst
2
About ThousandEyes
ThousandEyes delivers visibility into every network your organization relies on.
Founded by network
experts; strong
investor backing
Relied on for "
critical operations by
leading enterprises
Recognized as "
an innovative "
new approach
31 Fortune 500
5 top 5 SaaS Companies
4 top 6 US Banks
3
CDN Architecture from User to Edge to Origin
Access 
Networks
Origin 
Server
Edge 
Server
Edge 
Server
Users
Internet
4
Methods of Monitoring Your CDNs
Users
Access 
Networks
Internet
Origin 
Server
Edge 
Server
Edge 
Server
Origin to Edge
3
 Cloud Agent to Edge
2
Cloud Agent to Origin
1
Cloud 
Agent
Cloud 
Agent
5
•  Benchmark CDN performance on latency and fetch time
–  Before and after object/page is CDN-enabled
–  Ongoing with user to edge versus user to origin
•  Ensure the proper edge is serving the request
–  Geolocation and GSLB vs. Anycast 
–  Manage multiple CDNs
•  Cache issues
–  Identify objects and locations with high rates of cache misses
–  Identify corrupted caches
Key Aspects to Measure and Monitor
6
Response Headers Can Be Invaluable
7
Alert for Specific CDN-Based Objects or Geos
Isolate specific page
components using Page
Load alerts that are
domain-specific
Geo-target your alerts
with different thresholds
for different regions
8
•  Use Page Load Tests to monitor overall page performance
with objects from many domains and providers
–  Will include HTTP Server, Network tests to the root object
–  Break down DNS time, connect time, wait time
•  Use HTTP Server Tests to see response headers
•  Use Network Tests to monitor the edge or origin directly
–  Understand geo-selection with path visualization
–  Measure latency and throughput
•  Customize alerts for your SLA thresholds
Tips and Tricks
9
Demo
10
Setting Up Tests
Page Load Tests to see all
objects on a page
Network or HTTP Server
Tests to a specific CDN
domain
11
Comparing Paths to Origin and Edge
User to Amazon
(yellow) origin
User to Akamai
(green) edge
12
Look Out for GSLB vs. Anycast Destinations
Anycast CDN
Look at penultimate hops for geo data. Edges here are
located in London, Frankfurt, Warsaw, Dubai and Madrid
13
CDN Performance in Page Load Tests
See response time and
throughput by domain
and provider (like
Akamai or Amazon)
See a full month of
detailed data
Page load and DOM
load aggregate metrics
Select from over 100
vantage points or add
your own
14
Page Load Waterfall: Object-Level Detail
Domain and
provider data 
(by ASN)
Compressed
wire size,
header size and
content size
Blocked, DNS,
Connect, SSL,
Send, Wait,
Receive time
Page load and
DOM load
15
Alerting on CDN Performance – Page Load
Isolate specific page
components using Page
Load alerts that are
domain-specific
You can also make your
alerts geo-targeted, with
different thresholds for
different regions
16
See what you’re missing.
Watch the webinar:

https://www.thousandeyes.com/resources/cdn-webinar

More Related Content

Monitoring CDN Performance

  • 1. CDN Monitoring: 
 From Users to Edge to Origin Young Xu, Product Marketing Analyst
  • 2. 2 About ThousandEyes ThousandEyes delivers visibility into every network your organization relies on. Founded by network experts; strong investor backing Relied on for " critical operations by leading enterprises Recognized as " an innovative " new approach 31 Fortune 500 5 top 5 SaaS Companies 4 top 6 US Banks
  • 3. 3 CDN Architecture from User to Edge to Origin Access Networks Origin Server Edge Server Edge Server Users Internet
  • 4. 4 Methods of Monitoring Your CDNs Users Access Networks Internet Origin Server Edge Server Edge Server Origin to Edge 3 Cloud Agent to Edge 2 Cloud Agent to Origin 1 Cloud Agent Cloud Agent
  • 5. 5 •  Benchmark CDN performance on latency and fetch time –  Before and after object/page is CDN-enabled –  Ongoing with user to edge versus user to origin •  Ensure the proper edge is serving the request –  Geolocation and GSLB vs. Anycast –  Manage multiple CDNs •  Cache issues –  Identify objects and locations with high rates of cache misses –  Identify corrupted caches Key Aspects to Measure and Monitor
  • 6. 6 Response Headers Can Be Invaluable
  • 7. 7 Alert for Specific CDN-Based Objects or Geos Isolate specific page components using Page Load alerts that are domain-specific Geo-target your alerts with different thresholds for different regions
  • 8. 8 •  Use Page Load Tests to monitor overall page performance with objects from many domains and providers –  Will include HTTP Server, Network tests to the root object –  Break down DNS time, connect time, wait time •  Use HTTP Server Tests to see response headers •  Use Network Tests to monitor the edge or origin directly –  Understand geo-selection with path visualization –  Measure latency and throughput •  Customize alerts for your SLA thresholds Tips and Tricks
  • 10. 10 Setting Up Tests Page Load Tests to see all objects on a page Network or HTTP Server Tests to a specific CDN domain
  • 11. 11 Comparing Paths to Origin and Edge User to Amazon (yellow) origin User to Akamai (green) edge
  • 12. 12 Look Out for GSLB vs. Anycast Destinations Anycast CDN Look at penultimate hops for geo data. Edges here are located in London, Frankfurt, Warsaw, Dubai and Madrid
  • 13. 13 CDN Performance in Page Load Tests See response time and throughput by domain and provider (like Akamai or Amazon) See a full month of detailed data Page load and DOM load aggregate metrics Select from over 100 vantage points or add your own
  • 14. 14 Page Load Waterfall: Object-Level Detail Domain and provider data (by ASN) Compressed wire size, header size and content size Blocked, DNS, Connect, SSL, Send, Wait, Receive time Page load and DOM load
  • 15. 15 Alerting on CDN Performance – Page Load Isolate specific page components using Page Load alerts that are domain-specific You can also make your alerts geo-targeted, with different thresholds for different regions
  • 16. 16 See what you’re missing. Watch the webinar: https://www.thousandeyes.com/resources/cdn-webinar