Distribution Matters: Why a Distributed Cloud Architecture is Necessary for NFV
 

Distribution Matters: Why a Distributed Cloud Architecture is Necessary for NFV

on

  • 999 views

David Amzallag, VP Virtual Telecommunications and CloudBand CTO, discussed The 9th Fallacy of Distributed Computing at SDN & Openflow Congress, Oct 17, 2013

David Amzallag, VP Virtual Telecommunications and CloudBand CTO, discussed The 9th Fallacy of Distributed Computing at SDN & Openflow Congress, Oct 17, 2013

Statistics

Views

Total Views
999
Views on SlideShare
966
Embed Views
33

Actions

Likes
0
Downloads
8
Comments
0

2 Embeds 33

https://twitter.com 31
http://moderation.local 2

Accessibility

Categories

Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

Distribution Matters: Why a Distributed Cloud Architecture is Necessary for NFV Distribution Matters: Why a Distributed Cloud Architecture is Necessary for NFV Presentation Transcript

  • DISTRIBUTION MATTERS: WHY A DISTRIBUTED CLOUD ARCHITECHTURE IS NECESSARY FOR NFV? DAVID AMZALLAG VICE PRESIDENT, VIRTUAL TELECOMMUNICATIONS AND CLOUDBAND CTO, ALCATEL-LUCENT david.amzallag@alcatel-lucent.com SDN & OPENFLOW CONGRESS, OCTOBER 17, 2013 COPYRIGHT © 2013 ALCATEL-LUCENT. ALL RIGHTS RESERVED. @david_amzallag
  • THE 9TH FALLACY OF DISTRIBUTED COMPUTING • • FALLACIES OF DISTRIBUTED COMPUTING ARE A SET OF ASSUMPTIONS ORIGINALLY ASSERTED PROGRAMMERS NEW TO DISTRIBUED APPLICATIONS INVARIABLY MAKE… 1. THE NETWORK IS RELIABLE. BY PETER DEUTSCH (SUM MICROSYSTEMS > ORACLE, 1994) AND JAMES GOSLING (1996) 5. TOPOLOGY DOESN’T CHANGE. • 3. BANDWIDTH IS INFINITE. 4. THE NETWORK IS SECURE. 6. THERE IS ONE ADMINISTRATOR. THESE ARE ULTIMATELY PROVE FALSE • 2. LATENCY IS ZERO. WE ARE HAPPY TO ADD THE 9TH FALLACY  7. TRANSPORT COST IS ZERO. 8. THE NETWORK IS HOMOGENEOUS. 9. LOCATION IS IRRELEVANT. 2 COPYRIGHT © 2013 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
  • WHY DISTRIBUTE? WHAT ARE THE MAIN CHALLENGES? 7 REASONS 1 RT NETWORK AWARENSS • HOW TO UNDERSTAND NETWORK BEHAVIOUR AND TO “MANAGE AND ORCHESTRATE” THIS ACCORDINGLY? • INTEGRATION WITH SDN 2 3 LIFE CYCLE MANAGEMENT DISTRIBUTION INSIDE THE NODE • HOW TO ON-BOARD A VNF AND TO MANAGE ITS LIFE CYCLE CORRECTLY? • HOW TO MOVE VMS OF MULTI-TIER APPLICATION BETWEEN NODES? WHY TO DO THAT? • HOW TO ENABLE EFFICIENCY IN THE NODE AND TO ENABLE WORKLOAD MIGRATION FROM ONE “RACK” TO ANOTHER BASED ON LOAD? • HOW TO PLACE STORAGE OBJECTS “CORRECTLY” TO ENSURE OPTIMIZED UTILIZATION OF STORAGE? • OPENSTACK AND DISTRIBUTION? • EXAMPLES? 3 COPYRIGHT © 2013 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
  • WHY DISTRIBUTE? WHAT ARE THE MAIN CHALLENGES? 7 REASONS MADE BY CLOUDBAND 4 VM PLACEMENT 20 NODES, 1000 VMS/NODE, 40000 HOURS • WHAT’S THE OBJECTIVE? • TO FIND THE RIGHT BALANCE BETWEEN MAXIMUM TOTAL AVAILABLE RESOURCES (IF RESOURCES RUN OUT – NEW REQUEST CANNOT BE SATISFIED) AND TO AVOID RESOURCE BOTTLENECK (IF RESOURCES ARE SPREAD TOO THINLY – THEY ARE EFFECTIVELY WASTED) AWS? • POLICY-BASED AUTOMATIC PLACEMENT 4 COPYRIGHT © 2013 ALCATEL-LUCENT. ALL RIGHTS RESERVED.
  • WHY DISTRIBUTE? WHAT ARE THE MAIN CHALLENGES? 7 REASONS 5 6 7 END-TO-END RESOURCE MANAGEMENT SENSITIVITY FOR LATENCY RELIABILITY AND AVAILABILITY • DYNAMIC CAPACITY MANAGEMENT • EPC BY 3GPP? • WORKLOAD MIGRATION (WHICH VM TO MIGRATE? WHY? TO WHERE? • “BUDGETED” LATENCY, FOR INSTANCE, FOR A 3GPP EPC • OPTIMIZATION OF PLACEMENT IN “IDLE TIME” • OFFLINE CAPACITY PLANNING AND DESIGN: HOW MANY NODES? WHERE TO LOCATE? HOW MUCH CAPACITY FROM EACH RESOURCE? 5 COPYRIGHT © 2013 ALCATEL-LUCENT. ALL RIGHTS RESERVED. • WHEN FAILURE OCCURES (RACK, SERVER…), THE RIGHT VMS ARE TO MIGRATE TO A DIFFERENT SITE
  • SUMMARY • AS A USER, YOU DON'T NEED TO BE AWARE OF THE DISTRIBUTION. IT SHOULD BE AUTOMATIC AND SIMPLE • HAVING ALL OF THIS IN MIND, THERE ARE HIGH EXPECTATIONS FROM THE NFV PLATFORM! • CLOUDBAND IS A WORKING PRODUCT. DISTRIBUTION IS INHERENTLY PART OF IT • NFV IS ALL ABOUT OPERATIONS. WITHOUT DISTRIBUTION ONE WILL NOT BE ABLE TO DELIVER THE NFV PROMISES TO A BETTER OPERATIONALIZED NETWORK • LOCATION IS INDEED RELEVANT! 6 COPYRIGHT © 2013 ALCATEL-LUCENT. ALL RIGHTS RESERVED.