EU Phosphorus Project
Harmony
Advance Reservation features for Grids
(on behalf of UniBonn, SURFnet, UvA, CRC, Nortel, FHG, I2CAT)
E2E Workshop – Establishing Lightpaths
Amsterdam, December 2008
Phosphorus Architecture
HARMONY
Harmony system is…
• An multi-domain path provisioning system where users and Grid
applications can book in advance end-to-end paths with AA
• The Harmony Service Interface (HSI) allows the user, Middleware or ever
applications operate the whole inter-domain scenario
• Harmony allows heterogeneous domain interoperability by performing an
inter-domain resource brokering over the heterogeneous Network Resource
Provisioning Systems (NRPS) underlying
• NRPS considered are:
• Argia/UCLP (CA-CRC/Inocybe, Spain-i2CAT)
• ARGON (Germany –Viola Testbed)
• DRAC (Holland- Nortel/SURFnet)
Grid applications in Phosphorus
WISDOM - Wide In Silico Docking On Malaria
ƒ WISDOM consists of virtual screening techniques AutoDock and FlexX. both
are in silico docking techniques, where researchers use large computing power
(Grid systems) to compute the probability that potential drugs will dock with a
target protein.
ƒ Distribution of high volume input data from the specified Input data Server
ƒ High bandwidth need and low latency for data distribution of up to 240 GB
KoDaVis –Making Atmospheric Processes
visible
ƒ Simulations of physical and chemical processes in the atmosphere help to
understand the effect of human activities on the climate: data sets ~ 1000
GigaByte stored on 1 or more central servers
ƒ Collaborative visualization of data provides insight into processes
TOPS - Technology for Optical Pixel Streaming
ƒ Streaming of high resolution 2D scientific content
ƒ Huge data source (~50 TB), not practical to transfer data itself
ƒ Interactive queries on data source. Results (images) are streamed to scientist
for analysis
ƒ Visualisation of huge data-sets on a tiled panel display (TPD). The rendering
is performed at a source node
ƒ Requirements: 10 Gbit/s, 100 msec latency, layer 2, MTU 9000
ƒDDSS – Distributed Data Storage System
ƒ Huge data volumes moved over the network in parallel TCP/IP streams
ƒ Transfers both latency- and bandwidth-dependent
ƒ Single network links allocation for one-to-one scenarios
Harmony testbed
HARMONY
Architecture
Key points:
• Distributed (P2P) or
Hierarchical architecture
for the Network Service
Plane (NSP)
• The NSP is composed by
independent entities
(Inter Domain Brokers)
• Harmony Service
Interface common for the
adaptation layer and the
network service plane
Harmony Service Interface – HSI
Harmony AAI – Overview
ƒ The Harmony System implements an
Authentication (AuthN) and Authorisation
(AuthZ) Infrastructure based on the
Generalised AAA Toolkit [1].
ƒ AuthN
• Based on user certificate + user signature.
• PKI-based, using certificate X.509.
• Signature is exchanged using SAML assertions
among entities.
• Signature added as part of the SOAP header in
the service request message.
ƒ AuthZ
• Access control based on XACML obligations
using local policy databases.
• Implemented using GAAA-Toolkit (ver. 0.5).
• Session is kept exchanging tokens among
entities (token := GRI, value, validity)
[1] http://staff.science.uva.nl/~demch/projects/aaauthreach
• Malleable reservations stand behind the user
intention for transfering a certain amount of data within a
given deadline.
bandwidth
Advance Reservation Algorithm
• When a reservation request is received, the IDB loops
over different interdomain paths, possible start times
and bandwidths (inside user’s constraints range) to
find a set of available resources that fulfill the request.
time
• The steps followed by the IDB are:
earliest start
time
deadline
1. Path computation and choice of bandwidth – In a first step, a path between the
given user endpoints is computed. Then, only bandwidths in the range requested are chosen.
2. Bandwidth adaptation – The attempt here is made to adjust all the endpoints on the
given path to a common bandwidth level. First, it chooses the min-max bandwidth following
user’s constratins. In the second step, the highest bandwidths are reduced according to the
technology-induced granularity until the smallest bandwidth is equal to or greater than the minmax bandwidth.
3. Availability checking – In this step the availability of the network resources is checked.
4. Reservation – The final step commits the fixed reservation constraints that fulfils the original
malleable reservation request.
Harmony and the outer world
Current implement.:
• Harmony-IDC
gateway (signaling
demonstrated on
SC08, booth #2603)
• Harmony-G2MPLS
(signaling, working on
topology)
• HarmonyAutobhan (under
design, using IDC)
Key points:
• For any integration it is necessary to build an Harmony Gateway (HG), with the HSI on the
one hand, and the interface of the other system in the other hand
• HG translates requests in one system-language to HSI, bi-directionally. Modularity in the
internals of HG allow stateful or stateless interoperability, depending on HG implementation
• NSI OGF participation
Harmony-IDC Gateway
1. IDB in Harmony receives reservation request.
2. IDC is seen as a simple domain (limited IDB –request forwarding-)
3. Reservation request to IDC domain is sent to the Gateway
4. Gateway translates request from Harmony to IDC and initiates the reservation procedure in
IDC control plane.
Harmony administrator GUI
Admin GUI
Thanks for your attention
Sergi Figuerola (i2CAT Foundation, Barcelona)
[email protected] or +34.675.780.950