The Bridge - Issue 3, 2021 - 12

Feature
Quality-of-Service Architecture for Cloud Computing Networking
software. Recently, the increased capacity (10 Gbps)
of data networks has caused a new paradigm in data
communications. This shift is to perform protocol
processing in hardware at wire speed.
While layer network architecture provides a hardware
foundation for networking, the protocol layering is the
basis for internet operation and will remain so for some
time. The design requirement for protocol development
is to specify protocols that enables communications
without having to understand what aspects of the
network are processing below this protocol and which
aspects are above the protocol. In other words, protocol
independence is a requirement to communicate in a
multi-protocol environment. This protocol independence
is the crux of the protocol layering perspective as
described in the literature [7].
This protocol processing is quite complex. Each layer that
the message transverses has its own set of attributes and
requirements. Some of the attributes for each layer are
verifying the correctness of the message, classifying the
message and making a forwarding decision based on the
rules for this message.
III. QUALITY OF SERVICE WITH CLOUD PROVIDERS
One approach for QoS in cloud services is to identify
controllable and uncontrollable factors. A system designer
for a hybrid system QoS architecture must addresses the
controllable factors of QoS [8]. The controllable factors
include design decisions and implementation tradeoffs,
for example buffer size, scheduling algorithm, and
metering algorithm. The uncontrollable factors include
physical delays, for example line speed, propagation, and
contention delay.
A. Objective and Scope
The objective of QoS is the differentiation among different
services of user traffic. QoS can be an absolute guarantee
or a relative guarantee. The guarantees are on traffic
characteristics such as loss, delay, bandwidth, and burst
size. For example, burst size could have an absolute
guarantee of 96 kbytes or a relative guarantee that
normally the burst size would not exceed 96 kbytes. The
service model must translate user needs into controllable
technical limits.
This problem space would be simpler if all user traffic
services were individual flows that could be easily
identified. However, the nature of the internet, and
particularly IP, is to aggregate multiple individual flows
THE BRIDGE
at the entrance to the network and treat the flows as
an aggregate in the edge and core of the network.
Aggregation of individual flows greatly simplifies the
processing required in devices. However, the tradeoff
of this simplification is in the quality and reliability
guarantees. The hybrid system QoS architecture must
aim to honor QoS requirements of individual flows while
operating on flow aggregates.
In understanding the scope of this problem, traffic
engineering principles should be applied to define the
solution formally. Cloud network traffic engineering
considerations include defining attributes that associate
with aggregate flows in order to specify and constrain
behavioral characteristics. The traffic attributes can be
associated with resources that constrain the placement
of label switched paths (LSP) and the flows associated
with each LSP. Constraining the type of flows that can
be aggregated together brings about defining formal
arrangements for when aggregation and de-aggregation
occurs. Therefore, the scope of the QoS architecture
encompasses traffic engineering of cloud network flows
with diverse QoS requirements. The associated tasks are:
* To define a formal architecture for which traffic
engineering policies are attached and
* To define a set of policies associated with the formal
model that can be disseminated to each switch or
router [9].
Traffic engineering resources must be defined with regard
to the reference architecture to meet the end-to-end
serviced differentiation.
B. Service Definition
The current internet supports only the " best effort " service
class model. The desire is to change this model and
have the internet support other services in addition to
" best effort. " The motivation is to create pricing models
that enable the service provider to create more services
that attract and keep customers, as well as to create a
price structure for finer granularity of service quality. This
paradigm shift toward a new pricing model has been
called a service-enabled model. The hybrid system QoS
architecture addresses this need. User expectations are
key to any QoS development. User expectations are
defined by the perceived quality and service that the
users expect to receive as a value-paying customer,
i.e. one that is paying a particular rate for a guaranteed
service level. While there are no standardized service
https://hkn.ieee.org/
The Bridge - Issue 3, 2021

Table of Contents for the Digital Edition of The Bridge - Issue 3, 2021
