I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale. Applying Optimization Patterns to the Design of Real-time ORBs. In Proceedings of the ### Conference on Object-Oriented Technologies and Systems, pages 145--159, San Diego, CA, May 1999. USENIX.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....as well as the generic service provision, but they do not concentrate on efficient communication protocol processing or multimedia Quality of Service (QoS) support in the first place. The latter aspects started to be dealt by in recent work, however, have not been finished yet. The approach TAO [6] deals with investigations of CORBA based middleware for high speed networks and applications. The AQuA approach [7] develops adaptable, object oriented, distributed computing systems while applying quality objects to manage system characteristics. The support of computational grids for ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, W. Kachro, and A. Gokhale, "Applying optimization patterns to design real-time ORBs," in 5th USENIX Conference on OO Technologies and Systems COOTS'99, May 1999, San Diego, California, U.S.A.

No context found.

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale. Applying Optimization Patterns to the Design of Real-time ORBs. In Proceedings of the ### Conference on Object-Oriented Technologies and Systems, pages 145--159, San Diego, CA, May 1999. USENIX.

No context found.

Irfan Pyarali, Carlos O'Ryan, Douglas C. Schmidt, Nanbor Wang, Vishal Kachroo, and Aniruddha Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5

No context found.

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 th Conference on Object-Oriented Technologies and Systems, (San Diego, CA), pp. 145--159, USENIX, May 1999.

No context found.

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Realtime ORBs," in Proceedings of the 5 th Conference on Object-Oriented Technologies and Systems, (San Diego, CA), pp. 145--159, USENIX, May 1999.

No context found.

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

.... including static [25] and dynamic [26] scheduling, event processing [9] I O subsystem [27] and pluggable protocol [28] integration, synchronous [29] and asynchronous [30] ORB Core architectures, systematic benchmarking of multiple ORBs [31] and optimization principle patterns for ORB performance [32]. This paper extends our previous work by focusing on the following dimensions in the ORB endsystem design space: Identifying key aspects of CORBA implementations that deliver both real time and fault tolerant properties simultaneously . Evaluating the suitability of FT CORBA as the ....

Irfan Pyarali, Carlos O'Ryan, Douglas C. Schmidt, Nanbor Wang, Vishal Kachroo, and Aniruddha Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 Conference on Object-Oriented Technologies and Systems, San Diego, CA, May 1999, USENIX.

....(RTUs) which perform protocol processing in user space [45] RTUs were applied in the multi media server developed in project MARS [49] which provides QoS guarantees within the filesystem and data on disk. Figure 3 show how our real time I O (RIO) subsystem has been implemented in Solaris [50] using the leveraged technologies described above. The key optimizations used in our RIO subsystem to maximize efficiency and provide rQ 2 rQ 3 1 sQ 23 sQ sQ rQ rQ wQ Classifier user thread wQ wQ rQ (routing tables) IP Multiplexor wQ rQ rQ SYS:61 rkt3 kernel Best Effort ....

....Figure 4: Components in the TAO Real time ORB Endsystem 1. Highly scalable and predictable request demultiplexing strategies: TAO s Object Adapter demultiplexing strategies route requests to objects in constant, i.e. O(1) time regardless of the number of objects and IDL interface operations [50]. 2. IDL compiler optimizations: TAO s IDL compiler can generate optimized compiled and or interpreted stubs and skeletons, which enable applications to make fine grained time space tradeoffs at the presentation layer [57] 3. Efficient and predictable concurrency and connection models: TAO s ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

....and realtime applications and middleware. For instance, research conducted as part of the DARPA Quorum project [8] the QuO project at BBN [9] and the TAO [10] and TMO [11] projects at Washington University and UC Irvine, have identified key design patterns [12] optimization principle patterns [13], and frameworks [14, 15] that instantiate these patterns into highquality, QoS enabled DOC middleware components [16] 2. Maturation of standards: Over the past decade, the OMG s suite of standards has matured considerably, particularly with respect to high performance and real time systems. For ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

....of Work The scope of this effort is to create a real time ORB that s compliant with the new OMG Minimum CORBA [12] specification and embed this ORB within Cisco network elements, such as IP routers and ATM switches. This embedded ORB will leverage and enhance the advanced real time features [21, 19] developed by Washington University s Center for Distributed Object Computing to implement network element management and control protocols efficiently, scalably, and predictably using standard based middleware. This proposal describes the specific tasks to be performed during the 12 months of the ....

