1. Executive summary - LoCI

vainclamInternet και Εφαρμογές Web

14 Δεκ 2013 (πριν από 3 χρόνια και 8 μήνες)

136 εμφανίσεις


1

DOE/MICS Mid
-
Year Project Report: June 4, 2003

Project Title:

Optimizing Performance and Enhancing Functionality of Distributed Applications
Using Logistical Networking (DE
-
FC02
-
01ER25465)

Project Type:
SciDAC

PI: Institution: Micah Beck (lead), University

of Tennessee, Knoxville

Jack Dongarra, University of Tennessee, Knoxville

James S. Plank, University of Tennessee, Knoxville

Rich Wolski, University of California at Santa Barbara

1.

Executive summary

The multi
-
threaded research project in Logistical Network
ing for the DOE SciDAC project aims to create
advanced, storage
-
enabled network services that can provide reliable, fast, flexible, scalable, and efficient
delivery of data to support distributed and high performance applications of all types. During the
most
recent period, meaningful progress toward this goal was made on all key fronts


research and
development, publication and dissemination, and planning and interaction for productive collaboration
with the SciDAC research community. Highlights include
the following:



Released Logistical Runtime System Tools (LoRS Tools) v0.81, with improved end
-
to
-
end data
compression, encryption, and checksum services. LoRS Tools v0.81 includes access to
DataMover plug
-
in functionality for UDP point
-
to
-
point and multic
ast transfer mechanisms, and a
new version of LoRS View visualization software with an improved graphical user interface.



Released IBP Depot v1.3 which offers improved reliability, dynamic thread control, and IPv6
compatibility to allow IBP depots to run o
n IPv4 only, IPv6 only, or dual IPv4/IPv6 machines.



Developed the Read
-
Only Logistical File System (ROLFS) and the more functionally advanced
Logistical File System (LFS) to enable users to interact with Logistical Networking technologies
using a familia
r, recognizable file system.



Began collaboration with the Terascale Supernova Initiative (TSI). The Logistical File System
(LFS) has been ported with the Hierarchical Data Format (HDF) to allow TSI researchers to take
large data sets generated by advanc
ed computer simulations and output directly to IBP storage.



Deployed the first of five new high
-
powered, dedicated IBP servers to be located at major TSI
research sites around the country. These high performance machines will provide an additional
9.5 T
B of IBP storage space for use by TSI and the entire SciDAC research community.



Published “An End
-
to
-
End Approach to Globally Scalable Network Storage” at SIGCOMM 2002
Conference. One of only two position papers accepted in an extremely competitive field
, this
paper explains the unique architectural vision behind Logistical Networking.



Gave several major public demonstrations of the performance and functionality of Logistical
Networking technology. At the international iGrid2002 conference, multiple stand
ard TCP
streams were used to accomplish data transfers of 100 Mbps from the US to Amsterdam. At
SC2002, a distributed computing application running on NetSolve servers around the world used
Logistical Networking technology to seamlessly access blocks from
distributed data replicas for
enhanced application performance.


2

2.

Current accomplishments

2.1

Currently Deliverable Logistical Networking Technologies

Logistical Networking technologies offer a flexible, highly scalable means to manage distributed content
and d
ata of all kinds using shared network storage. Currently deliverable software tools allow the user to
deploy their own IBP storage “depot(s)” or utilize available public IBP storage deployed worldwide to
easily accomplish long haul data transfers, tempora
ry storage of large data sets (on the order of terabytes),
prepositioning of data for fast on
-
demand delivery, and high performance content distribution such as
streaming video.

Internet Backplane Protocol (IBP)

is a highly scalable, low
-
level mechanism
for managing remote
storage as a sharable network resource through deployment and shared use of lightweight storage
allocations called storage “depots.” IBP is the foundation of the network storage stack and essential to the
Logistical Networking approach
.

The External Node (exNode)

is a generalized data structure, analogous to a UNIX inode, holds metadata
necessary to manage distributed content on IBP depots. ExNodes are used to aggregate IBP storage
allocations and allow file
-
like structuring.

The Log
istical Backbone (L
-
Bone)

directory and resource discovery service catalogues registered IBP
storage depots for an international deployment of 204 depots that serve 15 TB of storage as a shared
resource for the scientific community. A second private direc
tory is being established to serve the ESnet
and SciDAC communities exclusively. This private L
-
Bone implementation currently offers 1.5 TB of
storage, and will grow to 8 TB as planned IBP deployments are accomplished over the next three months.

