Cloud Computing and Service Models

Prepared by,
N.Balaganesh,
Public, Private and Hybrid
Public, Private and Hybrid
Cost Model - Cloud
SAAS, PAAS & IAAS
• SAAS – Software as a Service
• PAAS – Platform as a Service
• IAAS – Infrastructure as a Service
SAAS, PAAS & IAAS
SAAS, PAAS & IAAS
Architectural Design of Compute and Storage
Clouds
Generic Cloud Architecture
Layered Cloud Architectural
Development
Market Oriented Cloud Architecture
 Architectural Design Challenges
Challenge 1 – Service Availability and Data Lock-in Problem

• Getting account from the multiple service providers may provide more protection
from the failures

• SAAS providers may get obstacle from distributed denial of service (DDoS) attacks

• But scale up of resources at those moment may defend these attacks

• Data Lock-in can be avoided by using same software environment in both public
and private cloud. The task which cannot be easily executed in data center can be
put in public cloud
 Architectural Design Challenges
Challenge 2 – Data Privacy and Security Concerns

• Many nations has laws to keep the data within national boundaries

• In cloud environment, newer attacks may result from hypervisor malware, guest
hijacking, VM rootkits

• Passive attacks – Stealing of sensitive data or passwords

• Active attacks – Manipulate kernel data structures which can cause major damage
to cloud servers
 Architectural Design Challenges
Challenge 3 – Unpredictable Performance and Bottlenecks

• I/O sharing between the VMs is problematic. The time taken for writing a file on
VMs storage in shared environment will be very slow

• The solution is to improve the I/O architectures and OS to effectively virtualize

interrupts and I/O channels

• Data placement and transport in cloud is getting complicated because of

continuous flow in data

• To overcome this limitation, the weak servers should be removed

 Architectural Design Challenges
Challenge 4 – Distributed Storage and Widespread Software Bugs

• Data center must meet the expectation in terms of scalability and data durability.
Data consistency checking in SAN connected data centers is a major challenge

• Debugging of software bugs should take place in large scale distributed production
environment. But this requirement is practically not possible.

• Debugging using simulators is better solution. But simulator should be well

designed
 Architectural Design Challenges

Challenge 5 – Cloud Scalability, Interoperability and Standadization

• Scalability is based on the SLA and each service providers provide the scalability in
different types. Eg. Google scales up and down the resources based on the need.
Amazon allots resources earlier and will be used when needed.

• Open Virtualization Format (OVF) describes the format for packaging and
distribution of VMs. It can apply to different virtualization platforms with different
levels of virtualization
 Architectural Design Challenges
Challenge 6 – Software Licensing and Reputation Sharing

• Mostly open source software are being used in cloud computing. But if licensing
structure are changed by commercial software vendors for cloud computing then
it could be used in pay and use model

• If customer misuses the cloud then reputation of cloud will come down

• Similar to trusted E-Mail services, the services can be offered by the customers of
the cloud. At SLA level, the cloud service providers want legal liability to remain
with the customer
Inter Cloud Resource Management
 Resource Provisioning and Platform
Deployment
1. Provisioning of Computing Resources

• According to SLA, the sufficient resources should be made available by the service
providers

• Underprovisioning leads to broken SLA. Overprovisioning of resources leads to

under utilization which is loss for the service providers

• So, efficient VM provisioning mechanisms should be available i.e. efficient

installation of VMs, live VM migration and faster recovery from failures. Therefore,
Amazon EC2 and IBM Blue Cloud uses XEN VMM (Customized OS for specific
applications)
 Resource Provisioning and Platform
Deployment
1. Provisioning of Computing Resources

• Power efficient schemes for caching, query processing and thermal management
are mandatory due to increasing energy wastage by heat dissipation from data
centers
 Resource Provisioning and Platform
Deployment
2. Resource Provisioning Methods

• Case (a) – Overprovisioning with the peak load causes heavy resource wastage
(shaded area)

• Case (b) – Underprovisioning of resources results in losses for both users and
providers
 Resource Provisioning and Platform
Deployment
2. Resource Provisioning Methods

• Case (c) – Constant provisioning of resources with fixed capacity to a declining user
demand could result in severe wastage of resources
 Resource Provisioning and Platform
Deployment
3. Demand driven Resource Provisioning

• This method adds or removes computing instances based on current utilization

level of allocated resources

• When a resource is below threshold level for some time period, the resources
could be decreased accordingly and increased when demand increases

• Amazon implements such auto-scale feature in its EC2 platform

 Resource Provisioning and Platform
Deployment
4. Event driven Resource Provisioning

• This method adds or removes computing instances based on time event

• The scheme better for seasonal or predicted events such as Christmas, New year

• If the event is predicted correctly, then this scheme will be good

 Resource Provisioning and Platform
Deployment
5. Popularity driven Resource Provisioning

• This method adds or removes computing instances based on popularity of certain

applications based on internet search

• It works well if prediction is correct. Otherwise, resource will get wasted

 Resource Provisioning and Platform
Deployment
6. Dynamic Resource Deployment

• This method adds resources from other grid to achieve scalability in performance

• Intergrid managed infrastructure was developed by Melbourne University Group

using Java for enabling dynamic resource deployment

• IGG (Intergrid Gateway) allocates resources from a local cluster to deploy

applications in 3 steps (1) Requesting VM, (2) Enacting the Leases and (3)
Deploying the VMs as requested
 Resource Provisioning and Platform
Deployment
6. Dynamic Resource Deployment
 Resource Provisioning and Platform
Deployment
7. Provisioning of Storage Resources

• Because of much volume of information storage in the cloud, the SSD (Solid State
Drives) are preferred than hard disks

• The file systems such as GFS (Google File System), Microsoft Cosmos and HDFS
(Hadoop Distributed File System) are preferred rather than traditional file systems

• The traditional database are unable to hold large volume of both structured and
semi-structured data. So BigTable from Google, SimpleDB from Amazon and SQL
Service from Microsoft Azure are used
Global Exchange of Cloud Resources
 Global Exchange of Cloud Resources
• IAAS service providers have established data centers in multiple geographical
locations around the world

• The location selection is also made available for customers so that they could
specify where to keep there data

• But customer may not know the best location for them to keep the data or
services in advance. Also, cloud service providers cannot setup a data center in all
countries around the world

• So to meet the QoS, the use of services from multiple cloud service providers is
made enabled
 Global Exchange of Cloud Resources
• Cloudbus project in Melbourne university has proposed InterCloud architecture
which supports brokering and exchange of cloud resources for scaling applications
across multiple clouds
 Global Exchange of Cloud Resources
• Cloud providers will be able to dynamically expand or resize their provisioning
capacity based on sudden spikes by using InterCloud architecture

• Cloud Exchange acts as a market manager for bringing together service providers
and consumers

• It aggregates the infrastructure demands from the application brokers and

evaluates them against the available supply currently published by the cloud
coordinators
Thank You