Special Features:
GGF MAY DRIVE FUTURE OPTICAL NETWORK SPECS
By Alan J. Weissberger, Contributing Editor
I. Summary
On Monday afternoon, Oct. 25, the Optical Internetworking Forum hosted a
half-day workshop with the GGF. The goal was for OIF participants to better
understand Grid network requirements and to identify joint (GGF-OIF) work
items. Attendance was limited to OIF member companies (GGF members had to
belong to an OIF member company in order to attend). There was particular
interest in determining if there was a match between the Grid network
architectures/management/policies and the work that the OIF has already done
on control and management interfaces.
Two speakers were invited to present their views on the GGF and Grid networks.
The OIF Technical Committee chair followed by providing a list of questions on
Grid networks that was to be the basis for an interactive Q & A/discussion
session. The provocative and stimulating discussions that ensued lasted beyond
the official 5 p.m. end time and carried over to the OIF Technical Meeting Oct
26-28. However, specific action was deferred until the next OIF meeting in
January 2005 (two months before the next GGF meeting in March 2005). The OIF
liaison to GGF stated, "Our current plan is to have further internal
discussion and possible follow-up actions at our January 2005 OIF meeting in
Dallas. We are open to any input between now and the meeting. On our side, we
will let you know of any specific OIF items that may be of possible interest
to the GGF."
The OIF Closing Plenary report of the Architecture and Signaling WG stated,
"GGF may become users/requirement drivers for OIF Implementation Agreements."
Note: all the OIF presentations and contributions are available only to those
affiliated with OIF member companies. The GGF drafts are publicly available
from the GGF at: www.ggf.org/documents/.
II. GGF Presentation by Franco Travostino of Nortel, chair of the GHPN (Grid
High Performance Network) WG
Franco opined that GGF was more diverse and effective than the IETF (which he
has also participated in). For that reason, he thought that OIF and GGF could
work very well together. Here are a few highlights from Franco's presentation:
- Grids will be a real business -- Grid global spending to hit $4.8 billion
in 2008 (Source: Insight Corp June 2003 report) with intra-Grids at $2.2
billion, extra-Grids at $1.3 billion, inter-Grids at $1.3 billion)
- The Grid Opportunity: "Resource sharing and coordinated problem solving in
dynamic, multi-institutional virtual organizations" (Ian Foster, ANL).
- Grids provide "a new turn of the crank for distributed systems." They are
the common foundation for the Virtual Organization:
- Grids co-ordinate on demand, secure access to distributed and
heterogeneous resources (CPU, storage, bandwidth/ network access).
- Grids use standard, open, general purpose protocols and interfaces
(developed and published by the GGF).
- Grids deliver non trivial QOS to the application.
- What's special about Grids? How do they differ from previous distributed
systems?
- User knows of a resource pool; the pool (but not its constituents)
knows about the user. However, there is not a prior knowledge of
specific resources by the user.
- Capability to "gang-schedule" a subset of a resource pool (CPU,
storage, sensors, etc).
- It may straddle over administrative boundaries, trust boundaries,
and large distances (Editors Note: early deployment of enterprise
Grids is usually contained within a single administration and trust
cloud. This is because Grid security has not been standardized yet and
is dependent on WS Security roadmap of IBM-MSFT).
- Dynamic matchmaking in time space between "virtual organizations"
and resource pools.
- Two important questions for telcos/ network providers interested in Grids:
- Will you use Grids internally -- for billing, accounting, inventory
management, and other operations support systems or management/control
functions?
- Do you plan to offer any connectivity services in support of Grids?
Will it be a managed or unmanaged service? Will it be combined with
compute/storage/Grid applications software?
- Two specific goals of the GHPN RG are to identify:
- Grid application requirements and implementations that are not
supported or understood by the networking community.
- advanced networking features that are not being utilized by Grid
applications.
- EScience is the dominant Use Case for the GHPN: early Grid adopters are
from this space.
- Three types of physical networks were considered by the GHPN RG, in support
of Grids:
- Wireless systems: mobility, resource constraints, etc.
- Optical networks: Circuits, scheduled connection services,
diversity routing.
- Sensor networks: CERN LHC and Berkeley's Motes are two kinds of
sensor-based Grid networks.
