Attendee Usage Abstracts

We write and support tools for micro-architecture development and chip design. We currently have about one thousand processors running NT split between the desktop and a batch pool.

System installation and administration issues, including * OS and applications deployment * adapting standard applications to "roaming user" environments * user and resource management * NT/Unix coexistence * etc.

My work often requires porting tools from Unix to Windows platforms. In some cases, the tools use graphical interfaces and non-trivial process architectures, while in other cases, the tools only require standard language support on the Windows platform. A primary goal in such porting is to preserve commonality, ideally building from a common source tree mounted on all Unix and Windows platforms. To this end, designs are adjusted to take advantage of similarities in the Windows vs.
Unix architectures and isolate the differences.

We have developed Quality Objects (QuO), a framework for providing quality of service (QoS) to applications based on distributed objects. We did our initial implementation using CORBA on Solaris and Linux (and C++ and Java), and will be porting it to NT very soon as well as considering ways to support DCOM.
This is research which is part of the DARPA ITO Quorum program.
(Mark Berman of BBN will be sending in a more detailed abstract, probably later today.
Please feel free to replace his more detailed one with mine. Also, our web pages are pretty skimpy now, but I hope to get them updated by August.)

I am currently heading up study cases involving 1) porting UNIX applications to Windows NT and 2) security and reliability of Windows NT in commercial and industrial applications.

Research involves clustered computing using the Windows NT Operating system. Presently developing a Virtual Interface Architecutre (VIA) for Windows NT on x86 based machines.

My primary involvement with Windows NT is in a system administration role for Fluke systems.
We have a mixed UNIX/NT environment. NT is used primarily on the clients and file, print, email, web, and other enterprise applications are stored on the UNIX systems. We are currently migrating many of these services to NT. Our current concerns are UNIX/NT interoperability and NT scalability and security.
On a personal level, I am also interested in application development, primarily in C++ and Java. I would also like to gain greater knowledge of some the internal structure of the Windows NT operating system.

My main interest is integration of NT-machines and an UNIX environment, mainly SUN- and SGI- machines and fileserver from Network Appliance.
This integration is two-sided: both platforms need to be adapted to each other.

My research goal is to advance the scalability and reliability of distributed systems. My research interests include failure detection/recovery, process replication/migration/load balancing, checkpointing/object persistence, group communication and transaction processing.
Currently I am working on the following projects, (1)DCOM-SWIFT: Software Implemented Fault-Tolerance for DCOM Applications, (2)NT-SWIFT: Software Implemented Fault-tolerance on Windows NT (3)COM on a Multicast Transport (4)DOORS: Reliable object service in CORBA (5)FilterFresh: Reliable and scalable distributed computing in JAVA

We are involved in a large R&D project invovling a distributed Java system on an NT platform.
I am also involved in designing lab exercises on an NT platform to help teach networking concepts to undergrad students.

The research my organization will be involved with while using Windows NT will be to implement an environment of integrated Unix and NT Workstations. Our environment is a CAD/CAM workgroup of approximately 50 users. Our current environment is mostly HP-UX workstations running SDRCs' I-DEAS Master Series 6 solid modeling software. We have a mixed network of ATM, FDDI and 10baseT. NFS is our primary means of distributing executable files to our Unix clients as well as the shared data in our I-DEAS data installation. On our NT platform running I-DEAS we use Samba to map our users data files from a Unix system and the data management directories are mapped drives from a central NT server.
We also run NCDs' Wincenter Pro to provide Windows NT applications to our Unix base. Our current configuration is extremely network intensive and requires a good network configuration. Eventually we plan to have our entire computer base on Windows NT only.

I work in a large, mixed UNIX/NT environment.
We are interested in monitoring and reporting on the performance of all of these machines in a consistent manner. Part of this work involves writing the NT portion of a legacy UNIX metrics gathering tool.

Security, video over ip, global application distribution, cluster servers

Administrator for 20 Banner "Forms" server that maintain front end access to SCT Administrative system. These system are moving administrative functions from mainframe environment to client/server. Write customized hardware for load balancing of netbios client access to said servers.
Maintain synchronization across servers on both NTFS and HFS partitions within NT. Defining a campus wide NT master accounts domain with which to help create a single-account signon be it AD or NDS. Wrote custom networking package for incoming students and staff for one-step internet connectivity.

I'm interrested in the integration of Windows NT infrastructure with existing distributed environments such as UNIX and DCE.

We work on UNIX / NT bridges, software which can make UNIX available in NT and vice-versa.

I am working in the Astrophysics, Nuclear Physics and Particles Physics Department of CEA, a governamental agency for research and development in the nuclear energy domain.
In our Department we have about 600 PCs and several NT servers. To manage our PCs we are using a modified version of a product developed at CERN (Geneva, Switzerland), which is called NICE (Network Integrated Computing Environment). This product allows the installation of a slightly modified version of the chosen OS (Windows 95 or Windows NT), along with the modifications of the registry and system files to allow the usage of a network shared version of the software applications.
Once installed, NICE allows to automatically update the client OS and software applications against a master copy.
We needed to modify this CERN product to be able to exploit it in our environment (French version of the OS, different software applications).
In general, our users are physicists and engineers who need to interoperate with the existing Unix systems.
So, my main interests are the management of a secure NT deployment, the management and monitoring of client PCs and the interoperability between Unix and NT.

I am using Window NT for my office productivity tools, for my research, and for my teaching environment. I currently run NT on my desktop (and Windows 95 on my laptop).
My research involves the numerical solution of partial differential equations. I use tools such as Mathematica to perform substantial symbolic computations.
This fall I will be teach Java and C++ programming to approximately 450 Purdue freshmen. We will be using Microsoft's Visual Studio including Visual C++ and Visual J++. The educational laboratory will be equiped with NT-based machines. Moreover, I have been involved in curriculum development at Purdue for almost a decade. Currently, we are studying the increased use of NT-based systems for educational laboratories.

Over the past two years we have migrated many of our internal systems to NT, including basic file and print services and email (converted to Exchange in the past year). We have brought up new systems on NT such as intranet/web, database, and application servers. We also have large VMS and Solaris environments and these have expanded as well. We have about 1000 internal users running mostly Windows 95, with some NT and Solaris for development and system administration.
My interests include:
- Improving cross-platform interoperability and integration with our Unix environment. Current examples (completed, in-progress, and planned) include a Sendmail STMP backbone and Exchange, DNS (Bind on Unix) with NT WINS and DHCP; file systems; NIS and NT Domain account databases; Unix and NT Web servers; Legato for common backup; etc. We are closely looking at NT 5.0 to assess its impact; for example, since Microsoft is migrating WINS functionality into its DNS servers, will we be able to support this with our Unix DNS?
- Using our Intranet to unify the components of our internal environment, such as databases, file systems, Exchange mailboxes/calendars and Public Folders, etc.
- Building redundancy across our several U.S. locations; since we are expanding our presence to both coasts and internationally, this narrows our already slim maintenance window. Our basic NT infrastructure of NT Single Master/Resource Domains and WINS is very reliable and we are looking at a combination of clustering, failover, and replication to provide other applications with greater availability/redundancy/disaster recovery across a WAN in our NT environment.
- Building tools for automating administration tasks and remote administration, since NT is generally lacking in these areas. I have recently been looking at using Perl or shell scripts for this purpose.