Logisti
cal Runtime System Tools (LoRS Tools)

software suite uses the underlying capabilities provided by
IBP, the exNode, and the L
-
Bone to implement high
-
level file management capabilities with strong
properties, including high
-
performance access, reliability, a
nd end
-
to
-
end services such as data
compression, checksums, and encryption. The LoRS View visualization tool, included with the LoRS
Tools package, provides graphical representations of LoRS Tools file management capabilities, allowing
the user to view up
load, download, and inter
-
depot data transfers in real time.

DataMovers

are auxiliary IBP depot modules that support all kinds of customized or special purpose
depot
-
to depot
-
transfers, including point
-
to
-
point, point
-
to
-
multipoint, and multicast transmis
sion. Since
the movement of large data sets is of immediate interest to our SciDAC application collaborators, we are
experimenting with depots equipped with high
-
performance data movers for massive, long
-
haul transfers
among remote collaborators (e.g. ORN
L and CERN).

2.2

Research and development

Our most recent research has expanded the effectiveness of Logistical Networking in both performance
and functionality. The entire software tools suite is now available and in use by our collaborators. Below
are some

of our accomplishments:



Released Logistical Runtime System Tools (LoRS Tools) v0.81, with improved end
-
to
-
end
support, including data compression, default DES encryption, and checksum conditioning of
stored data. Includes support for the use of TCP DataMo
ver and reliable UDP DataMover plug
-
in
features.



Released LoRS View v0.80, including an improved graphical user interface with easier to use
location dialogs, more customizable preference parameters, and a “Route” control panel for
executing single
-
threade
d, pipelined routed data augmentations. This new version supports all
features available with the LoRS Tools command line interface.


3



Released IBP Software version 1.3, for IPv6 compliant depots. IBP v1.3 may be run on dual
stacks, IPv4 only, or IPv6 only
machines. New features also include dynamic thread control, new
DataMover plug
-
ins, GNU compliant installation procedure, and greater stability and reliability
due to recent fixes.



Two new DataMover plug
-
in features for increased data transfer performance

are included with
IBP v1.3. The UDP/IP multicast DataMover uses unreliable UDP/IP multicast to accomplish
point to multipoint transfers. The SABUL (Simple Available Bandwidth Utilization Library)
DataMover uses a reliable UDP transfer stream along with
a TCP flow control channel to provide
very high throughput over long
-
haul transfers. SABUL was developed by Robert Grossman and
his group at the University of Chicago and the National Center for Data Mining, who also
collaborated with us in the development

of the SABUL DataMover.



Investigated the efficacy of using on
-
line forecasting to achieve high
-
throughput and reliability
levels without the overhead of redundant, parallel TCP streams.



IBPvo, an experimental personal video recording application, is curre
ntly available for use via the
internet. IBPvo technology can be generalized to provide content delivery services to the
SciDAC community, for instance streaming of large video files.



Read
-
Only Logistical Files System (ROLFS) is currently being tested as
a supporting technology
for IBPvo. ROLFS provides file management services, including automatically refreshing the
time
-
limited IBP storage allocations used to store file content.



Logistical File System (LFS), currently a baseline library/user
-
space file s
ystem, has been ported
with NCSA’s Hierarchical Data Format (HDF) v4.1 scientific data management library to allow
the Terascale Supernova Initiative’s complex modeling applications to release large output data
sets directly to the logistical network.



The
first of five new 1.9 TB high
-
performance dedicated IBP servers, to be deployed at principle
TSI research sites (ORNL, SDSC, Stony Brook, LBL, NCSU) around the country, has been
successfully installed at ORNL and is currently in use by TSI group members.
(More detail in
Section 4, below.)



Three IBP depots located topologically close to the Abilene backbone allow for significant
overlay routing possibilities. Clusters of public depots, thirty
-
one in California and eight in North
Carolina, serve as backup t
ransfer paths for TSI data transfers.

