Blogging in the Clouds Intel Blogger Day 2008 Sivadon Chaisiri http://www.blognone.com/user/javaboom http://javaboom.wordpress.com
Hot HPC Topics
Hot HPC Topics • Cloud Computing: Buzz- and Fuzz (y) Word!
Hot HPC Topics • Cloud Computing: Buzz- and Fuzz (y) Word! • Virtualization:Virtualized Season
Hot HPC Topics • Cloud Computing: Buzz- and Fuzz (y) Word! • Virtualization:Virtualized Season • Green IT: Think Green, Do Green, Feel Green
Hot HPC Topics • Cloud Computing: Buzz- and Fuzz (y) Word! • Virtualization:Virtualized Season • Green IT: Think Green, Do Green, Feel Green •
Parallel Programming is Important
Cloud Computing: Buzz- and Fuzz (y) Word!
Cloud Computing: Buzz- and Fuzz (y) Word! “... computation may someday be organized as a public utility ...”
John McCarthy, 1960
Cloud Computing: Buzz- and Fuzz (y) Word! “... computation may someday be organized as a public utility ...”
John McCarthy, 1960
“Cloud Computing can take on different shapes depending on the viewer, and often seems a little fuzzy at the edges.”
JamesO’Brien
Cloud Computing: Buzz- and Fuzz (y) Word! “... computation may someday be organized as a public utility ...”
John McCarthy, 1960
“Cloud Computing can take on different shapes depending on the viewer, and often seems a little fuzzy at the edges.”
JamesO’Brien
“A cloud is a pool of virtualized resources that can host variety of different workloads, allow workloads to be deployed and scaled-out quickly, allocate resources when needed, and support redundancy. ”
Greg Boss et al., IBM
Cloud Computing: Buzz- and Fuzz (y) Word! “... computation may someday be organized as a public utility ...”
John McCarthy, 1960
“Cloud Computing can take on different shapes depending on the viewer, and often seems a little fuzzy at the edges.”
JamesO’Brien
“A cloud is a pool of virtualized resources that can host variety of different workloads, allow workloads to be deployed and scaled-out quickly, allocate resources when needed, and support redundancy. ”
Greg Boss et al., IBM
My Own Definition
•
Visualizable, Configurable, Scalable, and also Payable Internet-based Computing Model
Public Clouds vs Private Clouds
Something about Private Clouds
Something about Private Clouds •
Need more benefits so invest more resources
Something about Private Clouds • •
Need more benefits so invest more resources Buy more and own more (Good ?)
Something about Private Clouds • • •
Need more benefits so invest more resources Buy more and own more (Good ?) Objective: ROI [+] while TCO [-]
Something about Private Clouds • • •
Need more benefits so invest more resources Buy more and own more (Good ?) Objective: ROI [+] while TCO [-]
•
TCO = Investment (budget, energy, man power, maintenance, ...)
Something about Private Clouds • • •
Need more benefits so invest more resources Buy more and own more (Good ?) Objective: ROI [+] while TCO [-]
• •
TCO = Investment (budget, energy, man power, maintenance, ...) ROI = Benefits
Something about Private Clouds • • •
Need more benefits so invest more resources Buy more and own more (Good ?) Objective: ROI [+] while TCO [-]
• •
•
TCO = Investment (budget, energy, man power, maintenance, ...) ROI = Benefits
Let’s see
Something about Private Clouds • • •
Need more benefits so invest more resources Buy more and own more (Good ?) Objective: ROI [+] while TCO [-]
• •
•
TCO = Investment (budget, energy, man power, maintenance, ...) ROI = Benefits
Let’s see
•
Need someday but own forever
Something about Private Clouds • • •
Need more benefits so invest more resources Buy more and own more (Good ?) Objective: ROI [+] while TCO [-]
• •
•
TCO = Investment (budget, energy, man power, maintenance, ...) ROI = Benefits
Let’s see
• •
Need someday but own forever You are a new investor / a small business
Something about Private Clouds • • •
Need more benefits so invest more resources Buy more and own more (Good ?) Objective: ROI [+] while TCO [-]
• •
•
TCO = Investment (budget, energy, man power, maintenance, ...) ROI = Benefits
Let’s see
• • •
Need someday but own forever You are a new investor / a small business Trial and Error!