Tandem's ServerNet architecture is a low-cost, high-performance, reliable cluster interconnect technology, supporting the Virtual Interface Architecture. Tandem Labs is using the Servernet I VIA emulator to develop several large-scale clustered HPC and commercial applications on top of Microsoft Windows NT. Servernet II will provide the same full featured VIA implementation, implemented in hardware, with much higher performance.

At QUALCOMM, we have a mixed environment of 8,000 NT/95 clients, 2000 Unix Workstations, and 1000 Macintosh clients. We are working towards a master directory to manage our users and group accounts, as well as providing a single sign-on interface to our users while integrating this mixed environment. Currently we are utilizing Novell's NDS, Microsofts Domain Structure, AFS Authentication, and NIS for this task. We are also using SAMBA and NAC storage solutions to bridge some of the multiplatform issues, and utilizing native storage solutions when ever possible. We are utilizing SMS for our automated software distribution. This is extremely useful in our environment, since end users are not granted 'Administrator' privledges, and therefor SMS must be used to properly install applications for the user.
In recent months, I have been devoting my energy towards utilizing PERL5 for NT to automate several manual functions that are un-reasonable to do in a large environment. By exectuting external programs such as 'net user' to gather information about our user accounts, and cross referencing that information with other user databases we can track inactive accounts, and stale accounts. Since our deployment of SMS is only as good as the machines that actually have the client software, I have also written a script that 'audits' the network, and compares the list of machines with what SMS sees, and if need be, pro-actively tries to fix the machine automatically, and if not possible, it sends e-mail to our support team so they can look into the problem.
My next project is to work on setting up perl scripts that parse the event logs looking for failed access attempts, and failed logins. I am also interested in seeing what can be done to create a 'trip-wire' utility for NT machines hosted in a 'untrusted' environment.

My research interests are in the area of resource scheduling/management for cluster environment: resource scheduling algorithms/paradigms, scheduling structures for distributed system, schemes to support dynamic resource allocation and workload balancing and etc..
NT and Unix are the two platforms that I use to study these resource scheduling/management issues for cluster systems of heterogeneous hardware and operating systems.
Resource and system management for high availability clusters (e.g. Wolfpack) is one of my new areas of investigation.

I am a senior computer science major at the University of Wisconsin Madison. In my studies and my job as a system administrator I use Windows NT on a daily basis, and I am very interested in attending the LISA NT conference to not only expand my knowledge of the operating system itself, but to become acquainted with and speak to others who administrate Windows NT. I feel that this conference is an excellent chance to exchange ideas and to see how others have dealt with the problems that arise when administering a large network.
After my graduation I plan to focus on network security, and this conference is a prime oportunity to see how existing systems are kept secure. I am very excited to attend the conference and become further educated on this widely used operating system.
Sincerely, Ryan M. Fruit

I am not currently using Windows NT. I have in the past, done some Internet application devlopment on NT. The major motivators for my desire to attend are 1) to talk with my peers/fellow Usenix members to gain perspecitve on whether I want to to embrace "the dark side" 2) to get a pulse of the current state of research using Windows NT (i.e., what's being done, what's possible, what's research, what's product....)

NT is becoming popular as an internet server, we would like to be able to provide our customers secure internet server with NT like we do with unix.
Hopefully this conference will help me better understand security issues with NT and how to prevent any security configuration holes.
Joe

Currently, I am in charge of the administration of Windows NT 4.0. The application running on top of Windows NT is MCIS v2.0. The components running are SMTP servers, POP servers, LDAP servers, Web Severs (IIS 4.0), Membership Directories and Mailstores. Our current mail server configurations is supporting about 120,000 users with an average of 5 email messages per second.

I am currently researching ways of preparing my existing enterprise to leverage the advanced features of NT 5.0. I have an opportunity to create a new domain and want to make sure it fits in line with our NT 5.0 rollout strategy especially when it comes to the integration between DNS and the Active Directory.
My enterprise is filled with NT and unix machines (IBM, HP, Sun, SGI), and I am always looking for a better solution to share files between them. Currently we use commercial implementations of samba and are evaluating Auspex's CIFS solution.

Looking at the automatic generation of native-NT transaction/database applications from 4GL/non-procedural system specifications. The system being prototyped will take an application system specification (rules, rocedural pieces of code, interfaces, persistent data) and build the corresponding sets of components, database schemas and interfaces. This will all be done using current Microsoft technologies - COM, MTS, Wolfpack - build systems that are scalable and resilient. The proposed system will let users build compex application systems without exposing them to the complexities inherent in creating and managing such systems under Windows NT.

I am interested in the viability of Windows/NT, and particularly Microsoft Network Protocols, on the Internet. Concerns include scalablilty, security vulnerabilities, standards compliance, and the general ability of these systems to get along in an open, heterogeneous environment.

My current research is focused on providing highly-customizable quality of service (QoS) for distributed systems and applications. The goal is to provide an integrated framework for QoS that includes various aspects of quality such as reliability, real time, security, and performance. Our current primary research platform is MK operating system from Open Group Research Institute, but we are exploring the option of moving part of our research to the Windows NT platform. Thus, the major reason for attending this conference is to learn about the feasibility of doing such work (in particular, real time) on the Windows NT platform.

Design solutions for our NT environment.

I'm a Sun Unix System Administrator and also MCSE.
I'm working on integrating Unix and NT resources to provide single log-in, single storage for our users.
My daily duty is to provide user support ( both to users and my colleague), improve installation/configuration process, develope in-house tools to enchance our work.
I have been working on SunOS, Sun Solaris, HP-Unix, Linux, and MS WinNT/95 System Adminstration for more than two years.

The OS Research group at Microsoft Research is currently using Windows NT as the base for Millennium, our research project in self-configuring, self-tuning distributed systems. As part Millennium, we are exploring how best to implement DCOM over SAN, how to support live migration in object-oriented systems, and how to create system which automatically tune their behavior at very fundamental layers to adapt to changes in system architecture or user usage patterns. On this front, I am particularly interested in the optimization of applications using just the program binary files.

