9. Agility for your business
●More efficient teams
●Pay as you go
●Pay for consumption, not capacity
10. Who uses this stuff?
●This is old news for web companies.
●Moving into enterprise data centers globally and being used for
all types of workloads.
●Even telcos are moving to cloud infrastructure.
11. NFV - Network Function Virtualization
●Major shift in telco equipment industry over the last few years
●Goal: take expensive, inflexible hardware appliances and turn
them into VMs running on cloud infrastructure.
●Deliver new services in a matter of days, not months
12. NFV Timeline
●NFV white paper - October, 2012
●2014 -
oOPNFV (Open Platform for NFV) launched by Linux Foundation to
develop an open reference platform
oTelcos and NEPs (Network Equipment Providers) doing POCs
●2015
oProduction deployments expected
21. Mission: to produce the ubiquitous Open Source Cloud
Computing platform that will meet the needs of public and
private clouds regardless of size, by being simple to
implement and massively scalable.
22. OpenStack History
●Launched in 2010 by NASA and
Rackspace
●Foundation created in 2012
●Current sponsorship:
o8 Platinum
o17 Gold
o75+ Corporate
23. OpenStack Foundation Members
●Platinum Members:
oAT&T
oCanonical
oHP
oIBM
oNebula
oRackspace
oRed Hat
oSUSE
●Subset of Gold:
oCisco
oDell
oEricsson
oHuawei
oIntel
oJuniper Networks
oVMware
oYahoo!
●Subset of Corporate:
oAlcatel-Lucent
oAvaya
oComcast
oNokia
oOrange
oOracle
oPaypal
oSAP
24. OpenStack Contributors
●6 month release cycle
●Latest release (Juno - October 16, 2014)
o1500+ individual code contributors [1]
o125+ companies (code contributors) [1]
o20,000+ commits [1]
o120,000+ code reviews [1]
o1.8 million test jobs [2]
ostored and analyzed 18 TB of log data [2]
[1] http://www.stackalytics.com
[2] https://twitter.com/tcarrez/status/522777380604346368
25. Give OpenStack a try!
●RDO - community OpenStack distribution
ohttps://openstack.redhat.com/
●OpenStack based public clouds
ohttp://www.openstack.org/marketplace/public-clouds/
27. Scaling in Cloud
●Horizontal scaling
●Auto-scaling based on load (using Heat)
●Design for failure
28. CPU and Memory Performance
●Beware of overcommit.
●Requirements from NFV use cases
oNUMA awareness
oDedicated CPUs
oDedicated memory (large pages)
29. Network Performance
●Data plane applications want network performance as close to
bare metal as possible
●Requirements from NFV use cases:
oSupport for SR-IOV network devices
oIntel DPDK enablement
30. Consider your storage options
●Ephemeral local block storage
●Persistent block storage (Cinder)
●Relational database service (Trove)
●Object Storage (Swift)