2.3

Application
-
Driven Research

2.3.1

Terascale Supernova Initiative

The TSI project has adopted Logistical Networking as a key component of their data management
strategy. TSI is currently using the LoRS command line tools t
o share large data sets and computational
results between collaboration sites. Previously, using traditional FTP tools, TSI collaborators tolerated
transfer rates of 8 Mbps at best. Using the LoRS tools, they can now transfer data at speeds up to 220
Mbps

between key research sites at ORNL and NCSU. See Figure 1, below.

The LoRS Tools software package allows SciDAC users to store, manage, and retrieve data via the
Logistical Network. The latest LoRS Tools package includes a graphical user interface to a
llow
straightforward mastery of capabilities, as well as the LoRS View visualization tool for viewing data
manipulations in real time. These additions to the LoRS Tools package enhance the usability of the tools
by making them accessible to researchers an
d students at all levels of expertise. The complete LoRS
Tools package is available for download from our website.


4

New features of the LoRS software include dynamic thread control for the download command. While
downloading data, LoRS may now use the p
rogress
-
driven redundancy algorithm to maximize download
performance by making informed decisions about thread allocation.

2.3.2

Logistical File Systems

We are currently focusing our efforts to develop two new Logistical file systems to aid TSI and other
SciDA
C collaborators in managing and sharing large data sets. The Read Only Logistical File System
(ROLFS) is intended to facilitate file sharing across organizational and geographical boundaries. The
Logistical File System (LFS) is a more full
-
featured file
system designed for use on workstations or file
servers. Both of these approaches provide a familiar file system interface for our novel network storage
infrastructure.



ROLFS version 1.0, to be released in early June, uses

a standard client
-
server model to provide an
exNode directory that will allow users to freely share data with the SciDAC community. Full deployment
of a private ROLFS directory for the use of TSI collaborators, with completed performance upgrades, will
b
e in place by the end of the summer. A dedicated ROLFS server will actively manage TSI data set
exNodes and provide a central directory for exNode storage and retrieval.

Figure 1: The diagram below illustrates the way that the components of the Logistical Runtime System Tools
(LoRS Tools) work together in the high performance distribution of data between IBP depots at Terascale
Supernova Ini
tiative sites at ORNL and NCSU.


5

A key feature of ROLFS v1.0 is active exNode management, which relieves the user of
the burden of
monitoring and refreshing stored data. ROLFS performs the scheduled renewal of time
-
limited IBP
storage allocations, as well as maintaining preset fault
-
tolerance and performance levels through
automatic striping and redundancy. ROLFS resto
res degraded allocations by automatically replicating
stored data fragments in order to maintain a minimum number of redundant copies of the data.

When mature, LFS will be a complete file system implementation
with full capabilities to manipulate files
and

directories, including create, open, read, and write functionalities
. LFS provides a much more
scalable and flexible foundation for distributed data management than a traditional file system. LFS
stores data on the logistical network, i.e. on IBP depots
, while traditional file systems store data on local
disk or a file server attached to the local network. LFS transparently handles the tasks of finding an
appropriate IBP storage depot, allocating IBP storage, and storing data on the Logistical Network.

The current implementation of LFS has been ported with NCSA’s Hierarchical Data Format (HDF) v4.1
scientific data management library, to allow complex modeling applications used by TSI and other
SciDAC researchers to release output directly to the Logisti
cal Network. As TSI scales up the size of its
simulations, the capability to output directly to the network without waiting for the entire simulation to
finish will be critical to storing and moving data effectively. HDF is widely used among SciDAC
resea
rch groups, hence using Logistical Networking functionality to enhance HDF will be an important
segue for reaching the broad SciDAC community.

We are also extending LFS to include single
-
writer consistency, automatic replication generation, and
automatic

replica scheduling. This new system will use a dynamic scheduler that forecasts future
performance and availability levels for IBP depots to determine the degree of replication and placement
of replicas necessary to insure a specified availability and pe
rformance goal. The system is being tested
using GridSAT, a Grid enabled satisfiability solver used in curcuit design and verification. GridSAT has
been able to achieve new satisfiability results using dynamically allocated resources. We plan to use our

new file system capabilities, which provide a standard Unix interface for programming ease, to implement
a checkpointing facility for GridSAT.

In the next three months, we intend to make LFS compatible with the Condor project’s Pluggable

File
System (PFS)
. This will allow LFS to be deployed to applications quickly and ease the burden of porting
LFS to operating systems whose file system hooks are different from those provided by Linux.