We are developing a high performance platform for developing and deploying network applications.
The first version of this is based on Windows NT.

Scandinavia Online System is a company who is mixing Windows NT and Unix environment today.
We have several NT servers and Workstations, Unix(mostly SUN Solaris, Linux and IRIX) which we want to integrate as close as possible. We have several Network Appliance filers which uses both NFS and CIFS. We are interested in knowing about other companies experience with the same solutions.
We are also especially interested in NT security and large-scale distribution of Windows NT. We also want to know more about scalability of Windows NT and IIS.

One of the research projects underway at Panasonic Technologies, Inc. is an investigation into scalable video servers and their guaranteed real-time video delivery in a distributed multi-media environment. There are two main research aspects that are being considered: the file system, and the network.
The file system needs to be resilient to error conditions and fully recoverable while staying on-line. Because the server is scalable, the file system can also be distributed. A final aspect to the file system is transparent support for a tertiary archive.
The network studies concentrate on two aspects: within a distributed server and for media delivery to/from the client. The server has to optimize quality of service while maintaining both its communication requirements to the various server parts and the data needs of the client. There are also the requirements to support both intranet and internet service compatibilities.
Windows NT has been chosen as the test bed for our investigations, both for the server and the clients.

I have implemented a Mainframe file system on NT, which enables you to browse mainframe data stored on Hitachi 7700 disk shared with mainframes, using ordinary Windows applications.
You can share/transfer bulk of data used for data warehousing, or you can edit COBOL source code by MS-Word.
I am ready to provide a hands-on demo using my note PC.
See you on Poster and Demo session.

The Johns Hopkins University Applied Physics Laboratory (JHU/APL) is a non-profit research division of JHU which supports the Department of Defense and other government agencies through applied research, technology development, and innovative problem solving. In the Advanced Systems Development Group, I work with a diverse group of electrical engineers, computer scientists, mathematicians, and even a few physicists in applying leading-edge information concepts, architectures, and technologies to complex systems-level problems for our sponsors.
Over the last decade, most of our research systems have been based on Unix variants and/or Real Time Operating Systems (RTOSs). Our hardware has ranged from single-board computers, to "cheap PCs," to high-end parallel machines such as transputer arrays, the Goodyear ASPRO, Intel Touchstones, and an occasional high-end SGI. More and more, the "cheap PC" has been chosen as the hardware of choice, with x86-based O/Ss ranging from Mach to Linux, LynxOS, QNX, VxWorks, and some ---x and --s O/Ss I'd rather forget.
But in the last two years, we have had to sit back and give NT a hard look, for ourselves as well as for our sponsors. It has many of the features we desire for our research: a real O/S, at reasonable cost, with very good tools, did I mention the plethora of 3rd party applications and hardware?
But it also has a few warts: less than perfect determinism for real-time, security holes, sudden lockups, and system configuration problems. From a community that includes folks that worry about 20+ year life cycles, we also worry about openness, long-term commitment and support, open vs. defacto standards, and interoperability. Some of these issues will simply take time to fix as NT matures, while others may be addressed through evolution or random mutations made through NT-based research.
While my systems-level work takes me across broad swaths of research, my current NT-related research interests include: · Making NT more responsive for distributed real-time applications · Making NT more intrusion tolerant and secure in distributed environments · Making NT more fault-tolerant through clusters, group communications protocols, and dynamic resource sharing · Reducing latency while increasing determinism and throughput of compute and communicate cycles through NT using various high-speed networks such as ATM, Fibre Channel, Myrinet, and Gigabit Ethernet

Westinghouse has been developing a product that is to be used to ensure that NT systems and UNIX systems maintain critical software and hardware configurations. This product provides operational support as well as auditable reports. The auditable reports satisfy NRC requirements for Nuclear Power plant design processes. I am the technical lead for this product.

We are using Windows NT in a growing number of computer science research projects, including computer graphics and geometric modeling using OpenGL, image analysis and numerical analysis using MPI, symbolic computing using Mathematica, and security testing and evaluation. Much of this work has been done traditionally using UNIX-based workstations and servers. Our current efforts involve porting code to the NT environment, while maintaining compatibility with and access to UNIX resources.
Some of the system administration issues we are facing with NT include using SMS to manage software installations, running systems in secure or "locked-down" mode, allowing non-administrators to install and maintain software, sharing files with NFS-based UNIX file servers, and managing accounts across NT and UNIX platforms.

I am responsible for the deployment of all NT Workstations with in GE Medical Systems. I configure and test all the Standard configurations used in the business. I am also responsible for deciding on new platform and NT applications for our Engineering group.

I am intereseted in the development of applications that manage Windows NT Clusters.

Studying the feasability of using NT as a client test bed for stressing servers in a message switch system. Exploring the use of winsock2 as the underlining network stack.

At Qualcomm, we have a mixed environment of 2,000 UNIX and 8,000 NT clients. We are working towards a master directory to manage our user and group accounts as well as provide a single sign-on interface to our users while integrating our mixed resources. Currently we’re utilizing Novell’s DS and Microsoft’s SAM to maintain our user accounts. We also use Samba and Network Appliance file servers (SMB/CIFS-based servers) to provide an enterprise solution for our NT clients to share files securely between our UNIX and NT platforms while maximizing our multi-platform operations to deliver enterprise-wide file sharing. We are also evaluating and implementing clustering and fail-over systems for redundancy and maximum uptime as well as to help distribute our workload across our many resources effectively and efficiently.
My position at Qualcomm is to integrate both platforms (UNIX and NT) and to come up with a standard setup. I'm working on such projects as building an NT and UNIX cluster farm which involves workload management to support our applications that are critical to our enterprise.
My focus to to provide maximum utilization of hardware and software resources. My other projects include deploying CIFS-based file servers that speak NFS while participating in the NT domain and providing file services to NT via CIFS and evaluating DFS.

I am a Manager in the University Research Programs group within Microsoft Research. I have responsibility for maintaining a formal relationship with roughly half of the top 50 research universities in the US, and a few in Canada.
I also have varying degrees of responsibility within MSR for a few government institutions; namely the DoD, DOE and DARPA and NSA.
In addition to general relationship management, I have research interests that I pursue in the following areas:
Network and Application Security Mobile and Wireless Research Embedded Systems Research (Win CE based)

We are investigating the use of NT-based compute servers in a heterogeneous environment.
I'm focusing on the development of tools for managing special purpose remote execution servers.