Timeline Grid Computing Volunteer Computing
(e.g., GIMPS and SETI@home)
Globus Toolkit (from GT2 GT4) [virtual server]
Early 1990s
Utility Computing Software-as-a-Service Cloud Computing HP’s Utility Data Center (UDC) Sun Grid Computing Utility (Network.com)
(SaaS)
Google App Engine Amazon EC2
Google Apps
IBM Blue Cloud
saas-showplace.com
Microsoft SSDS GoGrid Singapore’s AxSaaS
[virtual platforms]
[virtual applications]
Late 1990s
2000
[virtual things]
2008 Reference: wikipedia.com
Learning from the Past Grid Computing
Utility Computing
[virtual server]
[virtual platforms]
Software-as-a-Service
[virtual applications]
Learning from the Past Grid Computing
Utility Computing
Software-as-a-Service
“A single VO with a single virtual servers” VO = Virtual organization
PO = Physical organization
Advantage • Collaboration • Sharing • Saver • Independent policies / PO Disadvantage • Inflexible in other POs • Need the isolation • Who’ll give me more • Who’ll take care of me • No incentive [virtual server]
[virtual platforms]
[virtual applications]
Learning from the Past Grid Computing “A single VO with a single virtual servers” VO = Virtual organization
PO = Physical organization
Advantage • Collaboration • Sharing • Saver • Independent policies / PO Disadvantage • Inflexible in other POs • Need the isolation • Who’ll give me more • Who’ll take care of me • No incentive [virtual server]
Utility Computing
Software-as-a-Service
“An economic model: how to manage idled time (80%)” Advantage • Sell your garbage Disadvantage • Inflexible (on customer-side)
[virtual platforms]
[virtual applications]
Learning from the Past Grid Computing “A single VO with a single virtual servers” VO = Virtual organization
PO = Physical organization
Advantage • Collaboration • Sharing • Saver • Independent policies / PO Disadvantage • Inflexible in other POs • Need the isolation • Who’ll give me more • Who’ll take care of me • No incentive [virtual server]
Utility Computing “An economic model: how to manage idled time (80%)” Advantage • Sell your garbage Disadvantage • Inflexible (on customer-side)
[virtual platforms]
Software-as-a-Service “Less on-premise Less budget” Advantage • Less (or no) software development • Less (or no) hosting • Less maintenance (e.g., update) • Choose the right app Disadvantage • Limitation on the other sides • Only applications
[virtual applications]
• • • • • • •
Cloud Computing
Virtualized Scalable Flexible
Multi-tenancy SLA-based Self-healing Pay-per-use
Google Trends : 2004 - October 2008
Reference: http://www.google.com/trends, Oct 7, 2008
Google Trends : 2008
Reference: http://www.google.com/trends, Oct 7, 2008
Google Trends : 2008
Reference: http://www.google.com/trends, Oct 7, 2008
Google Trends : 2008 (Only Grid)
สวัสดี
Reference: http://www.google.com/trends, Oct 7, 2008
Players in the Clouds Cloud Customers Cloud Distribution * (e.g., Akamai, LimeLight, CDNetworks) Software-as-a-Service (SaaS)
Google Apps, Microsoft S+S
Platform-as-a-Service (PaaS)
Google App Engine, Force.com, IBM IT Factory, 10gen, Mosso
Infrastructure-as-a-Service (IaaS) data-Storage-as-a-Service (dSaaS)
IBM Blue Cloud, Amazon EC2, GoGrid, Sun Grid Amazon S3, Nirvanix SDN, Flickr, Mosso Cloud FS
Cloud Providers **
Cloud Supporters Platform VMO, Citrix XenServer, MS Windows Cloud, MS VMM, Citrix Cloud Center, Intel VT vs AMD-V, VMware VDC-OS, Google MapReduce, Intel Parallel Studio, Oracle Database 11g and etc.
* Based on “Benefiting from Cloud Compuitng Technology Trends”, Gear6 ** Based on “Cloud Computing with Linux”, IBM
Success Stories Amazon S3
Amazon EC2
Success Stories
Blognone’s News • • • • • • • • •
พบกับ Windows บน Amazon EC2 เร็วๆนี้ เปิดหน้าต่างมองดูกลุ่มเมฆผ่าน Windows Cloud ไอบีเอ็มทุ่ม 300 ล้านตั้งศูนย์ข้อมูล 13 แห่ง Oracle จับมือ Intel ทะยานสู่กลุ่มเมฆ บิล เกตส์ชี้ Cloud Storage ไปได้รุ่งกว่า ห้าเทคโนโลยีแห่งปีในกระแส Hype Cycle รายงาน Google Developer Day 2008 (กลยุทธ์ Client Cloud Connectivity) ศึกสามก๊กมหาสงครามกลุ่มเมฆ Microsoft VMM, VMware VDC-OS, และ Citrix’s C3 โปรดติดตามต่อไป
Ten Challenges 1. Security
6. Power Consumption
2. Logging
7. Latency
3. Platform-dependency
8. Legacy Internal Cloud
4. Reliability
9. National Laws
5. Standard (Inter-Cloud)
10. Internal Politics
Based on Stacey Higginbotham, “10 Reasons Enterprises Aren’t Ready to Trust Cloud”, Gigaom
Virtualization: Virtualized Season
• Characteristics • Applying the Virtualization • Virtualized Resources • • • Heterogeneity • • Snapshot • • Migration • Server Consolidation
Working outside firewall
Sharing & Isolation + Security Resource Planning Load-balancing
• •
Green IT
High Availability
•
Green IT: Think Green, Do Green, Feel Green Green Approaches *
• • • •
Green use Green disposal Green design
•
Blognone’s News •
กูเกิลรักษ์โลกสีเขียว
•
จัดอันดับซูเปอร์คอมพิวเตอร์ประหยัดไฟ
•
Xeon ใหม่ไร้สารฮาโลเจน
•
Data center จะสร้างมลพิษมากกว่าอุตสาหกรรมการบิน ในปี 2020
•
เปลี่ยนแผ่นซิลิคอนเวเฟอร์ให้เป็นเซลล์รับพลังงานแสง อาทิตย์
•
OLPC เป็น "สุดยอดแล็ปท็อปสําหรับสิ่งแวดล้อม" ของโลก
•
โปรดติตตามต่อไป
Green manufacturing
* Reference: San Murugesan “Harnessing Green IT Principle and Pratices”, IEEE ITPro, 2008
Google’s Data Centers
Parallel Programming is Important! Unsustainable (overheating + power consumption) Parallelism is the primary method
Doubling every 18 months
Intel x86 8086, 8088
J. Hennessy and D. Patterson, “Computer Architecture: A Quantitive Approach”, 4th edition, Morgan Kauffman, 2007
Parallel Programming • • • • • • • •
MPI / PVM OpenMP Multithreading Intel Threading Building Blocks Intel Parallel Studio Microsoft Robotics Studio Microsoft Parallel FX Google’s MapReduce / Hadoop
Any Questions ? Contact Me Sivadon Chaisiri http://www.blognone.com/user/javaboom http://javaboom.wordpress.com EMAIL:
[email protected] MSN:
[email protected]