PFS
capability will allow SciDAC collaborators to take advantage of LF
S without encountering system
incompatibilities.

In addition, we are presently involved in a number of smaller projects related to file systems. Of particular
note is an experimental IBP device driver for Linux. The /dev/ibp device allows applications wit
h no
knowledge of IBP to take advantage of Logistical Networking by simply opening the device file and
reading or writing to it.

2.3.3

Adaptive Forecasting as a Strategy for Robustness

We are investigating efficient networking methodologies, capable of withstand
ing intermittent network
and host failures. The typical approach to maximizing both throughput and connection reliability using
TCP is to use parallel and redundant communication streams. If one stream fails, or becomes slow (due
to ambient network conge
stion) the missing data is fetched or stored over another, better performing
stream. Unfortunately, when redundant data is moved as a precautionary measure and then discarded, this
approach can result in wasted bandwidth

a valuable commodity in today’s ne
tworks. We have
developed a novel approach to managing replicated communication streams that relies on the on
-
line
performance forecasts generated by the Network Weather Service (NWS). By dynamically ranking the
connectivity between data source and sink,

and adaptively discovering appropriate time out values, our
approach achieves higher performance than the redundant streams approach, with the same robustness

6

characteristics, using a small fraction of the bandwidth. We have prepared a paper on the subje
ct and
submitted it for consideration to SC2003.

2.3.4

Integration with Distributed Computing Middleware

Support for distributed computing that facilitates the research efforts of collaboratories and other
advanced applications is an important part of the SciDAC

vision. Our experiments with the integration of
Logistical Networking technology and NetSolve middleware are designed to support that vision.
NetSolve allows remote users, working with familiar interfaces such as Matlab and Mathematica, to
access distribu
ted hardware and software resources in order to perform complex calculations. Integration
with Logistical Networking enables users to store (and replicate) data objects in IBP depots near the
locations of NetSolve servers, and then point NetSolve servers
to these depots to find data to use in
computations. The proximity of the IBP depots to the NetSolve servers, as well as the existence of
multiple replicas that can supply the needed data, will improve the performance of NetSolve, especially
across the wid
e area. The user can thus run computations on remote data and retrieve only the pertinent
portion of the output at the client.

The ability to utilize IBP for remote storage has been incorporated into the current release of NetSolve
(v1.4), but full integra
tion with exNode and LoRS technology is not complete. A prototype version
showing the power of this enhanced functionality was exhibited at SC2002 (see Section 2.4.2 below).

2.4

Major Public Demonstrations

2.4.1

High Performance Data Transfer


iGrid 2002

A demonstr
ation showcasing high performance data transfers using Logistical Networking was given at
iGrid 2002 in Amsterdam, the Netherlands. The presentation began with a brief introduction to the
network storage stack, the hierarchical arrangement of Logistical N
etworking technologies analogous to
the IP stack. After an explanation of the technologies involved, a demonstration was performed in which
video content was streamed directly from IBP storage to a video player for immediate viewing. The
stored video con
tent had been fragmented and distributed over several IBP depots around the world. The
LoRS Tools were used to retrieve the scattered data blocks from storage and reassemble them in the
proper order as content was released to the video player. During the

demonstration, the LoRS View
visualization tool displayed the status of the downloading data blocks and the video stream. LoRS View,
which provides a real
-
time visual representation of individual data manipulations, showed the
downloading of each data bl
ock as it happened and tracked the overall progress of the content stream as it
was released to the video player.

2.4.2

Logistical Networking in Distributed Computing


SC2002

The integration of Logistical Networking technology with NetSolve was demonstrated at

Supercomputing
2002. NetSolve enables users to accomplish complex computations while taking advantage of distributed
resources, by sending work out to a pool of servers scattered around the world. The SC2002
demonstration clearly showed how this process
can benefit from Logistical Networking. Instead of
sending a huge data set directly to NetSolve servers, Logistical Networking allows the user to send only
an exNode, a pointer to data store on IBP depots. The NetSolve server then retrieves the data, ope
rates on
it, and stores the results into a new exNode that is returned to the user. If the results of the first operation
are to be used as input for a second operation, then the necessary data will already be stored near the
NetSolve server and may be re
trieved by the server directly. Enabling this capability will be part of our
future work.

