CS 6250 EVALUATION TEST ANSWERS AND
QUESTIONS SET A+
✔✔What happens when a packet arrives in an openflow device? - ✔✔- lookup process
in first table
- ends with match in a table OR. a miss (no rule found)
(8)
✔✔What are the possible actions for a packet in openflow device? - ✔✔- forward packet
to a port, encapsulate and forward to controller, drop, send to next flow table, etc.
(8)
✔✔Are southbound APIs a barrier for acceptance of SDN technology? What will fix
this? - ✔✔Yes. API proposals that standardize these southbound APIs, like Openflow,
address this
(8)
✔✔What are 3 info sources provided by Openflow protocol? - ✔✔1. EVENT-BASED
messages sent by forwarding devices to controller when LINK OR PORT CHANGE
2. FLOW STATS generated by forwarding devices
3. PACKET MESSAGES sent by forwarding devices to controller when they don't know
what to do with new flow
(8)
✔✔What is ForCES and how is it different from Openflow? - ✔✔ForCES provides
flexible approach to traditional networking WITHOUT changing architecture of network
NO logically centralized controller
(8)
,✔✔What is biggest drawback of traditional networks? - ✔✔- configured using low level,
device specific instructions
- proprietary NOS (network operating systems)
(8)
✔✔what are core controller functions (functions all controllers should provide) - ✔✔-
topology , stats, notifications, shortest path forwarding, security, device management
(8)
✔✔Why use centralized cluster of controllers instead of single entity controller? - ✔✔-
single entity can have scaling issues
- enterprise class networks may deal with millions of flows per second
(8)
✔✔Why use physically distributed controllers? - ✔✔Distributed can scale to any
environment
(8)
✔✔What are 2 ways distribution can occur in SDNs? - ✔✔1. centralized cluster of
nodes
2. physically distributed set of elements
(8)
✔✔How do hybrid distributed controllers work? - ✔✔- clusters of controllers in a data
center, distributed nodes in different sites
(8)
✔✔Do distributed controllers have weak consistency of semantics? - ✔✔Yes
(8)
✔✔Are distributed controllers fault tolerant? - ✔✔Yes
(8)
✔✔What is an ONOS? What is it an example of, and what is its goal? - ✔✔Open
networking operating system
- DISTRIBUTED SDN control platform
- aims to provide GLOBAL VIEW OF NETWORK to apps, scale out performance, fault
tolerance
(8)
, ✔✔How is shared view of network state built with ONOS? - ✔✔- using network topo and
state info when an instance is discovered
(8)
✔✔How are decisions for forwarding and policy made within ONOS? - ✔✔info is
consumed from view
policy decisions are made
update the decisions made back the view
(8)
✔✔What handles propagation of state changes between switch and network with
ONOS? - ✔✔The MASTER openflow controller is the ONOS instance for a group of
switches
Decisions are made within the master
(8)
✔✔How can work be distributed within ONOS? - ✔✔Add more instances in ONOS
cluster
(8)
✔✔How is fault tolerance achieved with ONOS? - ✔✔REDISTRIBUTE the work of a
failed instance to the other remaining instance
When ONOS instance fails, election is held to choose a new master for each switch
(8)
✔✔What is P4 and why was it developed? - ✔✔P4 is a high level programming
language
helps configure switches in conjunction with SDN control protocols
helps with programmability on DATA PLANE
(8)
✔✔What plane does P4 help program? - ✔✔The data plane
(8)
✔✔Why does openflow include multiple stages of rule table, and an increasing number
of header fields? - ✔✔Helps to expose the functionalities of the switch to the controller
(8)
✔✔How does openflow address the increasing number of header fields? - ✔✔flexible
approach to match headers
P4 allows configuring switch programmatically
QUESTIONS SET A+
✔✔What happens when a packet arrives in an openflow device? - ✔✔- lookup process
in first table
- ends with match in a table OR. a miss (no rule found)
(8)
✔✔What are the possible actions for a packet in openflow device? - ✔✔- forward packet
to a port, encapsulate and forward to controller, drop, send to next flow table, etc.
(8)
✔✔Are southbound APIs a barrier for acceptance of SDN technology? What will fix
this? - ✔✔Yes. API proposals that standardize these southbound APIs, like Openflow,
address this
(8)
✔✔What are 3 info sources provided by Openflow protocol? - ✔✔1. EVENT-BASED
messages sent by forwarding devices to controller when LINK OR PORT CHANGE
2. FLOW STATS generated by forwarding devices
3. PACKET MESSAGES sent by forwarding devices to controller when they don't know
what to do with new flow
(8)
✔✔What is ForCES and how is it different from Openflow? - ✔✔ForCES provides
flexible approach to traditional networking WITHOUT changing architecture of network
NO logically centralized controller
(8)
,✔✔What is biggest drawback of traditional networks? - ✔✔- configured using low level,
device specific instructions
- proprietary NOS (network operating systems)
(8)
✔✔what are core controller functions (functions all controllers should provide) - ✔✔-
topology , stats, notifications, shortest path forwarding, security, device management
(8)
✔✔Why use centralized cluster of controllers instead of single entity controller? - ✔✔-
single entity can have scaling issues
- enterprise class networks may deal with millions of flows per second
(8)
✔✔Why use physically distributed controllers? - ✔✔Distributed can scale to any
environment
(8)
✔✔What are 2 ways distribution can occur in SDNs? - ✔✔1. centralized cluster of
nodes
2. physically distributed set of elements
(8)
✔✔How do hybrid distributed controllers work? - ✔✔- clusters of controllers in a data
center, distributed nodes in different sites
(8)
✔✔Do distributed controllers have weak consistency of semantics? - ✔✔Yes
(8)
✔✔Are distributed controllers fault tolerant? - ✔✔Yes
(8)
✔✔What is an ONOS? What is it an example of, and what is its goal? - ✔✔Open
networking operating system
- DISTRIBUTED SDN control platform
- aims to provide GLOBAL VIEW OF NETWORK to apps, scale out performance, fault
tolerance
(8)
, ✔✔How is shared view of network state built with ONOS? - ✔✔- using network topo and
state info when an instance is discovered
(8)
✔✔How are decisions for forwarding and policy made within ONOS? - ✔✔info is
consumed from view
policy decisions are made
update the decisions made back the view
(8)
✔✔What handles propagation of state changes between switch and network with
ONOS? - ✔✔The MASTER openflow controller is the ONOS instance for a group of
switches
Decisions are made within the master
(8)
✔✔How can work be distributed within ONOS? - ✔✔Add more instances in ONOS
cluster
(8)
✔✔How is fault tolerance achieved with ONOS? - ✔✔REDISTRIBUTE the work of a
failed instance to the other remaining instance
When ONOS instance fails, election is held to choose a new master for each switch
(8)
✔✔What is P4 and why was it developed? - ✔✔P4 is a high level programming
language
helps configure switches in conjunction with SDN control protocols
helps with programmability on DATA PLANE
(8)
✔✔What plane does P4 help program? - ✔✔The data plane
(8)
✔✔Why does openflow include multiple stages of rule table, and an increasing number
of header fields? - ✔✔Helps to expose the functionalities of the switch to the controller
(8)
✔✔How does openflow address the increasing number of header fields? - ✔✔flexible
approach to match headers
P4 allows configuring switch programmatically
