1 / 40

High Performance Computing – Beowulf

High Performance Computing – Beowulf. Robert Whitten Jr. Welcome!. Today’s Agenda: Beowulf. Pop Quiz. What is a beowulf cluster? What are the advantages of a beowulf cluster? What are the disadvantages of a beowulf cluster? How would you create a beowulf cluster?. What is beowulf?.

dixon
Télécharger la présentation

High Performance Computing – Beowulf

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. High Performance Computing – Beowulf Robert Whitten Jr

  2. Welcome! • Today’s Agenda: • Beowulf

  3. Pop Quiz • What is a beowulf cluster? • What are the advantages of a beowulf cluster? • What are the disadvantages of a beowulf cluster? • How would you create a beowulf cluster?

  4. What is beowulf? • Mythical Old-English hero who defeats Grendel, the green dragon • Movie starring a CG-animated Angelina Jolie • Cluster architecture

  5. Beowulf cluster history • 1993 - Becker & Sterling started Beowulf project • 1994 – Becker & Sterling constructed a 16 node cluster • 1996 – DOE and NASA demonstrate beowulf cluster that surpasses 1GFlop/s • 1997 – Caltech demonstrates beowulf cluster that surpasses 10GFlop/s • Present – beowulf clusters are used at many universities, industries, and government sites

  6. What is a beowulf cluster? • Collection of stand-alone computer networked together • Network equipment to connect computers • Software • Operating system software • Application software • Communication software (MPI, OpenMP)

  7. Hardware • Any old computer will do • Laptops • Desktops • Playstation 3s ?? • Network • Switches / hubs • Cable • Ethernet adapters (most computer already have this, including the PS3)

  8. TOPOLOGY Reference: www.teach-ict.com/, uva.ulb.ac.be/

  9. Star Topology • Ethernet • IEEE 802.3 • CSMA / CD • Switch vs. Hub • Broadcast vs. Cut-Through • Cable: CAT V, UTP, RJ45

  10. Ethernet Cable Reference: www.uga.edu/netinfo/netguide/

  11. Model for Networking OSI Model – Developed by ISO O Open I International S System S Standards I Interconnect O Organization

  12. OSI Model

  13. Data Flow Reference: http://catalyst.washington.edu/help/computing_fundamentals/networking/img/osi_model.jpg

  14. Back to The Model

  15. Post Office Analogy What is the purpose of creating a network? How did we used to communicate (get a message from one person to another)? What scheme was used to accomplish this?

  16. Layer 1 : Physical(Bits across Wire) This is the Data – The message you need to send across the way.

  17. Layer 2: Data Link(MAC Address and LLC) Bank of Granny Corporation Receiver Dear Grandma, Please send Money! Sender How are you going to get this letter to the Post Office? Broke Me

  18. Layer 3: NetworkIP Addressing and Best Path Selection Grandma 1010 Geek Drive Hexville, OH 11110 Sender • Regular Stamp ? • 2nd Day Air ? • Overnight ? Receiver

  19. Layer 4: TransportTCP vs. UDP • How do you know Grandma got your Letter? • How does Grandma know you got the Money? • Certified Mail – Cost more, and can take more time, but ensures delivery.

  20. Layer 5 : SessionStart & Stops the Session When the Flag is Down – What does the mail carrier do? When the Flag is Up – What does the mail carrier do

  21. Layer 6: Presentation Data Represented Looks Greek To Me ! ! !

  22. Layer 7: Application(Email, File Transfer, Web Page) • What did you send? • Does it make a difference ? • What are some different things you can send in the mail? • Do different items need different packaging?

  23. Basic Network Diagram

  24. TCP/IP: Addressing Scheme IP = Internet Protocol Protocol is What? Set of Rules 4 Main Rules of IP • Dotted Decimal Notation • No Decimal Number can be >255 • Must Have Subnet Mask • Broken Down into Classes based on High order of bits in the first octet.

  25. Rule #1 : Dotted Decimal Notation Example of IP address: 192.16.32.5 255.255.255.0 IP SM Dotted . Decimal . Notation . Rule Remember – 4 Octets , a Decimal number that represents the value of an 8 bit Binary number

  26. Rule #2 : No Number can be 255 Sample IP Address: 192.16.32.5 255.255.255.0 IP SM The first Octets = 192 & 255 What is 192 in Binary? What is 255 in Binary?

  27. Subnet Mask Divides IP Address Into Two Parts Grandma 1010 Geek Drive Hexville, OH 11110 Sender • Regular Stamp ? • 2nd Day Air ? • Overnight ? Host Receiver Network

  28. Rule #4 : IP Address Classes

  29. Network Bits vs. Host Bits Sample IP Address: 192.16.32.5255.255.255.0 N H Remember This Guy ? By Performing a Binary AND Function IP Address AND Subnet Mask Network Bits 1 1 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0

  30. Network Bits vs. Host Bits Sample IP Address: 192.16.32.5 255.255.255.0 N H Remember This Guy ? By Performing a Binary AND Function IP Address AND Subnet Mask Network Bits 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

  31. Research Question: • How Many Networks can you have in a Class A Network ? Clue # 1 • It has to do with the number of Network bits Clue #2 • It has to do with an exponent Clue #3 • Yep – it has to do with binary!

  32. Answer: Use This Equation to figure out how many networks there are in a Class: 2n– 2 (N = number of bits used) Class A – Network bits used are 8 28 – 2 = ? Answer = 126

  33. Any Questions ? • Why take away 2? Because the 1st and Last networks are reserved: 1st Network: 0.10.10.1 – reserved Last Network: 127.0.0.0 – reserved for loopback testing. • We only take away 2 in a Class A address … The other Classes use the equation 2n

  34. Research Question: Can you figure out the number of hosts available for each network? Clue # 1 • It has to do with the number of Host bits Clue #2 • It has to do with an exponent Clue #3 • Yep – it has to do with binary! Clue #4 • Yep – it is the same equation! 2n - 2

  35. Remember This? Sample IP Address: 192.16.32.5 255.255.255.0 N H Remember This Guy ? By Performing a Binary AND Function IP Address AND Subnet Mask Network Bits 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

  36. Any Questions ? • Why take away 2? Because the 1st and Last Host bits are reserved: 1st Host bit: Network ID Last Host bit: Broadcast ID • Network ID = All Host bits will be all binary 0’s • Broadcast ID = All Host bits will be all binary 1’s

  37. SAMPLE IP Address: Sample IP Address: 192.16.32.5 255.255.255.0 N H By Performing a Binary AND Function IP Address AND Subnet Mask 11000000.00010000.00100000.00000101 11111111.11111111.11111111.00000000 11000000.00010000.00100000.00000000 192 . 16. 32. 0

  38. Sample IP Address: 11000000.00010000.00100000.00000101 11111111.11111111.11111111.00000000 11000000.00010000.00100000.00000000 192 . 16. 32. 0 192.16.32.0 = Network ID 192.16.32.255 = Broadcast ID 192.16.32. 1 – 254 = Host Addresses

  39. Software • Operating system software • Usually Linux-based • Typically has all the software pieces needed • Microsoft does have Windows for clusters • Application software • Not many commercial applications are parallel • Usually custom built applications • Something like homework 3 • Communication software • MPI or OpenMP or something else

  40. Questions? http://www.nccs.gov Oak Ridge National Laboratory U. S. Department Of Energy40

More Related