| M. Hiltunen and R. Schlichting, "Constructing a configurable group RPC service," in Proceedings of the 15th International Conference on Distributed Computing Systems, Vancouver, BC, Canada, May 1995, pp. 288-- 295. |
....time out and cancel mechanisms to ensure communication correctness. DCE does not include group RPC communication, useful for some consistency protocols. But it is still possible to implement group RPC with point to point RPC mechanisms. For more details the interested reader is referred to [11]] As is shown in [8] 9] 10] RPC can be effectively applied for coherence protocols implementation. 4.4 DCE Security Services Executing remote procedures over a computer network imposes serious requirements for security. The DCE RPC includes an interface to the DCE Security Services. It ....
Hiltunen M. A., and Schlichting R. D.: Constructing a Configurable Group RPC Service, 15 th IEEE IC on Distributed Computing System, Vancouver, May 1995, Pages 288--295.
....to adapt to application needs is now a growing concern in the software engineering domain. This has led to the definition of notations to ease the development of customized systems using existing software. Examples of environments offering such a facility can be found in (Batory and O Malley 1992, Hiltunen and Schlichting 1995, Struman and Agha 1994) These proposals differ from ours in that we are addressing customization of execution platforms based on the refinement of requested dependability properties, while they provide a way to construct such platforms based on its adequate structuring. Thus, these environments ....
Hiltunen M. A. and Schlichting R.D. (1995) Constructing a Configurable Group RPC Service. Proceedings of the 15th IEEE International Conference on Distributed Computing Systems, pages 288-295.
....the specification of required customization by application programmers, and the provision of adequate base operating systems so as to implement safe and efficient customization. Much work has been undertaken both in the operating system (e.g. see [2, 3, 4] and the programming language (e.g. see [5, 6, 7]) communities in order to address these needs. Our research is aimed at providing a framework easing both the specification of system level requirements (such as those related to availability, security, timeliness) within distributed applications, and the construction of corresponding customized ....
....tools to help her him to select the appropriate customized system. Proposals made in the programming language community relate to the system structuring issue. They define programming paradigms easing the development of customized system services from the description of system level software [5, 6]. Here too, the issue of selection of customized services within applications is not dealt with. Furthermore, to our knowledge, existing solutions address only one specific type of customization, such as fault tolerance, while the Aster system provides a framework for various kinds of ....
M. A. Hiltunen and R. D. Schlichting, "Constructing a configurable group rpc service," in Proceedings of the Fifteenth International Conference on Distributed Computing Systems, pp. 288--295, 1995.
....group [Che86, Coo90, CGR88, SS90] This is especially useful for constructing fault tolerant services using replicated servers. It is also possible to have a group of clients interacting with a group of servers. GRPC has several semantic aspects in addition to those found in point to point RPC [HS95a] These include: ffl Ordering. FIFO order guarantees that all calls issued by a single client are executed by each server in the same order. Total order guarantees that all calls by all clients are executed in the same order by each server. Causal ordering of client requests preserves the causal ....
....then one server acts as a coordinator that determines the ordering of requests. Thus, for each request, the coordinator sends an ordering message, Order(Req(x) to all other servers. 5. 1 Group RPC Micro protocols The micro protocol suite is based on the semantic variations of GRPC described in [HS95a] the categories that follow represent semantic variations of termination, ordering, communication, collation, call style, membership, and failure. 5.1.1 Termination Semantics Termination semantics specify the guarantees that are given about the termination of a call. Included in the client ....
M. Hiltunen and R. D. Schlichting. Constructing a configurable group RPC service. In Proceedings of the 15th IEEE Symp. on Distributed Computing Systems, Vancouver, BC, May 1995.
....approaches are aimed at lower level customization, they do not provide the application programmer with tools to help her him to select the appropriate customized system. Work has also been done in the programming language community so as to ease construction of customized systems (e.g. see [18, 19, 4, 7]) However, the existing proposals do not address correctness of the customization with respect to application requirements. More generally, they do not consider the formal verification of matching of application requirements with customized systems. From the perspective of the formal method ....