Our research group (CSAG) has realized a Windows NT version of High Performance Virtual Machines (HPVM), a software providing communication services on cluster PCs connected with a high-speed Myrinet network.
The HPVM software solves the key challenges in making a cluster of Windows NT machines useful as a parallel computer. First, HPVM provides users with a collection of standard APIs like MPI, Shmem, Global Arrays with supercomputer class performance (13 us minimum latency, 84 MB/s peak bandwidth for MPI), efficiently delivering the underlying Myrinet's performance to application programs. Second, HPVM provides cluster management and scheduling (thru Platform Computing's LSF). Finally, HPVM solves Windows NT's remote access problem, providing convenient remote access and job control (thru a graphical Java-applet front-end).
HPVM is currently running on a 192-processor cluster of Pentium II PCs (dual-processors), built by the CSAG in collaboration with NCSA at the University of Illinois at Urbana-Champaign.

Our group is developing NT-based multimedia applications.
I/O performance is crucial. NT appears to provide adequate hooks to deal with I/O bandwidth, but there are significant holes in the Microsoft supplied documentation.
We need to take advantage of opportunities to share knowledge of and experience with NT from a developers point of view.

We are developing filesystem and NDIS driver on Windows NT to support the continuous multimedia transmission and QoS issues on TCP/IP and RTP.
We also use Windows NT to provide a framework of digital library framework. It supports video on demand, CD-title service, audio CD listening and other types of media integrated by Microsoft SQL server with a WEB based front-end.

The Laboratory of Hygiene is the state public health laboratory for Wisconsin, attached to the University of Wisconsin - Madison.
We provide clinical, environmental, and industrial laboratory services to state and county agencies, hospitals, and private citizens.
We have around 350 users of 250 NT workstations, 10 NT servers, and a variety of unix and proprietary minicomputers. These are supported by an IS staff of roughly a dozen. Our NT network is a single domain, geographically dispersed over 4 sites up to 5 miles apart. The remote sites connect to the UW campus network via T1 or ISDN lines. 3 of the 4 buildings have local NT servers providing file and printer sharing, domain control, and DHCP. We use SMS to push packages to workstations, and either NT backup or Cheyenne Arcserve for server backup. Workstations are standardized as much as possible to use the same drive letters and application suites and are not backed up.
We use a mixture of server and workstation based applications, including Eudora e-mail, Netscape web browser, MS-Office, SPSS, Aldus Pagemaker, WRQ Reflections terminal emulation, Oracle, etc. We use MS-Exchange to provide POP3 e-mail.
We use roaming user profiles and mandatory policies, with a kixstart login script which maps shared software off the nearest file server. Workstation installs are network based, using a global "workstation administrator" group with rights on all workstations. We set up an NT administrators user group at the UW-Madison to share NT problems and solutions.
Our three biggest problems with NT are with slow performance booting and logging in, high administrative overhead, and the difficulty of tightening up NT security.
Two of this years projects affecting our NT network are setting up a firewall, and deploying Outlook as an e-mail and calendaring client.

My Windows NT involvement includes research and development in security, NT/UNIX system administration, collaborative tools, secure remote dial-up, and mobile devices. My focus is on applying commercial applications for military use.

Interactivity in high performance computing is loosely defined as providing a means by which end users can interact with their high performance applications while the applications are executing. It includes research into steering, visualization, collaboration, and dynamic adaptation. Our previous work established an interactivity framework for UNIX based applications[Pla98]. Our current work is in adapting the framework to NT applications. The effort includes a pilot project wherein CS researchers work with Physics researchers in integrating the interactivity framework into simulation code. Further efforts are underway to develop an experiment manager that can handle experiment setup, analysis, and adaptation in an NT environment.
Interest in NT is high, partly because the applications most benefiting from interactivity (e.g., large-scale optimization, and cancer treatment) are ultimately transferred to or adapted for end-users that are running NT on their desks. In addition, the computing power of Pentium PCs running Windows NT offers a readily accessible and inexpensive platform for large-scale applications. Industry is moving in this direction. In particular, we see it in our work with U.S. airlines.
The IHPCL project, through a grant received from the Intel Corp., has created several high performance cluster computers, each with substantial computational speeds. These next generation clusters provide a low-cost solution to high performance computing of complex scientific applications. A 64-node NT cluster of 200 MHz Pentium Pro machines connected via 100Mbps Fast Ethernet is the testbed upon in which the interactive framework is being developed.
Our effort has already resulted in ports of tools that facilitate communication[DE], binary I/O[PBIO], and monitoring and steering[MOSS]. Infrastructures often require additional support, particularly for interactive applications involving multiple scientists. These scenarios frequently consist of multiple computational components: data sources that provide the data to be disseminated, clients that are the net consumers of the information, and information brokers (i.e., agents or mediators) that acquire the information from other sources, add value, then distribute it to downstream brokers or clients. An example of such a broker occurs for query processing across heterogeneous data sources. The user issues a query in a general database query language which is then manipulated by intermediate sources which might break it down into pieces understandable by the individual sources, and optimize the pieces. A broker also asembles the results. Conducting an experiment involving nume!
rous data sources, clients, and information brokers is facilitated by the presence of a experiment manager.
Experiment management is important when dealing with mismatches between brokers and clients or other brokers and with changes in the needs of individual end users. Experiment management may be even more important when dealing with large-scale experiments comprising many computational instruments, components, and end users. For such experiments, experiment management provides the single point of control from which experiments may be initiated and controlled, users can view current resource characteristics, allocate resources to tasks, start component pieces of the experiment, and view experiment progress.
Toward these ends, an experiment manager must perform the following tasks:
-- experiment setup, control, and teardown,
-- data management,
-- configuration management, and
-- configuration analysis and adaptation.
The first item concerns the explicit definition and control of experiments by end users, including the definition of suitable input and output files for computational instruments, the specification of experiment repetition and completion, etc. The data manager maintains information about resources and computational components, and it provides efficient access and update methods to this information. The configuration manager determines appropriate locations for computational components, using knowledge of the information maintained by the data manager. It also manages remote execution control, that is, it provides low-level services for tasks like starting components on remote machines. Finally, configuration analysis and adaptation implements the necessary feedback loop. By gathering information from the experiment, it couples this information with information retrieved from the data manager to make informed decisions that affect future placement and adaptation in the experim!
ent environment.
[Pla98] Beth Plale, Volker Elling, Greg Eisenhauer, Karsten Schwan, Davis King, and Vernard Martin, "Realizing Distributed Computational Laboratories", to appear in The International Journal of Parallel and Distributed Systems and Networks.
[DE] Greg Eisenhauer, Beth Plale, Karsten Schwan, "DataExchange: High Performance Communications in Distributed Laboratories", to appear in Journal of Parallel Computing.
[PBIO] Greg Eisenhauer, "Portable Self-Describing Binary Data Streams", Technical Report GIT-CC-94-45, College of Computing, Georgia Institute of Technology, Atlanta, GA 30332-0280.
[MOSS] Greg Eisenhauer and Karsten Schwan, "An Object-Based Infrastructure for Program Monitoring and Steering", to appear in Proceedings of the 2nd SIGMETRICS Symposium on Parallel and Distributed Tools (SPDT'98), August 1998.