....QoS requirements, as well as best effort requirements. 18] TAO s features and optimizations include an ORB Core that minimizes context switching, synchronization, dynamic memory allocation, and data movement [22] a highly scalable Object Adapter that demultiplexes requests in constant time [19]; an optimizing IDL compiler [6] real time I O subsystem [9] and a global resource allocation and scheduling framework [21, 5] CORBA protocol model synopsis: CORBA Inter ORB Protocols (IOP)s define interoperability between ORB endsystems. IOPs provide data representation formats and ORB ....

[Article contains additional citation context not shown here]

Irfan Pyarali, Carlos O'Ryan, Douglas C. Schmidt, Nanbor Wang, Vishal Kachroo, and Aniruddha Gokhale. Applying Optimization Patterns to the Design of Real-time ORBs. In Proceedings of the 5 Conference on Object-Oriented Technologies and Systems, San Diego, CA, May 1999. USENIX.

.... including static [10] and dynamic [11] scheduling, event processing [12] I O subsystem [13] and pluggable protocol [14] integration, synchronous [15] and asynchronous [16] ORB Core architectures, systematic benchmarking of multiple ORBs [17] optimization principle patterns for ORB performance [18] and high performance architectures for Fault tolerant CORBA [19, 20] This paper focuses on another dimension in the ORB endsystem design space: providing dependability using Faulttolerant CORBA (FT CORBA) to Real time CORBA (RTCORBA) based DRE systems. This paper is organized as follows: ....

Irfan Pyarali, Carlos O'Ryan, Douglas C. Schmidt, Nanbor Wang, Vishal Kachroo, and Aniruddha Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 Conference on Object-Oriented Technologies and Systems, San Diego, CA, May 1999, USENIX.

....ORB share components that are not in the critical path (such as configuration factories) or that do not contribute to priority inversion (such as connectors) In standard CORBA, POAs are shared, which could be a source of non determinism. However, our previous work on CORBA request demultiplexing [22] describes techniques to ensure predictable POA demultiplexing and reduce average and worstcase overhead regardless of organization of an ORB s POA hierarchy or number of POAs, servants, or operations. Likewise, our work on efficient, scalable, and, predictable dispatching components [23] shows ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

....TAO is an open source, high performance real time ORB written in C . TAO had the first implementation of the POA specification [18] The design of the POA is based on several design patterns, many of which have been adopted in ZEN. TAO also uses an optimized set of request processing strategies [22], e.g. active demultiplexing and perfect hashing. These strategies allow TAO s POA to provide constant time lookup of servants based on object keys and operation names contained in CORBA requests. In ZEN, we have based our design on several optimizations used in the design of TAO s POA, e.g. ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

....ORB share components that are not in the critical path (such as configuration factories) or that do not contribute to priority inversion (such as connectors) In standard CORBA, POAs are shared, which could be a source of non determinism. However, our previous work on CORBA request demultiplexing [22] describes techniques to ensure predictable POA demultiplexing and reduce average and worstcase overhead regardless of organization of an ORB s POA hierarchy or number of POAs, servants, or operations. Likewise, our work on efficient, scalable, and, predictable dispatching components [23] shows ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

.... including static [10] and dynamic [5] scheduling, event processing [8] I O subsystem [11] and pluggable protocol [12] integration, synchronous [13] and asynchronous [14] ORB Core architectures, systematic benchmarking of multiple ORBs [15] patterns for ORB extensibility [7] and ORB performance [16]. This paper extends our previous work [8] on real time extensions to the CORBA Event Service to show how this service can support the QoS requirements of largescale distributed interactive simulations by using IP multicast to federate multiple Event Channels and conserve network resources. In ....