- Two GHPN drafts are in the final ballot state, prior to becoming a GGF GFD:
- draft-ggf-ghpn-opticalnets-2: Convergence of Optical Networks and
Grids.
- draft-ggf-ghpn-netissues-4: Pain points in attaching Grids to the
network.
- Three more GHPN Drafts in the making:
- draft-ggf-ghpn-transportsurvey-1: a survey of L4 protocols for bulk
data transfer other than TCP (to overcome TCPs thruput deficiencies
under heavy load).
- draft-ggf-ghpn-netservices-usecases-2-3: Black-box use cases of
Grids setups exploiting network behaviors, and quantitative analysis
of the same.
- draft-ggf-ghpn-netservices-1: Scoping of Grid Network Services that
elevate the network resource to being a Grid Managed Resource akin to
processing and storage.
- Network related Grid specifications from GGF (selected by this editor):
- GFD.7: A Grid Monitoring Architecture.
- GFD.23: A Hierarchy of Network Performance Characteristics for Grid
Applications.
- GFD.7: Overview of Grid Computing Environments (for reference
only).
- Surfnet 6 (Netherlands) was presented as an example of a hybrid optical and
packet switched Grid network:
- Based on customer-owned and managed dark fiber.
- Native IPv4, IPv6 and light path provisioning over a single
transmission infrastructure.
- Light path provisioning using a single control plane: collapsed 20
routed locations to two locations. Enables users/apps to make service
level changes for their layer 1 or 2 paths, while providing excellent
quality on high speed (1-10 Gb/s) point to point paths. Light paths
not constrained by traditional framing, routing and transport
protocols. Enables creation of optical virtual private networks
(future).
- Managed dark fiber infrastructure to be extended with new routes.
- Observation: Testbeds show two Control Plane approaches:
- A fixed optical mesh between users with slow "automated fiber patch
panel" switching (OBGP).
- A shared optical "cloud" with rapid switching between users (GMPLS,
ASON, Optical Burst Switching). (The OIF UNI is based on GMPLS/ASON
approach for point to point light path provisioning and rapid
restoration).
- Grid User Network Interface (GUNI) is seen building upon and extending the
OIF UNI. Some additional requirements:
- One important missing piece -- not now included in the OIF UNI -is
Scheduled (vs immediate) light path connectivity.
Dynamic network resource allocation and reservation are both
needed.
(Editors Note: OIF did consider scheduled point to point connections, but
deferred to a future version of the OIF UNI/NNI.) - Security controllability to provide a trusted and efficient
communication environment where required.
- High availability when expensive computing or visualization
resources have been reserved.
- Multicast to efficiently distribute data to group of resources.
- Other issues have been pointed out by the networking community:
- What is the Grid traffic impact on infrastructures and other
traffic types?
- How to integrate wireless network and sensor networks in Grid
environment?
III. Presentation by John Strand of AT&T Labs
** Disclaimer: The views expressed do not necessarily represent AT&T's
positions.
John's key observations:
- Grid Leaders recognize importance of "agile" optical networks (ONs).
- e-Science is making heavy use of optical networks; mostly over government
funded fiber networks.
- GGF GHPN Research Group is actively pursuing ON opportunities ("Optical
Network Infrastructure for Grid" -- draft-ggf-ghpn-opticalnets, and
"Networking Issues for Grid Infrastructure" -- draft-ggf-ghpn-netissues).
- Grid community is largely unaware of/indifferent to industry directions
(new SONETSDH features, GMPLS, ASON, O-UNI, O-NNI).
- GHPN optical "Center of Mass" is research-driven and focused on future:
- all-optical networks.
- optical burst/packet switching.
(Editors Note: These might be considered "bleeding edge" vs. leading edge
technologies. They are currently a solution to the day after tomorrow's
networking infrastructure.)
John's proposal for moving Optical Grid networks forward:
- An attractive Optical Grid networking offering could be assembled TODAY
using standardized capabilities that are being widely deployed in carrier
networks.
- Key Features might include:
- bandwidth up to 10-40 gbps in 50 Mbps increments.
- very low jitter/delay/lost data.
- deterministic throughput/delivery times.
- improved security -- TDM, separate control/data planes.
- agile provisioning with connect times ~ 1 second.
- existing connectivity to most metro areas In United States and
Europe.