NT for transaction processing systems. Design, implementation.
NT performance analysis. Measurement, monitoring, tuning

Typical medium to large scale NT desktop, UNIX backended environment with all the issues you would expect to be present.

Not real research done here.
I'm involved in managing a Windows NT-based network integrated with unix hosts and workstations. Integration involves DNS-DHCP issues, file and print servers/clients solutions, distributed database and database/web integration issues, scripting for cross-platform automated managing, network performance analysis, backup strategies.

Evalutate transition of a UNIX based development environment to a Windows NT based development environment

Specific, purchase and manage an enterprise system based on Windows NT and Novell Netware.
Current issues include user database management integration with Unix systems. Especially interested in cross platforma dn Web based management tools.
Enterprise system is used for Exchange Mail service, network FAX services and central file server used for software distribution. Active in designing and deploying courses on Internet Technology, Web ASP technology and firewalls.

Test, integrate, and implement Windows NT in a mixed UNIX and Windows environment.
1) Need to know how to implement Windows NT 5.0.
2) Need to know how Windows NT 5.0 works with UNIX.

We're primarily struggling with security, user authentication and administration issues. We want to integrate NT with our existing account infrastructure (athena/kerberos). We've also implemented AFS and while we can load user's home directories, we don't have it truly integrated.
For remote administration I've looked at SMS, but find it sadly lacking consider MS wrote it and the operating system. I'm strongly considering writing a series of perl scripts to perform software distribution and maintenance.

Texaco is looking at using Windows NT machines clustered together to build a supercomputing capability out of machines that lie on users desktops. In addition NT is critical to our computing infrastructure

Our site currently has a mixture of platforms -- Windows, Novell, Macintosh, Digital VMS and Sun Sparcs. We are moving to supporting only Windows NT and Unix. At present, we have about 200 PCs, and over 300 Sparcstations. The systems are managed/adminstered by teams according to platform.
I administer NT systems and provide user support as well. Our main areas of interest are in NT security, and using SMS to push software and assist in troubleshooting. I am interested in meeting with people who have multiple servers, run SMS and SQL, and IIS. I would also like to discuss problems with the Microsoft Licensing software.

Constructing Dependable Distributed Software on a Network of Windows NT and Unix Workstations
With increasing popularity of Windows NT operating system, most modern distributed computing systems consist of a network of computing nodes running Windows NT or Unix operating systems. This heterogeneous computing environment is gaining popularity in commercial, research, and academic environment. Even the distributed computing environment that are used for running critical applications are becoming more and more heterogeneous with individual nodes in the network running either Windows NT or Unix operating systems. Examples include air traffic control systems, space applications, and several distributed process control systems.
My main goal is to investigate and design fundamental techniques for constructing dependable distributed software on a heterogeneous network of nodes running Windows NT or Unix operating system.
Constructing dependable distributed software is hard for several reasons including the uncertain nature of communication or processor failures and the asynchronous nature of most distributed computing systems. Heterogeneity of nodes in a distributed computing system adds a new level of complexity in constructing dependable distributed software.
Several useful techniques have been developed that simplify the design and implementation of dependable distributed software. Examples include group communication services, process checkpointing and rollback recovery, N-version programming, and so on. Most of the work done in designing these techniques has focussed on a homogeneous distributed computing environment with nodes running Unix operating system. While these techniques provide some powerful abstractions, they cannot be directly applied in a heterogeneous distributed computing environment. For example, a checkpoint of a failed process on a Unix workstation cannot be used to recover that process on a system running Windows NT. Similarly, a group communication service designed for a network of Unix workstations cannot be used to detect member failures if some of the group members are running on a Windows NT platform. Software tools such as CORBA help in designing and implementing distributed software for a heterogeneous computing platform. However, these tools do not address two very important issues: high performance and dependability.
I am attending this symposium for two purposes.
First, I would like to find out the tools and techniques that have been developed for designing dependable distributed software on a network of nodes running Windows NT operating system. Second, I would like to discuss the ways to extend the existing techniques for constructing dependable distributed software to a heterogeneous distributed computing environment that includes nodes running Windows NT or Unix operating systems.

I currently work for a NY State utility. We have just migrated to NT 4.0 on the desktop and are proceeding to migrate from Netware3.12 to NT file and print. Our enterprise consists of aprox 5500 clients.

I work in a large software development organization that currently develops it's software in a UNIX environment. We are making plans to take the development environment into native Windows NT.
I need to learn from the experiences of other people who have already taken such a journey.
Our development infrastructure consists of four to five hunderd workstations, and several file and compute servers.
My focus is largely on infrastructure, system administration, and capacity planning. I hope to find out about tools for sys admin, and job scheduling. I also would like to know how others have solved their needs for file sharing with a heterogeneous UNIX environment. (Large scale)

My primary interests are with transitioning a multiple UNIX DNS domain WAN environment from UNIX-based services to NT-based services. Where services are comprised of:
1. DNS; 2. Print; 3. File; 4. E-mail.
I am wanting to research ways in which I can help my company effectively migrate an existing NT client base of 1000+ from a UNIX authentication-based, SMB filesharing, non-NT- domain environment while still being able to support existing UNIX file and print services for UNIX clients (UNIX file services are also required by NT clients, though would like alternate solution to TAS [would like to avoid SAMBA as well]).

I manage Microsoft Research's external research programs. As such, my primary responsibility is acting as a liaison between university/government research groups and Microsoft/Microsoft Research. This includes research funding, our guest lecture program, technical briefings, and working with government/public funding and standards organizations.

My research interests are in operating systems, high performance network, and performance evaluation.
The use of NT in our laboratory and research is still limited, which is why I wish to attend the conference and NT internals workshop. Our intent is to include NT as a platform in networking and performance evaluation research that we currently do.
Specifically, we conduct network level performance evaluation experiments using many hosts and would like to include NT in the set of hosts used in these experiments. Somewhat longer term, we gather information from the OS internals during these experiments and learning to do the same under NT would be of interest. In addition I work with real-time systems, and adding NT to the set of support platforms considered is desirable. Thus my interest in NT internals, since knowledge of it will improve interpreting NT performance data, and eventually being able to instrument parts of the OS would be even more interesting.

