Neal N. Xiong @ GSU Slide 1
June 25, 2011
Cloud Computing
Services and Architecture
Neal N. Xiong
Georgia State University
Neal N. Xiong @ GSU Slide 2
June 25, 2011
Know exact case for the routers group: If, good for packets transmission Otherwise, miss packets, reduce QoS of packets transmission Networks resource are not extensive shared (partly shared)
UserUserUserUser
Traditional network application
Router
Cloud
Computing
Neal N. Xiong @ GSU Slide 3
June 25, 2011
What is a cloud? Definition [Abadi 2009]
shift computer processing, storage, and software away from the desktop and local servers
across the network and into next generation data centers
hosted by large infrastructure companies, such as Amazon, Google, Yahoo, Microsoft, or Sun
Neal N. Xiong @ GSU Slide 4
June 25, 2011
Dynamic cloud-based network model
North Carolina State University VCL modelhttp://vcl.ncsu.edu/
User/applicationsVCL Software and
Management nodes
Servers
Neal N. Xiong @ GSU Slide 5
June 25, 2011
VCL model in George Mason University
Neal N. Xiong @ GSU Slide 6
June 25, 2011
GMU model GMU uses configured IBM® BladeCenter® technology
and open source VCL software Involve the provision of systems and network security,
high-speed network services, a Web portal, a database server, a software image library and management nodes
Web portal manages the end-user reservation interface Database server holds a SQL database used for
managing reservations, images, and the VCL code configuration.
Management nodes are servers used in the
configuration of the overall VCL system.
Neal N. Xiong @ GSU Slide 7
June 25, 2011
Examples of Cloud Service
A student taking a statistics class might access the Mathematica to complete a homework assignment from a residence hall.
Another student might access Mathematica from home, before heading to work.
Yet another student taking that same statistics class might access Mathematica through a
wireless hub at a local restaurant. VCL applications included Stata v10, SPSS v17,
Matlab 2008a, Maple 13, and Mathematica 7.0.1.
Neal N. Xiong @ GSU Slide 8
June 25, 2011
VCL login interface in GMU
Neal N. Xiong @ GSU Slide 9
June 25, 2011
Examples of Cloud Service
For students, computer requirements to access the VCL are very small (remote student, public library).
Any point of minimal Internet access is sufficient to open the door for service, online, goes wherever students go.
For IT professionals, VCL helps address reduced budgets, increased power costs, aging equipment, software management, and the need to provide (software for teaching and learning).
Cost factors include hardware (e.g. bladecenter, workstations), software (e.g. VCL code), software management and imaging, HVAC, electrical, lighting, data storage, Internet access, network wiring, furniture and staff.
Neal N. Xiong @ GSU Slide 10
June 25, 2011
Improvements in areas of software licensing and maintenance are made through the VCL model as well.
Software use is tracked and metered allowing license purchases to match actual software demand.
Ensure equity in services to distance learners, a component of accreditation of distance education programs.
VCL connects users of very specific applications, exposing opportunities for sharing license costs.
Operational cost per service hour for VCL hardware is less than 1 cent (NCS, $13,000 were saved by reducing hours in 1 lab/1 week)
green Power costs: $11,594/year as compared with $3,944 in 1 lab/1 week
Neal N. Xiong @ GSU Slide 11
June 25, 2011
Reduced Costs
North Carolina State University: five students use one seat in a physical computer lab, On average, 25 students use one virtual seat in a virtual computing lab.
George Mason Univ: computer labs are available almost 93 hours a week. The VCL is available 168 hours a week. Accounting for maintenance, there are 7884 potential useable hours per year per virtual seat.
offers sophisticated tracking of professional software. The VCL connects users of very specific applications,
exposing opportunities for sharing license costs.
Neal N. Xiong @ GSU Slide 12
June 25, 2011
Barriers to virtual computing: outdated software licensing model, License costs
vary, change with little notice … add further complexity licensing and licensing
negotiation open doors for coordinated purchases across
departments and institutions.
Neal N. Xiong @ GSU Slide 13
June 25, 2011
Georgia State University Student Tech Fee Hardware/Software Cost Cycles: Requested vs.
