logistical networking for energy sciences
DESCRIPTION
Logistical Networking for Energy Sciences. Scott Klasky Oak Ridge National Laboratory Terry Moore The University of Tennessee. The Internet Backplane Protocol. A common service for state management in a shared network. A basis for asynchronous communication. - PowerPoint PPT PresentationTRANSCRIPT
Presented by
Logistical Networking for Energy Sciences
Scott KlaskyOak Ridge National Laboratory
Terry MooreThe University of Tennessee
2 Beck_LNES_SC07
The Internet Backplane Protocol A common service for state management in a shared
network. A basis for asynchronous communication. Scalability comes from weak assumptions:
Maximum size and duration of allocation. A highly generic, “best effort” service. “A weak network version of malloc.”
Robust services are built on top in an end-to-end manner ! The goal is scalability analogous to the Internet.
2 Beck_LNES_SC07
3 Beck_LNES_SC07
Internet Backplane Protocol
Na
Allocate!
Capability
Depot
Nw
Store!DepotData
Nr
Load!Depot
4 Beck_LNES_SC07
Na
Copy!Capability
Depot
Depot
Nr
Compute!Depot
Internet Backplane Protocol
5 Beck_LNES_SC07
Three kinds of files Files stored locally (1)
Attached disk (direct, LAN or storage network)
Files represented as exNodes exNodes stored locally (2) exNodes stored in LoDN directory (3)
LoDN/L-Store
AppHost
6 Beck_LNES_SC07
One infrastructure serves all
Logistical networking provides a “bits are bits” infrastructure Standardize on what we
have an adequate common model for
Storage/buffer management
Coarse-grained data transfer
Leave everything else to higher layers
End-to-end services: checksums, encryption, error encoding, etc.
Enable autonomy in wide area service creation: security, resource allocation, QoS guarantees…
Gain the benefits of interoperability today!
VideoSatellite
Bioinformatics
Experimentaldevices
Medicalimaging
Simulation
~PByte/sec
7 Beck_LNES_SC07
Fusion: Content distribution
Ewokcluster
Depots
NYU
PPPL
UCI
MIT
Portals
DirectoryserverJaguar
Cray XT4
8 Beck_LNES_SC07
Fusion: Location independent I/O
NYU
PPPL
UCI
MIT
end user
Retrievemetadata
Retrieve data from best-performing depots
Directoryserver
2.
1.
8 Beck_LNES_SC07
9 Beck_LNES_SC07
SRM interoperability
10 Beck_LNES_SC07
11 Beck_LNES_SC07
Contacts
Micah BeckLogistical Computing and Internetworking (LoCI) Lab Computer Science Department University of Tennessee(865) [email protected]
Scott A. KlaskyScientific ComputingNational Center for Computational Sciences(865) [email protected]
11 Beck_LNES_SC07