In preparation for the SC2002 demonstration, multiple copies of various matrices were stored in IBP
servers in the US, Europe and Australia; while the NetSolve team
set up several servers in those three
locations. The client then made calls to NetSolve servers in each of the three regions, passing the same

7

exNode to each server. The servers read the exNode and downloaded the data. The exNode contained
location info
rmation about copies of the data spread around the global network, allowing each NetSolve
server to retrieve data from depots of close proximity. In other words, NetSolve servers in California
retrieved matrices from servers in the western part of the US;

servers in Europe requested data mostly
from European depots; Australian servers got data mostly from Australia, and so on. During the live
demo, the LoRS View visualization tool and a NetSolve Monitor were used to show the process in action.

2.5

Publications



S.Y. Mironova, M.W. Berry, S. Atchley, M. Beck, T. Wu, L.E. Holzman, W.M. Pottenger, and D.J.
Phelps,
Advancements in Text Mining
, In "Data Mining: Next Generation Challenges and Future
Directions," H. Kargupta, A. Joshi, K. Sivakumar, and Y. Yesha (Eds
.), AAAI/MIT Press, Menlo Park,
CA, 2003.

M. Beck, T. Moore, and J. S. Plank, "An End
-
to
-
End Approach to Globally Scalable Programmable
Networking," to be presented at Future Directions in Network Architecture (FDNA
-
03), an ACM
SIGCOMM 2003 Workshop, Karls
ruhe, Germany, August 27, 2003 (to appear).

M. Beck, Y. Ding, E. Fuentes and S. Kancherla, “An Exposed Approach to Reliable Multicast in
Heterogeneous Logistical Networks,” the 3rd IEEE/ACM International Symposium on Cluster
Computing and the Grid (CCGrid
2003), Tokyo, Japan, May 12
-
15, 2003.

A. Bassi, M. Beck, T. Moore, and J. Plank, “The Logistical Backbone: Scalable Infrastructure for Global
Data Grids,” Asian Computing Science Conference 2002, Hanoi, Vietnam, December, 2002. Springer
Verlag.

S. Atchley,

M. Beck, H. Hagewood, J. Millar, T. Moore, J. S. Plank, and S. Soltesz, “Next Generation
Content Distribution Using the Logistical Networking Testbed,” Technical Report UT
-
CS
-
02
-
498,
University of Tennessee, Department of Computer Science, December 30, 20
02.

S. Atchley, M. Beck, J. Millar, T. Moore, J.S. Plank, and S. Soltesz, “The Logistical Networking
Testbed,” Technical Report UT
-
CS
-
02
-
496, University of Tennessee, Computer Science Department,
December 16, 2002.

S. Atchley, S. Soltesz, J. S. Plank, and
M. Beck, “Video IBPster,” Technical Report UT
-
CS
-
02
-
490,
University of Tennessee, Department of Computer Science, October 31, 2002.

J. S. Plank, S. Atchley, Y. Ding, and M. Beck, “Algorithms for High Performance, Wide
-
Area,
Distributed File Downloads,” T
echnical Report UT
-
CS
-
02
-
485, University of Tennessee, Department of
Computer Science, October 8, 2002.

K. Meyer
-
Patel and M. Beck, “A Logistical Networking Model for Video
-
On
-
Demand,” IEEE
International Conference on Multimedia and Expo, Lausanne, Switzer
land, August 26
-
29, 2002.

M. Beck, T. Moore, and J. S. Plank, “An End
-
to
-
End Approach to Globally Scalable Network Storage,”
ACM Sigcomm 2002 Conference, Pittsburgh, PA, August 19
-
23, 2002.

A. Bassi, M. Beck, J. Gelas, and L. Lefevre, “Logistical Storage i
n Active Networking: a promising
framework for network services,” the 3rd International Conference on Internet Computing (IC 2002), Las
Vegas, NV, June 24
-
27, 2002.

A. Bassi, M. Beck, G. Fagg, T. Moore, J. Plank, M. Swany, and R. Wolski, “The Internet Back
plane
Protocol: A Study in Resource Sharing,” the 2nd IEEE/ACM International Symposium on Cluster
Computing and the Grid (CCGrid 2002), Berlin, Germany, May 21
-
24, 2002.