M. A. Hiltunen and R. D. Schlichting, "Constructing a configurable group rpc service," in Proceedings of the Fifteenth International Conference on Distributed Computing Systems, pp. 288--295, 1995.
....written in Java that runs on multiple platforms. CTP is being implemented using the C version of Cactus on a cluster of Pentiums running Red Hat Linux version 6.2. Other prototype services that have been successfully implemented using Cactus or the predecessor Coyote system [14] include group RPC [15], membership [16] and a real time channel abstraction [8] This paper has several goals. The first is to argue that finegrain configurability and extensibility are valuable characteristics for transport protocols. The second is to describe a realization of this philosophy in the form of CTP. The ....
M. Hiltunen and R. Schlichting, "Constructing a configurable group RPC service," in Proceedings of the 15th International Conference on Distributed Computing Systems, Vancouver, BC, Canada, May 1995, pp. 288-- 295.
....manner with other protocols. Micro protocols within a given composite protocol are executed in an event driven manner, which enhances configurability by minimizing explicit references between micro protocols. This approach has been applied in a number of areas related to fault tolerant protocols [13, 15, 14], and is currently being implemented as an extension to the x kernel [16] on the Mach operating system [4] The major components of this approach are micro protocols, events, and a software framework: 1. Micro protocols: Implement a well defined property; structured as a collection of event ....
M. Hiltunen and R. Schlichting. Constructing a configurable group RPC service. In Proceedings of the 15th International Conference on Distributed Computing Systems, pages 288--295, May 1995.
....written in Java that runs on multiple platforms. An initial prototype of SecComm has been implemented on the MK version of Cactus on a cluster of Pentiums. Other prototype services that have 4 been successfully implemented using Cactus or the predecessor Coyote system [BHSC98] include group RPC [HS95] membership [HS98] and a real time channel abstraction [HSH 99] 3 SecComm Design 3.1 Application programming interface The SecComm service allows a higher level service or application to open secure connections and then send and receive messages through these connections. The specific ....
M. Hiltunen and R. Schlichting. Constructing a configurable group RPC service. In Proceedings of the 15th International Conference on Distributed Computing Systems, pages 288--295, Vancouver, BC, Canada, May 1995.
....multiple platforms. CTP is being implemented using the C version of Cactus on a cluster of SUBMITTED FOR PUBLICATION 101 Pentiums running Red Hat Linux version 6.2. Other prototype services that have been successfully implemented using Cactus or the predecessor Coyote system [14] include group RPC [15], membership [16] and a real time channel abstraction [8] This paper has several goals. The first is to argue that fine grain configurability and extensibility are valuable characteristics for transport protocols. The second is to describe a realization of this philosophy in the form of CTP. ....
M. Hiltunen and R. Schlichting, "Constructing a configurable group RPC service," in Proceedings of the 15th International Conference on Distributed Computing Systems, Vancouver, BC, Canada, May 1995, pp. 288--295.
....We then describe the design of a suite of micro protocols supporting the construction of custom composite protocols in these areas. Although the design approach supported by Coyote has been used previously to construct configurable communication oriented services such as group RPC and membership [6, 8, 9], this is its first application outside the area of highly dependable computing. 2 Overview of Mobile Systems All mobile systems have a number of common core elements, as shown in Figure 1. These include: ffl Mobile Hosts. A roaming device equipped with a wireless communication link. May vary ....
M. Hiltunen and R. Schlichting. Constructing a configurable group RPC service. In Proceedings of the 15th IEEE Conference on Distributed Computing Systems, pages 288--295, Vancouver, BC, May 1995.
....p i and p j : ffl ind(p i ; p j ) C = fp i ; p j ; p i p j g. ffl dep(p i ; p j ) C = fp i p j ; p j g. ffl con(p i ; p j ) C = fp i ; p j g. Naturally these rules generalize to any number of properties. Consider the abstract properties of a configurable group RPC service as an example [15]. An RPC call is typically blocking (synchronous) but non blocking (asynchronous) variants of RPC have also been implemented. A call cannot be both blocking and non blocking at the same time, however, thus these two properties of RPC conflict by definition. Let atomicity be the property that each ....