- powerful, mature OAM tools.
- multiple administrative/control relationships possible.
- customer-owned (and operated).
- customer-controlled resource management.
John's suggestion for OIF work on GGF networks:
Grid optical network connectivity might make use of: Virtual Concatenation,
Optical VPN, O-UNI, O-NNI, Security extensions, Generic Framing Procedure (for
encapsulating Physical or Link layer protocols over SONET/SDH framed optical
networks), ON discovery and management.
(Editors Note: the Optical VPN work is being done in ITU SG13. There appears
to be a lot of interest in this from network providers that desire faster
throughput and lower transit delay then with an IP-MPLS VPN. This is because
optical VPNs are based on Physical layer switching and forwarding vs. MPLS
tag-based forwarding.) - GGF Seems Like A Very Appropriate Partner For Collaborative work on:
- Problem definition.
- technical collaboration.
- interop agreement specification (possibly).
- interop demo (possibly).
- OIF should treat GGF As "Early Adopters" of advanced optical
networks.
IV. Interactive Discussion led by Jim Jones of Alcatel, and OIF TC Chair
Among the numerous questions Jim proposed, the most important (in this
editor's opinion) were probably:
- What types of end users are most significantly driving deployment of Grids
that would require dynamic optical interconnection (not just supplying
capacity)?
- Is there any analysis or specification about the anticipated
characteristics of end user traffic, including:
- Physical interface types and layers used.
- Bandwidth granularity.
- Holding time of connections.
- Latency requirements.
- Pre-emptability of traffic (e.g. after protection/ restoration).
- Predictability (e.g. whether connections could be scheduled in
advance).
- Will the end user needs result in traffic types having different
constraints that may result in more flexible resource sharing?
Is Grid computing targeted more at a traditional carrier model (where users
contract for connection services with a carrier and pay based on usage) or an
enterprise model (where users buy or lease the assets and do not pay a usage
fee)?
(Editors Note: this is the key question that network providers need to ponder:
will they provide Grid specific connectivity and/or a managed Grid network
service? Or, will they simply sell or lease dark fiber to enterprise users who
will then set-up, manage and control their own private network?) - How could the Grid application be opened to a broader group of users --
mainly to small, medium and global business customers?
- Today, Grids are somewhat limited to special, government-funded
research or academic networks.
- This relates strongly to the need for open networking and network
support for Grids.
- What are the main impacts of the global Grid on network operations and
management (OA&M)?
- Does it have potential for cost reductions?
- Will the global Grid bring about a significant paradigm shift in
network management?
- If so, what are the major challenges for migrating from the current
management environment to one more suited for the global Grid?
- What are the Security requirements of Grid users?
- Transport plane security.
- Control plane security.
- Interfaces from NE to management systems or Grid middleware.
Postscript
While these questions provided "food for thought," there was insufficient time
to properly address any of them during the workshop. This editor suggested
that the first questions that needed to be answered were related to the
business model -- an OIF focus on the carrier model of providing optical Grid
networks (vs. the enterprise/private network model) was recommended. I also
suggested formation of a "Carrier RG" within the GGF to Franco and noted that
carriers met privately at GGF12 in Brussels.
About Alan J. Weissberger
As the founder and Technical Director of Data Communications Technology (DCT),
a technical consulting firm started in March 1983, Alan J. Weissberger
specializes in telecommunications standards and their implementation. His
clients have included network providers (AT&T, NTT, Pacific Bell, US West,
Entel and CTC in Chile, Telkom South Africa, Moroccan PTT, others), equipment
and semiconductor manufacturers, and large end users. In 1995 and 1996 Alan
was the principal architect for the European Commission's multi-service,
multi-country ATM network -- the largest private network in Europe (that
network has now evolved into Gig Ethernet over CWDM). In 2000-01, he was
Ciena's lead ITU-T delegate, contributing to the standardization of the
optical control plane in SG13 and SG15. Alan now represents NEC Corp in
several OASIS TCs dealing with Web Services, while also attending the Global
Grid Forum and the Optical Internetworking Forum (OIF).
Weissberger can be reached via e-mail at aweissberger@sbcglobal.net or
ajwdct@technologist.com. To read his entire biography, please visit
www.gridtoday.com/04/1011/bio.html.
|