8

M. Beck and T. Moore, “Logistical Networking: When Institutions Peer,” the 2nd Interna
tional Workshop
on Global and Peer
-
to
-
Peer Computing on Large Scale Distributed Systems, part of CCGrid 2002, Berlin,
Germany, May 21
-
24, 2002.

A. Bassi, M. Beck, E. Fuentes, T. Moore, and J. S. Plank,

Logistical Storage Resources for the Grid
,”
in
the Pr
oceedings of the International Conference on Computational Science (ICCS 2002), Part II, vol.
2330, LNCS. Amsterdam, the Netherlands: Springer Verlag, 2002.

S. Atchley, S. Soltesz, J. S. Plank, M. Beck, and T. Moore,

Fault
-
Tolerance in the Network Storage

Stack
,”
presented at the IEEE Annual Workshop on Fault
-
Tolerant Parallel and Distributed Systems (held
in conjunction with the International Parallel & Distributed Processing Symposium), Ft. Lauderdale, FL,
USA, April 15
-
19, 2002.

J. S. Plank, M. Beck and

T. Moore, “Logistical Networking Research and the Network Storage Stack,”
USENIX FAST 2002 Conference on File and Storage Technologies, work in progress report, January,
2002.

M. Beck, T. Moore, J. Plank, “Scalable Sharing of Wide Area Storage Resources,”

Technical Report UT
-
CS
-
02
-
475, University of Tennessee, Department of Computer Science, January, 2002.

M. Beck, T. Moore, and J. S. Plank, “Exposed vs. Encapsulated Approaches to Grid Service
Architecture,” presented at 2nd International Workshop on Grid
Computing, Denver, CO, Nov. 12, 2001.

J. S. Plank, A. Bassi, M. Beck, T. Moore, M. Swany, and R. Wolski, “Managing Data Storage in the
Network,” IEEE Internet Computing, vol. 5, no. 5, pp. 50
-
58, September/October, 2001.

2.6

Publications in Process

S. Atchley,

S. Soltesz, J. S. Plank, and M. Beck, “Video IBPster,” accepted for publication in Future
Generation Computer Systems.

S. Atchley, M. Beck, J. Millar, T. Moore, J.S. Plank, and S. Soltesz, “The Logistical Networking
Testbed,” submitted for review by the A
CM Sigcomm CCR.

3.


Future Accomplishments

3.1


Next three months:



Read Only Logistical File System (ROLFS) version 1.0, to be released in early June, provides an
exNode directory that will allow users to freely share data with their community. ROLFS v1.0
will u
se a standard client
-
server model and feature unix
-
like access controls.



Full deployment of a private ROLFS directory for the use of TSI collaborators, with performance
upgrades completed. This dedicated ROLFS server will actively manage data set exNodes
and
provide a central directory for exNode storage and retrieval.



Deployment of four additional IBP servers at principle Terascale Supernova Initiative (TSI)
research sites around the country (SDSC, Stony Brook, LBL, NCSU), thereby increasing the IBP
sto
rage capacity available to the TSI and other SciDAC researchers by an additional 7.6 TB.



Further development of Logistical File System (LFS) with full capabilities to manipulate files and
directories, including create, open, read, and write functionalities
.




Make LFS compatible with the Condor project’s Pluggable File System (PFS). This will allow
LFS to be deployed to applications quickly and ease the burden of porting LFS to operating
systems whose file system hooks are different from those provided by L
inux.


9



Multiple resource capabilities for IBP will allow an IBP server to control depot storage
implemented on multiple device types, such as disk and RAM, within the same system.



Support for point to multipoint SABUL DataMover using UDP/IP multicast.



Impro
ved XOR encryption library for high throughput, faster performance.



Stronger encryption with 128 bit AES key.



Performance improvements including elimination of synchronization points and

better pipelining
of tasks.



Improved fault
-
tolerance and performance
of LoRS upload and augment commands.



Develop coding for exNode “intentions,” allowing the user to specify preferred fault
-
tolerance
parameters for exNodes including expected lifetime, fragmentation, and redundancy.



New L
-
Bone version to be released early s
ummer will provide additional metadata for improved
proximity resolution.



Further development of an experimental IBP device driver for Linux. The /dev/ibp device allows
applications with no knowledge of IBP to take advantage of Logistical Networking by sim
ply
reading or writing to the device file.

3.2

Next six months:



IBP Depot Software version 1.3 to be installed on the Italian arm of the European 6NET, an IPv6
testbed. IBP storage depots will be installed on three 140GB POP hubs in Rome, Milan, and
Bologna.

Local depots will also be installed at the twelve participating Italian universities and
research institutions.



Develop ROLFS
-
layer exNode management tool, a wide area service that will determine exNode
intent and then use that intent information to provi
de management services for the exNode.



Network Functional Unit (NFU) integrated with IBP server to provide ability to manipulate data
and perform computations remotely.



Incorporate internal DataMovers into IBP software, providing support for internal Dat
aMovers as
well as DataMover plug
-
ins.



Full integration of NetSolve and Logistical Networking technologies, using the LoRS API.



Porting the LoRS Tools software components to JAVA.



Extend LFS to include single
-
writer consistency, automatic replication gener
ation, and automatic
replica scheduling. These new file capabilities will be used to implement a checkpointing facility
for GridSAT.

3.3

Next twelve months:



Native Windows versions of the LoRS Tools software components.



Implement persistence connections bet
ween IBP client and IBP server to improve performance,
also pipeline requests to the server.




Add security features to IBP server, such as required authentication for allocation

and secure
connection for transmission of commands between client and server.



Conduct research into overlay routing to determine the optimal use of intermediate staging for
transferred data.



Research long
-
term storage scenarios for critical data, under high usage loads.


10

4.

Research interactions

4.1

Collaboration with the Terascale Supe
rnova Initiative (TSI)

Our primary research drive has been interactions with the TSI group. Recent Logistical Networking
research advances resulting from these interactions are detailed above, in Section 2.3.

We are working closely with TSI group member
s to build a new, private Logistical Networking
infrastructure modeled on the public L
-
Bone deployment, but to be used specifically for the advancement
of TSI and other SciDAC research endeavors. The first in a series of five high
-
performance, 1.9 TB
dedi
cated IBP servers has already been deployed at ORNL, with plans to deploy four more high powered
machines at principle TSI sites around the country (SDSC, SUNY
-
Stony Brook, LBL, NCSU) by mid
-
summer. These deployments will provide an additional 9.5 TB of I
BP storage space for SciDAC
researchers. Preliminary testing shows transfer speeds of up to 430 Mbps between depots at ORNL and
the UT Knoxville campus. While test transfers between depots at ORNL and NCSU reached speeds of
220 Mbps (see Figure 1), this
number is expected to improve with the upcoming depot installations.

Successful completion of the upcoming deployments will represent the culmination of several months
work negotiating with collocation services and support personnel at the five main TSI

sites. Securing
collocation agreements with the five proposed sites required efficient communication to establish clear
and accurate expectations on both sides, and our consistent attention to the needs, resources, and security
standards of the various s
ites. Although construction of this private infrastructure has been steered by
interaction with TSI, the hardware and technologies will be available for use by all members of the
SciDAC community.

4.2

Integration with Hierarchical Resource Management (HRM)

We are exploring the integration of Logistical Networking with Hierarchical Resource Management
(HRM) software, to allow HRM to take advantage of Logistical Networking overlay routing and point
-
to
-
multipoint transfer capabilities. The question of whether

the functionality of systems like HRM can be
usefully augmented by the addition of Logistical Networking resources and services "in the network" is a
central research topic for us, and the answer is of potentially great value to TSI and other DOE science
projects.

4.3

Participation in SciDAC
-
Sponsored Meetings and Workshops



Presentation and poster session at 2003 SciDAC PI Meeting, March 10
-
11, 2003, Napa, CA. The
goal of this second PI meeting was to give researchers the opportunity to access progress, sha
re
first year results, and develop collaborative goals for the future.



Participated in DOE Workshop on Ultra High
-
Speed Transport Protocols and Network for Large
-
Science Applications, April 10
-
11, 2003, Argonne National Laboratory, Argonne, IL.
The
obje
ctive of this “working” workshop was to address all aspects of network provisioning,
transport protocols, and application
-
level capability needed to craft ultra
-
speed networks
necessary to support emerging DOE distributed large
-
scale science applications.



Presentation and poster session at TSI Collaboration Meeting, February 5
-
6, 2003, Miami, FL.
This meeting allowed participants to interact, plan, and coordinate research efforts with TSI group
members and collaborators from across the country.

5.

Remarks

N
one