....event should be dispatched to a consumer, it must decide which thread will perform the dispatching. As shown in Figure 4, there are several alternatives. Using the same thread that received the event is efficient, e.g.it reduces context switching, synchronization, and data copying overhead [16]. However, this design may expose the Event Channel to misbehaving consumers. Moreover, to avoid priority inversions in real time systems, events must be dispatched by a thread at the appropriate priority. Likewise, highly scalable systems Although this design may increase context switching ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

.... including static [10] and dynamic [5] scheduling, event processing [8] I O subsystem [11] and pluggable protocol [12] integration, synchronous [13] and asynchronous [14] ORB Core architectures, systematic benchmarking of multiple ORBs [15] patterns for ORB extensibility [7] and ORB performance [16]. This paper extends our previous work [8] on real time extensions to the CORBA Event Service [17] as follows: We describe the patterns that guided the design and optimization of a flexible and scalable Event Service framework that allows developers to select implementation strategies that are ....

....event should be dispatched to a consumer, it must decide which thread will perform the dispatching. As shown in Figure 11, there are several alternatives. Using the same thread that received the event is efficient, e.g. it reduces context switch, synchronization, and data copying overhead [16]. However, this design may expose an event channel to misbehaving consumers. Moreover, to avoid priority inversions in real time systems, events must be dispatched by a thread at the appropriate priority. Likewise, highly scalable systems may want to use a pool of threads to dispatch events, ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 th Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

.... including static [10] and dynamic [5] scheduling, event processing [8] I O subsystem [11] and pluggable protocol [12] integration, synchronous [13] and asynchronous [14] ORB Core architectures, systematic benchmarking of multiple ORBs [15] patterns for ORB extensibility [7] and ORB performance [16]. This paper extends our previous work [8] on real time extensions to the CORBA Event Service [17] as follows: ffl We describe the patterns that guided the design and optimization of a flexible and scalable Event Service framework that allows developers to select implementation strategies that ....

....event should be dispatched to a consumer, it must decide which thread will perform the dispatching. As shown in Figure 11, there are several alternatives. Using the same thread that received the event is efficient, e.g. it reduces context switch, synchronization, and data copying overhead [16]. However, this design may expose an event channel to misbehaving consumers. Moreover, to avoid priority inversions in real time systems, events must be dispatched by a thread at the appropriate priority. Likewise, highly scalable systems may want to use a pool of threads to dispatch events, ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 th Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

....has been matched successfully against an event, we do not want to repeat this matching operation for each proxy connected to the same admin. Solution Optimization principle pattern of passing hints : We use a variation of the optimization principle pattern passing information between layers [16]. This pattern is commonly used in protocol stack optimizations where each protocol layer passes certain information to the layer on top to help it avoid demultiplexing overhead. Applying the solution in TAO: In TAO s Notification Service, we pass a hint to proxy objects to skip filter evaluation ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Realtime ORBs," in Proceedings of the 5 th Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

.... including static [10] and dynamic [5] scheduling, event processing [8] I O subsystem [11] and pluggable protocol [12] integration, synchronous [13] and asynchronous [14] ORB Core architectures, systematic benchmarking of multiple ORBs [15] patterns for ORB extensibility [7] and ORB performance [16]. This paper extends our previous work [8] on real time extensions to the CORBA Event Service to show how to support the QoS requirements of large scale distributed interactive simulations using IP multicast to federate multiple event channels and conserve network resources. In addition, we ....

....event should be dispatched to a consumer it must decide which thread will perform the dispatching. As shown in Figure 6, there are several alternatives. Using the same thread that received the event is efficient, e.g. it reduces context switching, synchronization, and data copying overhead [16], but potentially exposes the event channel to misbehaving consumers. Moreover, to avoid priority inversions in real time systems, events must be dispatched by a thread at the appropriate priority. In turn, highly scalable systems may want to use a pool of threads to dispatch the events, thereby ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 th Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

....reference count described above. 5 important benefit of this pattern is that the level of priority inversion does not depend on the duration of the upcall. In fact, priority inversion can be calculated as a function of the time needed to search the dispatching table. In our previous research [19], we have shown that very low and bounded search times can be achieved using techniques like active demultiplexing and perfect hashing. Implementations that use these techniques in conjunction with the serialization pattern described here can achieve predictable dispatching with bounded priority ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 th Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

No context found.