Funded
0500000
1000000
1500000
2000000
2500000
3000000
35000002001
2002
2003
2004
2005
2006
2007
2008
2009
2010
HW RequestHW FundedSW RequestSW Funded
Neal N. Xiong @ GSU Slide 14
June 25, 2011
VCL login interface in Georgia State University
Neal N. Xiong @ GSU Slide 15
June 25, 2011
Graph analysis for this data
Laptop (seconds)
VCL (seconds)
0.75000 0.73440
0.54690 0.23440
0.48440 0.04690
0.07810 0.03130
0.04690 0.01560
Window: the comparison
between Laptop and VCL
Neal N. Xiong @ GSU Slide 16
June 25, 2011
Graph analysis for this dataMac: the comparison
between Laptop and VCL
Laptop (seconds)
VCL (seconds)
1,449.70000 944.95310
1,924.40000 1,038.10000
1,952.30000 1,208.70000
2,185.70000 1,382.60000
2,520.50000 1,546.10000
2,780.20000 1,709.50000
1,648.00000 1,072.30000
Neal N. Xiong @ GSU Slide 17
June 25, 2011
Comparison between Laptop and VCL
Laptop (seconds) VCL (seconds) Improvement
0.75000 0.73440 2.08%
0.54690 0.23440 57.14%
0.48440 0.04690 90.32%
0.07810 0.03130 59.92%
0.04690 0.01560 66.74%
Laptop (seconds) VCL (seconds)
1,449.70000 944.95310 34.82%
1,924.40000 1,038.10000 46.06%
1,952.30000 1,208.70000 38.09%
2,185.70000 1,382.60000 36.74%
2,520.50000 1,546.10000 38.66%
2,780.20000 1,709.50000 38.51%
1,648.00000 1,072.30000 34.93%
Neal N. Xiong @ GSU Slide 18
June 25, 2011
Dynamic cloud-based network model
U.S.
southern
state
education
Cloud,
sponsored
By IBM,
SURA
&
TTP/ELC
Neal N. Xiong @ GSU Slide 19
June 25, 2011
Q & A
Thank You!
Neal N. Xiong @ GSU Slide 20
June 25, 2011
[1-RED] S. Floyd and V. Jacobson, ”Random early detection gateways for congestion avoidance,” IEEE/ACM Transactions on Networking, Vol. 1, pp. 397-413, Aug. 1993.
[2-RED] C. V. Hollot, V. Misra, D. Towsley, and W. B. Gong, ”A Control Theoretic Analysis of RED,” Proceedings of IEEE INFOCOM 2001, Anchorage, Alaska, vol. 3, pp. 1510-1519, April 2001.
[3- Adaptive RED] Sally Floyd, Ramakrishna Gummadi and Scott Shenker, ”Adaptive RED: An algorithm for increasing the robustness of RED’s active queue management,” Berkeley, CA, Technical Report, to appear, 2001.
References
Neal N. Xiong @ GSU Slide 21
June 25, 2011
References
[4-PD-RED] Jinsheng Sun, King-Tim Ko, Guanrong Chen, Sammy Chan, and Moshe Zukerman, ”PD-RED: to improve the performance of RED,” IEEE Communications Letters, vol. 7, no. 8, pp. 406-408, August 2003.
[5-PI-RED] C. V. Hollot, Vishal Maisra, Don Towsley and Wer-Bo Gong, ”On designing improved controllers for AQM routers supporting TCP flows,” Proceedings of IEEE INFOCOM 2001, Anchorage, Alaska, April 2001.
[6- SPI-RED] Naixue Xiong, Xavier Défago, Xiaohua Jia, Yan Yang, Yanxiang He: Design and Analysis of a Self-Tuning Proportional and Integral Controller for Active Queue Management Routers to Support TCP Flows. INFOCOM 2006.
Neal N. Xiong @ GSU Slide 22
June 25, 2011
References
[7-LRC-RED] Naixue Xiong, Laurence T. Yang, Yan Yang, Xavier Defago, Yanxiang He, “A Novel Numerical Algorithm Based onSelf-Tuning Controller to Support TCP Flows,“ Mathematics and Computers in Simulation, 79(4):1178-1188, 2008. (Elsevier)
[8-RED] H. Zhang, C. V. Hollot, D. Towsley, and V. Misra, “A self-tuning
structure for adaptation in TCP/AQM networks,” ACM SIGMETRICS
Performance Evaluation Review, Vol. 31, No. 1, pp. 302-303, June 2003.
[9-RED] W. Fang, Kang G. Shin, Dilip D. Kandlur, and D. Saha, ”The
BLUE active queue management algorithms,” IEEE/ACM Transactions on
Networking, Vol. 10, No. 4, pp. 513-528, Aug. 2002.
Neal N. Xiong @ GSU Slide 23
June 25, 2011
References[10-RED] Y. Gao and J. C. Hou, “A state feedback control approach to
Stabilizing queues for ECN-enabled TCP flows,” in Proceedings of IEEE
INFOCOM 2003, Vol. 3, pp. 2301-2311, San Francisco, CA, March 30 –
April 2, 2003.
[11-RED] Chonggang Wang, Bin Li, Y. Thomas Hou, Kazem Sohraby, Yu
Lin, ”LRED: A Robust Active Queue Management Scheme Based on
Packet Loss Ratio,” Proceedings of IEEE INFOCOM 2004, Hong Kong,
March 2004.
[12-RED] C. V. Hollot, V. Misra, D. Towsley and W. B. Gong, “Analysis
and design of controllers for AQM routers supporting TCP flows,” IEEE
Transactions on Automatic Control, Vol. 47, pp. 945-959, June 2002.
Neal N. Xiong @ GSU Slide 24
June 25, 2011
References[13] Z. Zhao, S. Darbha, and A. L. N. Reddy, “A Method for Estimating
the Proportion of Nonresponsive Traffic At a Router,” IEEE/ACM
Trans. on Networking, vol. 12, no. 4, pp. 708–718, Aug. 2004.
[1 book] Craig Partridge, “Gigabit Networking,” Addison-Wesley, ISDN 0-201-56333-9.
[2 book] James F. F. Kurose and Keith W. Ross,
"Computer Networking: A Top-Down Approach", 4th edition, Addison Wesley, (ISBN: 10: 0321497708)
Neal N. Xiong @ GSU Slide 25
June 25, 2011
25
XEx ' XEx ' XEx '
Protecting datacenters must first secure cloud resources
and uphold user privacy and data integrity.
Trust overlay networks could be applied to build
reputation systems for establishing the trust among
interactive datacenters.
A FD technique is suggested to protect shared data
objects and massively distributed software modules.
The new approach could be more cost-effective than using
the traditional encryption and firewalls to secure the
clouds.
Security and Trust Crisis in Cloud Computing
Neal N. Xiong @ GSU Slide 26
June 25, 2011
Computing clouds are changing the whole IT , service industry, and global economy. Clearly, cloud computing demands ubiquity, efficiency, security, and trustworthiness.
Cloud computing has become a common practice in business, government, education, and entertainment leveraging 50 millions of servers globally installed at thousands of datacenters today.
Private clouds will become widespread in addition to using a few public clouds, that are under heavy competition among Google, MS, Amazon, Intel, EMC, IBM, SGI, VMWare, Saleforce.com, etc.
Effective reliable management, guaranteed security, user privacy, data integrity, mobility support, and copyright protection are crucial to the universal acceptance of cloud as a ubiquitous service.
Security and Trust Crisis in Cloud Computing
Neal N. Xiong @ GSU Slide 27
June 25, 2011
Content: Reliable, Performance Distributed file system Bandwidth to Data • Scan 100TB Datasets on 1000 node cluster • Remote storage @ 10MB/s = 165 mins • Local storage @ 50-200MB/s = 33-8
mins • Moving computation is more efficient than moving data • Need visibility into data placement
Neal N. Xiong @ GSU Slide 28
June 25, 2011
Scaling Reliably • Failure is not an option, it’s a rule ! • 1000 nodes, MTBF < 1 day • 4000 disks, 8000 cores, 25 switches,
1000 NICs, 2000 DIMMS (16TB RAM) • Need fault tolerant store with reasonable availability guarantees • Handle hardware faults transparently
Neal N. Xiong @ GSU Slide 29
June 25, 2011
Hadoop Distributed File System (HDFS)
• Data is organized into files and directories • Files are divided into uniform sized blocks (default 64MB) and distributed across
cluster nodes • HDFS exposes block placement so that computation can be migrated to data
Neal N. Xiong @ GSU Slide 30
June 25, 2011
Problems of CPU-GPU Hybrid Clusters Scheduling Map tasks onto CPUs and
GPUs efficiently is difficult Dependence on computational resource
# of CPU cores, GPUs, amount of memory, memory bandwidth, I/O bandwidth to storage
Dependence on applications GPU computation characteristic
Pros. Peak performance, memory bandwidth Cons. Complex instructions
Hybrid Scheduling with CPUs and GPUs to make use of each excellence → Exploit computing resources