Windows NT Usage Abstract [RWTH Aachen, Lehrstuhl fuer Betriebssysteme]
The range of work of the chair comprises all questions of design, implementation, test, maintenance and analysis of complex software systems, especially for parallel and distributed computing environments.
We are using Windows NT as development and implementation platform for parallel and distributed applications. We operate a Windows NT based cluster of Dual Intel Pentium Pro based systems connected by FastEthernet as well as the Scalable Coherent Interface (SCI).
We develop Windows NT based system extensions to support parallel programming (e.g. shared virtual memory, message passing) and the efficient parallelization of industrial application codes (especially those from the scientific computing area) by using these system extensions.
Two current projects in this area are SVMlib - the shared virtual memory library, and SMI - the shared memory interface. These two projects reflect the two network architectures of our Windows NT cluster. While SVMlib runs on FastEthernet and TCP/IP, SMI provides a programming interface on a NT based SCI-Cluster. In more detail,
- SVMlib provides a page based, user-level shared virtual memory (SVM) subsystem for Windows NT. The emulation layer nt2unix (which is presented by us at the symposium) is used to compile and run the SVMlib code also on Unix based systems. SVMlib's main features are: - support for multithreaded user level code on heterogenous platforms (Windows NT, Solaris x86, Solaris SPARC, Linux)
- support for various API personalities (including SPLASH, CVM, SMI, JIAJIA, P4)
- support for various consistency models and distributed synchronization algorithms;
- run time visualization using the PABLO environment.
- SMI is a programming interface for data-sharing via explicitely installed SCI regions of shared memory, distributed synchronization, loop splitting and loop scheduling. Using SMI, various parallelization efforts are ongoing (including the GROMOS96 MD simulation, flight and passenger scheduling algorithms, and other codes from the scientifc computing area).
Additionally, we are using Windows NT for educational purposes. This includes the presentation of NT during OS lectures and using NT as the implementation platform for our students' Diploma and Doctoral Thesises.
For more information, please contact:
Univ.-Prof. Dr. habil. Thomas Bemmerl RWTH Aachen, Lehrstuhl fuer Betriebssysteme Kopernikusstr. 16, D-52056 Aachen, Germany Tel. : +49-241-80-7634, Fax.: +49-241-8888-339 e-mail: contact@lfbs.rwth-aachen.de WWW : http://www.lfbs.rwth-aachen.de

Intrested in NT security;Unix integation; NT Administration;

I have been involved with two projects that have been completely developed on NT. However, both of these projects were implemented fairly independently of the NT platform per se. Each was developed in Java, relying on development environments for code production.
Some of the interesting aspects of these projects include: 1) In both projects we heavily used the JDBC interface to interact with databases, both native to the NT environment, such as MS-Access, or running on networked Solaris machines like Oracle or MySQL.
2) Both projects rely on a fair amount of communication between hosts, which can be on heterogeneous platforms.
3) For one of the projects we developed a NSF server, which performed a la par of other NFS servers we had available.
The NFS server runs over the middleware created in the project, basically making a non file system based repository behave like a file system. The type of behavior we get is similar to the one achieved by the "Semantic File System" by Gifford et al. (SOSP'91).
Of the NT specific problems we have encountered throughout the process of developing on the platform are: 1) Having accurate resource management information. For example, it turned out that running one of our servers made software installation scripts hang, but we could not figure out which component could be causing the problem. Some of the components were out of our control (e.g., the ODBC driver.) 2) Similarly, when developing over applications like Access, in a number of occasions we ran into problems because our software would crash (due to our bugs) in the mids of the execution with the consequence that some resources were basically forever locked, like MS-Access databases.
How to manage NT resources, monitor their performance, etc. in a development environment is an issue that we are still struggling with. On the flipside, developing in Java has been an incredibly productive choice, but has also insulated us from a number of NT specific issues.

I am involved in the support of student computing at Georgia Tech. I would like to meet with other university people concerning security and other WindowsNT issues on campus. We are trying to determine how much freedom to give our students to roam the Georgia Tech network from their dorms rooms and how to assist them in using departmental servers to pick up class notes, assignments, syllabi, etc. and to submit their homework.

The LSU Libraries currently has 2 UNIX (SGI/IRIX) web servers. Its LAN conisists of a Novell Netware 4.1 server, which serves CD-ROM databases off Meridian CD-Towers to the staff and the public within the intranet. The LAN also consists of a Windows NT Server.
The Windows NT Server currently is used as an ERLServer which servers SilverPlatter databases off the web. It is also used as a NT gateway to the Netware Server and is used only to backup the Netware Server with its (NT Server's) tape drive. No users are currently logging on to the NT Server.
The LSU Libraries is currently looking to migrate completely to a NT Network, eliminating the Netware Server. We are also looking at integrating the UNIX Servers and the NT Servers using NFS, Freeware etc., Security of the Data Shared between the two is also an issue in consideration.

Primarily involved with the support of our administrative users. They are mostly running on Windows 95 connecting to NT Servers. Some NT Workstation.
Interests and concerns are:
- Developing or obtaining admin tools for the everyday mangement of approximately 15 domains. Have written Perl scripts for some tasks.
- Re-configuring to fewer (hopefully 1) domain in preparation for NT 5.0.
- Easy deployment of new workstations, including NT, with as little interaction as possible.

I am the system administrator of the Computer Science Instructional Labs for the CS Dept here at the University of Illinois. These consist of mainly 3 labs with 2 labs having Sparc, Sgi and HP systems running the vendor supplied version of unix and a third lab consisting of 75 PCs running NT4.0. The system management of these system in a cross platform environment is the main reason why I wish to attend these symposiums. Some of these lab systems are used by research assistants doing special projects for certaiin professors and trying to manage them amongst all the instructional systems can be quite a challenge.
Some of these challenges include, installing new OS on each platform, maintaining common user accounts accross all platforms. Passwd conformity accross platforms. File quota accountability accross platforms. Somba installation and upgrade.
Printing capability. Various other system management functionality.

I maintain a 500+ node network for the department of Molecular&Cellular Biology of Harvard University.
Clients are currently a 56/31/13 percentage of Macs, PCs, and Unix boxes. Nearly all of the PC's are Windows95 or NTW 4.0, but there are about a dozen NTS4.0, a few WfWg, and some OS/2 machines. The Unix machines are nearly all Solaris boxes, although there are DEC Alphas, HPs and SGI too.
The environment is a single-domain NT model, with a PDC and four BDC's (one for each of our subnets.) All PC's participate in and are authenticated by this domain. Most of the Macs are also incorporated, via DAVE 2.0 from Thursby Systems.
I am interested in reaching a single-user, single-logon environment for my department, with universal access to domain resources and centralized security and administration.

Developer of WinFrame/MetaFrame NT software.

Our current involvment in research is mainly in the integration of large scale NT based LAN services and office automation, including corporate messaging, file and print services, desktop automation and control, security control, performance monitoring tools, anti-virus tools, etc.
The focus is on the integration of the tools with the desktop and the client server applications. We use a distributed model with data base apps running on UNIX servers and clients running on Wintel PC's. The goal is to provide the adecuate level of centralized administration and control on key resources through the use of NT based tools and services for the desktop environment.

"Damn it, Jim, I'm an administrator, not a researcher!"
My interest in NT is purely practical. Presently I'm responsible for maintaining a mixed NT/Unix environment with approximately 40 Windoze clients.
Shortly I'll be adding a remote server.
I'm interested in coaxing maximum performance from the existing NT servers; gathering maximum diagnostic and performance data; re-allocating tasks to maximum effect between NT and Unix.

System installation and administration issues, including * OS and applications deployment * adapting standard applications to "roaming user" environments * user and resource management * NT/Unix coexistence * etc.

I am interested in the integration of NT within a UNIX environment. In particular, heterogenous large grain distributed computing.

Our management has decided that we will move our applications from Unix (DEC Unix / Linux) to Windows NT. We have several million lines of C using Motif and X. Our system is made up of about 10 main processes that interact via all available UNIX mechanisms (shared memory, pipes, sockets and signals).
While not enthusiastic about this move, I need to prepare myself for the ordeal.

As both a system administrator and a second year Masters student at the University of Wisconsin I have seen the dramatic growth of the Windows NT operating system in the academic environment. In just the past few years, we have transitioned from our separate Novell-Dos and Unix systems to a comprehensive Windows NT system consisting of over 170 machines working hand-in-hand with our existing Unix base. I am sure this trend will continue and thus I am interested in knowing the latest information on NT administration and what it holds for the future.
The thousands of users that we support each year constantly force us to deal with issues such as security and systems management. This, combined with the ever-present fact of limited resources puts emphasis on being able to monitor the current use of our resources in order to predict future demand. I feel that the LISA NT conference will help to expand my knowledge of NT administration in these crucial areas by imparting the knowledge and experiences that other system administrators have gained in sites similar to ours. It is also personally interesting for me to see the evolution of the NT operating system so that I can obtain a historical perspective of NT as I continue my career of system administration well into the future.

I am primarily involved in research around using workstation and server clusters running shrink-wrapped Operating systems as a scalable computing platform. Focus of our current research is around integrating user-level networking architectures intodistributed computing facilities in NT like RPC, DCOM, MTS etc.

We are using Windows NT machine for the Video Server and AV device Control Server which perform control PAN, TILT and ZOOM of some video cameras.
The server also performs image processings and/or picture coding.
And also the server is http server.
I am considering how much computing resources are assigned for each processes or threads. Because the processes are related single media such as video, audio or control, the resource are influence on the Quality of Service.

Due to their excellent cost/performance ratio, clusters of PCs can be attractive high performance computing (HPC) platforms. However, their limited communication performance over standard LANs is restrictive for parallel applications. It is the goal of the SMiLE project (Shared Memory in a LAN-like Environment) at LRR-TUM to overcome this performance gap by adopting the SCI interconnect technology (Scalable Coherent Interface). With hardware-based distributed shared memory (DSM) and high performance communication characteristics (throughput in the range of tens of Mbytes/s and latencies in the range of a few microseconds), a cluster built on top of this interconnection technology is regarded as well suited for HPC.
Our objectives within the SMiLE project are the construction of low-cost SCI-based cluster from commodity-off-the-shelf PCs with hardware DSM and the adaptation of various programmng models on this new architecture. So far, efforts include the implementation of fast socket communication, generic active messages, and the multithreaded, dataflow based runtime system MuSE which relies on messaging using our own active message implementation. Currently also under development is a port of the widely used PVM (Parallel Virtual Machine) package to a cluster of SCI connected PCs.
Besides these efforts in efficient message passing implementations on top of hardware DSM, we are also focusing on concepts for global virtual shared memory models on top of physically distributed NUMA-like memory. The result will be SCI-VM (SCI Virtual Memory), a flexible memory layer to abstract the distributed memory resources into one global memory segment in the virtual address space.
The SMiLE project is based on several operating systems, only restricted by the availability of SCI hardware and software. Currently we are using Solaris on SPARCs as well as Linux and Windows NT on x86 based PCs. Especially with the two currently ongoing projects, PVM and SCI-VM, we are targeting towards and developing for Windows NT due to its wide user base and acceptance in industry. In the future, however, these software layers will also be available on other platforms. This alone ensures the portability not only of the software, but also of the concepts and the general programming models.
For more information visit our SMiLE website: http://wwwbode.informatik.tu-muenchen.de/Par/arch/smile/.

Research into Checkpointing on Windows NT. Join author of paper which will be presented at the symposium. Also interested in distributed computing systems.

We are investigating a distributed, mixed-platform (WINNT and UNIX) environment using KIVA software for the deployment of an on-line yellow pages directory services application.

I assist in all web based infrastucture for Fluke Corporation. I had previously been a web based programmer/analyst for the Boeing Teledesic program. My primary role is defining the architecture for our current application migration to the web. This encompasses server and network configuration.

I'm interested in Windows NT performance, performance tuning, benchmarking, scalability, and hardware architecture.
especially in large system, like deschutes 8 way, TPC-C benchmark.

We have developed in-house, a set of products for the installation of software to windows NT clients and servers, as well as set of tools for the management of the software installation and the state of the clients and servers.
The basis is a Sys-V like package format (split into a server part, client part and user part) in which we re-package all our applications (internal as well as standard applications like Microsoft Office97).
The package format is the first level, the second level is the set of applications used for bringing the machines to a well defined state and the third level is the coordination (management of a server and group of clients) when applications are split between clients and servers.

I'm trying to implement a proportional-share CPU scheduler on NT for use by both real-time and non-real time applications. My general interests involve adaptive system software, so I'd be happy to talk with anyone who has similar interests, experiences, or problems in this area that they'd like to see solved.

Researching computing on clusters of NT systems. A related interest is performance of NT networking.
Currently building a prototype COTS NT cluster.

I am studying computer science at Hiram College.
This year the college is implementing private Windows NT accounts in the public computer labs as well as a massive change in the structure of the computer systems on campus.
My other interest in Windows NT lies with the research that is being done at NASA Lewis Research Center. I am a summer intern at Lewis and the area where I work is currently studying Internet Protocols over the Advance Communications Technology Satellite or ACTS. Part of this work will be done on Windows NT machines interfacing with other NT machines as well as UNIX machines.