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 th Conference on Object-Oriented Technologies and Systems, USENIX, (San Diego, CA), May 1999.

.... and realtime ORB endsystem design, including static [7] and dynamic [8] scheduling, event processing [9] I O subsystem [10] and pluggable protocol [11] integration, ORB Core architectures [12] systematic benchmarking of multiple ORBs [13] patterns for ORB extensibility [14] and ORB performance [15]. This paper focuses on a previously unexplored dimension in the high performance and real time ORB endsystem design space: the design and optimizations used to implement the standard CORBA asynchronous method invocation (AMI) callback model. The vehicle for our research on high performanceand ....

....model. The vehicle for our research on high performanceand real time CORBA is TAO [7] TAO is an open source 1 , CORBA compliant ORB designed to address applications with stringent quality of service (QoS) requirements. In addition to being the first ORB with a standard Portable Object Adapter [15], TAO was the first ORB to implement the standard CORBA AMI callback model. 1 The source code and documentation for TAO can be downloaded from www.cs.wustl.edu #schmidt TAO.html. Related work: The AMI polling model stems from research on programming language support for distributed computing. ....

[Article contains additional citation context not shown here]

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 th Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

.... a light workload, the half sync half reactive thread pool design will incur a dynamic memory allocation, multiple synchronization operations, and a context switch to pass a request message between the network I O thread and a worker thread, which makes even the best case latency unnecessarily high [4]. Moreover, if the OLTP server is run on a multi processor, significant overhead can occur from processor cache coherency protocols required to transfer command objects between threads [5] If the OLTP servers run on an operating system platform that supports asynchronous I O efficiently, the ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Realtime ORBs," in Proceedings of the 5 th Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

.... ########## Figure 1: Layering of TAO s A V Streaming Service Atop the TAO ORB Endsystem tegration, synchronous [Schmidt et al. 2000a] and asynchronous [Arulanthu et al. 2000] ORB Core architectures, event processing [Harrison et al. 1997] optimization principle patterns for ORB performance [Pyarali et al. 1999], and the performance of various commercial and research ORBs [Gokhale and Schmidt, 1996, Schmidt et al. 1998b] over high speed ATM networks. This chapter focuses on another important topic in ORB endsystem research: the design and performance of the CORBA A V Streaming Service specification. ....

Pyarali, I., O'Ryan, C., Schmidt, D. C., Wang, N., Kachroo, V., and Gokhale, A. (1999). Applying Optimization Patterns to the Design of Real-time ORBs. In Proceedings of the 5 th Conference on Object-Oriented Technologies and Systems, San Diego, CA. USENIX.

....developers to selectively trade off time and space, which is crucial for high performance, realtime, and or embedded distributed systems. Real time Object Adapter: An Object Adapter associates servants with the ORB and demultiplexes incoming requests to servants. TAO s real time Object Adapter [28] uses perfect hashing [29] and active demultiplexing [28] optimizations to dispatch servant operations in constant O(1) time, regardless 7 R R U U N N T T I I M M E E S S C C H H E E D D U U L L E E R R HIGH SPEED NETWORK INTERFACES (e.g. APIC, VME) Z Z E E R R O O C C O ....

....is crucial for high performance, realtime, and or embedded distributed systems. Real time Object Adapter: An Object Adapter associates servants with the ORB and demultiplexes incoming requests to servants. TAO s real time Object Adapter [28] uses perfect hashing [29] and active demultiplexing [28] optimizations to dispatch servant operations in constant O(1) time, regardless 7 R R U U N N T T I I M M E E S S C C H H E E D D U U L L E E R R HIGH SPEED NETWORK INTERFACES (e.g. APIC, VME) Z Z E E R R O O C C O O P P Y Y B B U U F F F F E E R R S S RT RT ....

I. Pyarali, C. O'Ryan, D. C. Schmidt, N. Wang, V. Kachroo, and A. Gokhale, "Applying Optimization Patterns to the Design of Real-time ORBs," in Proceedings of the 5 th Conference on Object-Oriented Technologies and Systems, (San Diego, CA), USENIX, May 1999.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC