1

Donghyun (David) KimDepartment of Mathematics and Computer ScienceNorth Carolina Central University

NSF Report on Support for Cloud Computing

Ref: http://www.nsf.gov/publications/pub_summ.jsp?ods_key=nsf12040

Spring 2012Seminar Series on Networking

2

Regarding the Reference• NSF Report on Support for Cloud Computing

In response to America COMPETES Reautho-rization Act of 2010, Section 524

• Release Date: February 6, 2012

3

Agenda• Definition of Cloud Computing (by NIST)

• Cloud Computing Research Enhancement Focus Area

• NSF CISE Cloud Computing Awards 2009 – 2011

• Initiatives on Cloud Computing 2009 – 2011

• Ongoing and Future Initiatives on Cloud Computing

4

Cloud Computing (by NIST)• A model for enabling • ubiquitous, • convenient, • on-demand network access

• to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider inter-action

5

Cloud Computing (by NIST)- cont’• This cloud model is composed of • Five essential characteristics• On-demand self-service• Broad network access• Resource pooling • Rapid elasticity• Measured service (pay per use)

• Four deployment models• Private Cloud• Community Cloud (sharing among several organiza-

tions)• Public Cloud (public over the Internet)• Hybrid Cloud (composition of two or more clouds)

6

Cloud Computing (by NIST)- cont’

• Three service models • Cloud Infrastructure as a

Service (laaS)• Physical or virtual machine,

raw storage, firewalls, load balancers

• Cloud Software as a Service (SaaS) • Computing platform/solution

stacks such as OS, runtime environment, database

• Cloud Platform as a Service (PaaS) – i.e. applications

Image source: http://en.wikipedia.org/wiki/File:Cloud_computing.svg

7

Cloud Computing (by NIST)- cont’

Image source: http://en.wikipedia.org/wiki/File:Cloud_computing.svg

8

Cloud Computing (by NIST)- cont’• Offers the promise of • massive cost savings • increased IT agility

• Cloud computing technology challenges many tra-ditional approaches to datacenter and enterprise application design and management.

• Major Challenges• Security• Interoperability• Portability

9

Cloud Computing Research En-hancement Focus Areas• New algorithms and technologies to define, assess, and es-

tablish large-scale, trustworthy, cloud-based infra-structures

• Optimizing the effectiveness and efficiency of cloud com-puting environments

• Models and advanced technologies to measure, assess, re-port, and understand the performance, reliability, energy consumption, and other characteristics of complex cloud environments

• Mitigating security, identity, privacy, reliability, and manageability risks in cloud-based environments, includ-ing as they differ from traditional data centers

• Advanced security technologies to protect sensitive or proprietary information in global-scale cloud environ-ments.

10

Data Center vs. Cloud Computing• Traditional Data

Center• Local network• Proprietary, custom-

ized• Economy of scale: Or-

ganization• Full control• Mostly secure • Limited capacity• Dedicated

• Cloud Computer Service• Internet• Standardized• Economy of scale:

Ecosystem• Partial control• Less secure • Nearly unlimited

capacity• Shared

Ref: http://blogs.zdnet.com/Hinchcliffe

11

NSF CISE Cloud Computing Awards 2009 - 2011• Active Awards per Division• CNS 76 (2.7% of new awards in FY09-FY11)• CCF 40 (4.1% of new awards in FY09-FY11)• IIS 9 (0.6% of new awards in FY09-FY11)

• Funded Areas• Computer Systems• Computer Networks• Security and Privacy• Algorithms and data management• Applications and software engineering• Computer science education

12

Awards in Computer Systems1. Cloud Architectures• Project 1 • Leverages the large number of servers in a data center, as to create

a dense network of connections between them without using high-end switches

• To provides the foundation for enhancing the scalability, inter-server capacity, and fault tolerance of the data center network.

• Project 2: exploring new memory architectures for data centers• Instead of dynamic random access memory (DRAM) , uses multiple

technologies, such as phase-change memory (PRAM) and flash mem-ory

• To construct a high-capacity, energy- efficient memory system for computer servers in the clouds.

• Project 3: multi-core aware cloud server• Designing and analyzing different methods and algorithms that can

be used to speed up the execution of cloud server software on multi-core chips.

13

Awards in Computer Systems2. Green Clouds• A Recent Report by US EPA• Data centers in the United States incur an an-

nual energy cost of $4.5 billion, which is com-parable to the total consumption of 5.8 mil-lion average US households

• The total energy consumption of data centers is expected to double every five years if no in-tervention

• Up to 80% of this projected energy expenditure may be avoidable

14

Awards in Computer Systems2. Green Clouds• Project 1• Reduce energy consumption & increase energy

efficiency• Hardware assisted parallel computing such as

multi-core chips• Exploits residual resources to run a secondary

computation in the background (consume less power than primary)

• Project 2: Manage idle time on servers• Most servers are only 10-30% busy• While idle, the servers still consume 60% or

more of their peak power consumption

15

Awards in Computer Systems2. Green Clouds• Project 3: Work load vs. active servers• Load oblivious but smart policies needed – automati-

cally scale data center resources to reduce power con-sumption

• Servers are shut down, put into sleep states, or run at different frequencies without knowing in advance how much work needs to be done.

• Project 4: Run on variable power from renewable sources

• Such as solar and wind power• Investigate techniques to mange and adapt the energy

and power foot print of a server to fluctuations in supply• Require a server to be “duty-cycled” – turned on and off in

a carefully controlled fashion

16

Awards in Computer Networks1. Network Support for Clouds• Investigate modern computer networks need to provide

new communication primitives (basic operations such as naming, routing, switching, congestion control, and content distribution) that are more suitable for clouds.

• Project 1: Service-level object based net operations• Exploring the possibility of basing network operations (such

as naming and routing) on service-level objects (such as ap-plications or databases) instead of basing them on computer hosts.

• Project 2: Service/application-aware routing• Allows each application in the datacenter network to create

its own custom routing tree

17

Awards in Computer Networks1. Network Support for Clouds• Project 3: Efficient content distribution in cloud• Optimal swarming algorithms for distribut-

ing content over multiple multicast trees• To deliver messages from the clouds to large

client populations.• Able to communicate while achieving several de-

sirable performance objectives

• Project 4: Precision time protocols in cloud • To achieve time-predictable delivery of con-

tent from clouds to clients

18

Awards in Computer Networks2. Marketplace for Clouds• Project 1:• Exploring the merits of adopting “co-location games”

as a framework for how clouds can be deployed and used in an economically sound manner.

• Consider how rational, self-interested parties in-teract to secure their share of cloud resources to support their applications

• Model and analyze the dynamics that result when these parties negotiate to minimize their individ-ual costs

19

Awards in Computer Networks2. Marketplace for Clouds• Project 2• Investigate the challenges associated with migrating

enterprise applications to clouds such as• Models to predict the impact of this migration on

the performance of the applications• Methodologies to ensure that policies for network access

to the applications on clouds are configured correctly• Methodologies to ensure that the applications, when ex-

ecuted on clouds, provide adequate quality of service guarantees

• Project 3: how to evolve the client machines • To make “virtual desktop clouds” accessible via light-

weight client machines

20

Awards in Computer Networks3. Resilient Networks for Clouds• Project 1• Quantifying the frequencies, durations, causes, and impacts of

faults that occur within datacenter networks, as well as other network classes.

• To generate a model of network faults that can be used to evaluate how suitable or effective different applications and pro-tocols are for various datacenter network architectures.

• Project 2• Becoming vulnerable to "soft errors" caused by external noise,

and are increasingly likely to fail early due to fatigue. • To investigate computational "resilience methods“ that scale

gracefully in the face of increasing hardware failures. • Use novel partitioned redundancy strategies that achieve

resiliency to network failures, and will function at different levels across hardware and software layers.

21

Awards in Security and Privacy1. Trustworthiness of Cloud Providers• The effectiveness of cloud computing services

is limited by the amount of trust that users have in providers. • how providers can provide more trustworthy

services?• how to prove their trustworthiness?

• Encryption of each data• Less available service from Cloud Provider (less

useful)• Homomorphic encryption• To provide without decryption (still impractical)

22

Awards in Security and Privacy1. Trustworthiness of Cloud Providers• Example: a company• Each uses an individual encryption key

→ if the individual dies, all data lost• A company shares a encryption key

→ difficult to revoke

• Project 1: Functional Encryption• Data can be encrypted so that only those users

currently holding appropriate credentials can ac-cess the data

• Fully access for auditors, restricted for the other works

23

Awards in Security and Privacy2. Protecting cloud providers from threats• New Security Threats & Challenges• Due to the large number of users in a cloud,

traditionally non-profitable attacks become profitable

• Cloud providers also have to protect their users from each other• if one client is a malware victim, other clients of

the same provider should not be endangered.• An attacker should not be able to buy ser-

vices from the same provider as its intended victim and then use those services to defeat the provider's defenses

24

Awards in Security and Privacy3. Leveraging the cloud to provide trustworthy applications• Project 1: Sharing health care records• Needs appropriate privacy and security mechanisms

are in place to protect patients, doctors, and insur-ance providers

• Project 2: Sharing information among com-panies• Every company has its own internal security and pri-

vacy policies, and different companies might have different regulatory restrictions on data usage

• Needs to understand and enforce these differing poli-cies, so that each company can benefit from collabo-ration without the risk of violating its own regulations

25

Awards in Algorithms and Data Management: 1. Algorithm• Major directions• Cryptography for security and privacy• Done

• Scheduling• Resource management• Coding theory• Formal methods

26

Awards in Algorithms and Data Management: 1. Algorithm – cont’• Scheduling• Multiple instances of multiple resources given multiple

competing objectives (e.g. energy consumption, ser-vice level agreement of multiple users)

• Each can be computational intensive, network in-tensive, and/or storage intensive

• Resource management• Map physical resources into logical resources to be

shared in dynamic environment• A kind of online problems, but unlike traditional ones,

resources can be redistributed• How this can be done efficiently?

27

Awards in Algorithms and Data Management: 1. Algorithm – cont’• Coding theory• Distributed novel distributed storage codes that use

network coding theory to address modern storage challenges

• Formal methods• The correctness of different components of cloud

computer systems can be verified using formal proof-theoretic frameworks for• Partition management protocol (which provides the

computing elasticity of the system)• Storage subsystem, and • Atomicity- guaranteeing protocol (which preserves in-

tegrity of data and computations).

28

Awards in Algorithms and Data Management: 2. Handling Massive Data• Clouds for data-intensive applications• Large-scale data management

• Projects 1 - 3• Scaling up and scaling out database opera-

tions to the extreme data sizes and levels of dis-tribution that will be present in the cloud envi-ronment

• Study security issues raised by multiple cloud users sharing large scale data resources

• Energy consumption should be considered

29

Awards in Algorithms and Data Management: 2. Handling Massive Data• Project 4: Theoretical foundations and practi-

cal control algorithms• To enable scalable design and efficient man-

agement of future extreme-scale data-in-tensive computing

• Identify fundamental design principles • Develop 10 distributed control strategies on opera-

tor placement, data storage, load shedding, and resource allocation to allow for efficient in-net-work information processing

• Require a collaborative effort spanning multiple disciplines including performance modeling, net-working, queueing theory, and optimization

30

Awards in Algorithms and Data Management: 2. Handling Massive Data• Project 4: • Investigating how critical data-driven com-

puter vision tasks, such as nearest neighbor searching and clustering in high-dimensional spaces, can be designed for cloud computing systems

31

Awards in Applications and Software EngineeringDeveloping Applications for Cloud Environments• Clouds for particular applications or particular

application domains• Hydrology/water resource management• Natural language processing on web-scale data

sets• Geographic information systems• Biological applications such as protein folding

and evolutionary biology• In-home control applications• Theoretical physics such as computational

string theory.

32

Awards in Applications and Software EngineeringDeveloping Applications for Cloud Environments – cont’• Project 1: Ubiquitous event reporting and data

gathering on the 2010 oil spill in the Gulf of Mex-ico and its ecological impacts• Smart phones (which have sophisticated sensor

packages, high-level programming APis, and mul-tiple network connectivity options)

• Cloud computing infrastructures to enable col-lecting and aggregating data from mobile applica-tions

• To develop a scientific basis for managing quality-of-service, user coordination, sensor data dissemi-nation, and validation issues that arise in mobile disaster monitoring applications

33

Awards in Applications and Software EngineeringSoftware Engineering for the Cloud Environment• Improving the programmability of parallel and distributed systems• Better methods of detecting and remedying errors • Programming and algorithm design models for cloud applications and

for energy-efficiency in large-scale systems• (soft or hard) real-time guarantees

• Project 1: Designing, implementing, and evaluating extensions to the services offered by cloud purveyors so that cloud services can be used by a broader developer base.• Scientists and students - require support for general compute-in-

tensive applications, to better work on the clouds• Fundamental technologies necessary for support of high-performance,

compute intensive applications within the clouds• These technologies include efficient support of shared- memory inter-

process communication, high-performance file system support, and system-wide performance monitoring and analysis tools

34

Awards in Applications and Software EngineeringSoftware Engineering for the Cloud Environment – cont’• Project 2• To tackle the problem of designing and programming large ap-

plications over the clouds: group compositional approach• First, develop novel protocols for smaller groups of nodes which

offer strong properties with minimal overhead• Second, they propose a coordination service and a suite of man-

agement algorithms to adaptively organize these smaller groups, composing them together into a large application.• Will address the problems of dynamic load balancing, topological con-

trol, and security• Project 3• Investigating cloud computing systems serving multiple users

with differing incentives• To ensure the reliability of these systems in face of unrespon-

siveness or selfish behavior by the underlying competing users

35

Awards in Computer Science Edu-cation• Parallel and Distributed Computing Curriculum• Parallel and Distributed Computing (PDC) now per-

meates most computing activities; this is especially true in the cloud computing environment

• Undergraduate Research in Cloud Computing• Focused on “Computer Systems Research in High

Performance Cloud Computing Environments”• Undergraduate researchers are immersed in de-

partmental research areas of expertise, which in-cludes computer architecture, energy-aware comput-ing, virtualization, security, and cloud and grid com-puting

36

Initiatives on Cloud Computing 2009- 2011• 3.1CiC: Computing in the Cloud• A collaborative cloud computing agreement that Microsoft Cor-

poration and NSF (announced April 20, 2011)• Microsoft is providing 13 cloud computing research projects with

access to Windows Azure, a cloud computing platform that provides on-demand computing and storage to host, scale and manage Web applications on the Internet through Microsoft data centers.

• PROBE: A National Facility for Hosting Cloud Test-Beds• Funded by NSF-CISE-CNS, the New Mexico Consortium, the Los

Alamos National Laboratory, the University of New Mexico, Carnegie Mellon University, and the University of Utah

• Start with a 1000-node cluster with 2000 processors, and will grow

• Created at a fraction of the cost of a typical supercomputing fa-cility

37

Initiatives on Cloud Computing 2009- 2011• NEBULA: A Future Internet Architecture to Support the Clouds• Funded by NSF-CISE-CNS-NeTS • The NEBULA architecture surrounds a highly available and exten-

sible core network with trustworthy transit and access networks that enable many new forms of communication and computing

• Achieves the three security properties of confidentiality, in-tegrity and availability • the NEBULA Data Plane that establishes policy-compliant paths

and provides both flexible access control and defense against avail-ability and denial-of-service attacks

• the NEBULA Control Plane that provides access to application-se-lectable service and network abstractions such as redundancy, con-sistency, and policy routing

• the NEBULA Core that redundantly interconnects data centers con-taining replicated data with ultra-high availability core routers (with Cisco)

38

Critical research areas identified by recent PI meeting• Cloud Architectures and Systems• Network Support for Clouds• Data Portability, Consistency, availability, and Man-

agement• Programming Models for Clouds• Fault Masking in the Clouds• Cloud Security, Privacy, and Auditing• Cloud Debugging, Certification, Diagnosis, and Update• Cloud Self-Monitoring and Autonomic Control• Cloud Inter-Operability and Standardization• Green Clouds• Cloud Test-beds

39

Ongoing and Future Initiatives on Cloud Computing• CISE Core and Cross-cutting Programs• Active awards: CNS 76, CCF 40, IIS 9 • FY 2012 solicitation called out cloud computing as a

“highlighted area” for CSR program• CCF awards focus on• Algorithmic Foundation (AF)• Communication and Information Foundations (CIF)• Software and Hardware Foundations (SHF)• CiC

• IIS awards focus on Information Integration and In-formatics

• Also from cross-division Trustworthy Computing (TC)