Operate in a heterogeneos environment, integrating nt, mac and unix systems numbering in the thousands across a lan and wan covering the pacific from the west coast of the US to Japan and Korea. Develop applications on nt and unix platforms migrate to web based information publishing and discemination.

Working to improve interoperability and usability between UNIX and NT.

I work at the University of Wisconsin-Madison as a software developer and a project manager for the Condor High Throughput Computing System (http://www.cs.wisc.edu/condor). Condor is an advanced software package for harnessing the power of a networked cluster of computers, by utilizing CPU cycles which would otherwise sit around idle. Condor has been a research project at the UW Computer Science Dept for many years in the Unix world, and we are now engaged in the development of a WinNT version (deep-port).
Research interests on WinNT include: Win32 process checkpoint and restart, transparent process migration from one NT workstation to another, transparently trapping system calls of a Win32 process and sending them as a remote procedure call to a remote workstation (part of process migration), discovery of system and process utilization statistics, translation of Win32 and/or NT Executive system calls to equivalent POSIX calls and vice-versa. Also, general interests in maintaining a large production piece of code on both Unix and WinNT while trying to keep as much code as possible in common.

My lab is beginning to use NT as a platform for our research systems. I am involved with projects that are using NT for user interfaces, experimental middleware, and network services. We are primarily developing in Java and Python at this point.
I am particularly interested in the ways in which we can interface our research prototypes with the application environment that so many people know and use. This requires extending the operating system and replacing functional components with our own enhanced versions. It can also require understanding how applications behave and exercising programmatic control over them.

I am involved in supporting a growing number of Windows NT workstations and in planning for smooth integration of the NT5.x DNS paradigm with the legacy DNS installation Motorola already supports behind our firewall.

Usage Abstract for participation to the Windows NT workshop J.A. Verdoes Shell Services International P.O. BOX 60 2280 AB Rijswijk The Netherlands e-mail: j.a.verdoes@siep.shell.com fax +31-70-3113110
Shell International Exploration and Production - Research and Technology Services mission is to provide services, consultancy and technology to the Royal Dutch Shell s Operating Units.
The technical and IT side of this involves:
- developing new algorithms for all aspects of exploration and production.
- exploiting new IT technologies to business advantage.
- developing new ways of running very CPU and IO intensive applications
- developing new ways of running applications that require the very large data sets that are common in seismic processing.
- disseminating these technologies to the Operating Units
- provide services with the above technologies, or special studies, for Operating Units that don t have the facilities themselves.
To day 500+ UNIX systems, including a number of IMB SP2 towers, are used to do large to very large scale scientific computing in the services as well as in research areas, and for the systems development of the program suites. 1500+ customers run a standards Windows 3.11/Novell environment. Some Windows 3.11 PCs are also used for software development and providing operational support to some remote locations, some are used for lab automation systems. The PC environment is now migrating to a standard Windows NT/Office 97 environment.
As a result of this are the behaviour, limitations and possibilities of the various UNIXes, platforms, peripherals, network etc. for the above usage well understood.
The decision of Shell s EP sector to standardise on NT for its personal productivity tools in each desktop are drivers for the use PCs and thus NT in technical and research computing.
The business wants it, but experience in applying NT in all of the above areas still has to be acquired before NT can be deployed for this on a larger scale.

I work as a systems & security engineer at CEA, France. I currently participate to the integration of Windows NT in our IT infrastructure.
My interests linked to Windows NT are :
- securing Windows NT,
- centralizing administration of Windows NT,
- integrating Unix & NT,
- learning about future directions for Windows NT,
- understanding Windows NT internals.
I expect from this conference to get the best information available on these subjects.
I also participate in a group about Windows NT security ( OSSIR, a French non-profit IT security association ).

I'm interested in finding more about Windows NT.

At QUALCOMM, we have a mixed environment of 8,000 NT/95 clients, 2000 Unix Workstations, and 1000 Macintosh clients. We are working towards a master directory to manage our users and group accounts, as well as providing a single sign-on interface to our users while integrating this mixed environment. Currently we are utilizing Novell's NDS, Microsoft’s Domain Structure, AFS Authentication, and NIS for this task. We are also using SAMBA and NAC storage solutions to bridge some of the multi-platform issues, and utilizing native storage solutions when ever possible. We are utilizing SMS for our automated software distribution. This is extremely useful in our environment, since end users are not granted 'Administrator' privileges, and therefor SMS must be used to properly install applications for the user.
For the last few months most of my time and effort has been the development of our in-house backup solution for Windows NT, and Windows 95. This new system uses a highly intelligent data management system that reduces or eliminates the need for redundant information being sent across the wire during a backup and looks to reduce our overall on-the-wire data transfers for backups almost %70 when the system is fully deployed. Look for my paper during the Lisa NT conference on Saturday “Designing an Optimized Enterprise Backup Solution with Intelligent Data Management.
I have also been spending some time looking into clustering and highly available systems in the Windows NT platform centering on Vinca, Veritas, and Wolfpack. As always working with some next generation beta servers and storage solutions from Compaq on the hardware side.

NetApp filers communicate with their clients using NFS, CIFS and HTTP. Finding ways to harmonize the locking metaphors and state requirements of these competing protocols is a challenging problem. Some key areas of conflict are seemingly not resolvable from first principles. Using market pressure to resolve them is uncomfortable for a researcher, but necessary to advance the product.

>From the perspective of a CPU architect, the Windows NT operating system stresses a processor's resources in different ways than application benchmarks do. To characterize how NT exercises a processor's branch prediction structures, caches, TLBs, and multi-processing features, we use performance monitor driven analysis tools (e.g. VTune) as well as trace-based simulation and performance analysis tools running on Windows NT.
I am particularly interested in how these techniques can be extended and combined with existing methods to:
(a) Better understand the behavior of the NT operating system.
(b) Improve the performance of the NT operating system.
(c) Make better use of processor capabilities.
(d) Improve future processor designs.
Rumi Zahir, CPU Architect, Intel Corporation rzahir@mipos2.intel.com RN6-16, 2200 Mission College Blvd., Santa Clara, CA95052-8119

I am a member of the NDIS group within the NT Networking and Communications Group at Microsoft. My primary responsibility is investigating high-performance network communication architectures.
I've recently joined the group after completing my PhD (May'98) from the Real-Time Computing Lab at the University of Michigan where I worked on real-time operating systems (synchronization, communication, and task scheduling issues).