M. Hiltunen and R. Schlichting. Constructing a configurable group RPC service. In Proceedings of the 15th International Conference on Distributed Computing Systems, pages 288-- 295, May 1995.
....a modular and configurable implementation of RPC using the model described in the previous section. All major micro protocols are described, with pseudo code for several shorter micro protocols given to illustrate the programming style. Detailed pseudo code for all micro protocols can be found in [10]. For brevity, the focus here is on group RPC. Point to point RPC can be seen as a special case in this implementation, although in practice it would likely be implemented separately to obtain a more compact and efficient protocol. In following the tenets of the model, the RPC service is ....
M. A. Hiltunen andR. D. Schlichting. Constructing a configurable group RPC service. Technical Report 94-28, Department of Computer Science, University of Arizona, Tucson, AZ, Oct 1994.
....included, then one server acts as a coordinator that determines the ordering of requests. Thus, for each request, the coordinator sends an ordering message, Order(Req(x) to all other servers. 4. 1 Micro protocols The micro protocol suite is based on the semantic variations of GRPC described in [HS95a] the categories that follow represent semantic variations of termination, ordering, communication, collation, call style, membership, and failure handling. 4.1.1 Termination semantics Termination semantics specify the guarantees that are given about the termination of a call. Included in the ....
....to the stream of request and ordering messages. Servers must communicate reliably or communication would halt if an ordering message was lost. This functionality is provided by CRET. Formal dependency and configuration graphs for a similar suite of group RPC micro protocols can be found in [HS95a, Hil96] UNIQUE or or or BND UBND 1AC AAC no order, unreliable no order, reliable no order, unreliable, unique no order, reliable, unique FIFO, reliable, unique FIFO, reliable total, reliable, unique total, fifo, reliable, unique (no protocols) TOTAL SMEM ACK RET RET ACK RET FIFO FIFO ACK RET ....
M. Hiltunen and R. Schlichting. Constructing a configurable group RPC service. In Proceedings of the 15th IEEE Conference on Distributed Computing Systems, Vancouver, BC, May 1995.
....are ordered, and how partitions are handled. This work builds on earlier papers that describe the abstract properties of membership services [HS95b, HS95c] as well as related work in which this approach has been used for other services such as atomic multicast and group RPC [GBB 95, HS93, HS95a] The membership service will eventually be ported to an x kernel based system that is currently under development [BS95] 2 Preliminaries 2.1 System structure A membership service can be viewed as an underlying software layer that generates messages indicating changes in membership and ....
M. A. Hiltunen and R. D. Schlichting. Constructing a configurable group RPC service. In Proceedings of the 15th International Conference on Distributed Computing Systems, pages 288-- 295, Vancouver, BC, Canada, May 1995.
.... from a previous system called Coyote [BHSC98] which has been used to construct highly customizable versions of communication services without real time 1 Note that the accuracy of the actual delivery time will depend on the system clock granularity, however constraints, including group RPC [HS95, BS95] membership [HS98] and atomic multicast [GBB 95] Although CactusRT is linked with the x kernel model, the concept of microprotocols and the associated execution model can be implemented using any number of different vehicles. For example, a second version of Cactus without support ....
M. Hiltunen and R. Schlichting. Constructing a configurable group RPC service. In Proceedings of the 15th International Conference on Distributed Computing Systems, pages 288--295, Vancouver, BC, Canada, May 1995.
....included, then one server acts as a coordinator that determines the ordering of requests. Thus, for each request, the coordinator sends an ordering message, Order(Req(x) to all other servers. 4. 1 Micro protocols The micro protocol suite is based on the semantic variations of GRPC described in [HS95a] the categories that follow represent semantic variations of termination, ordering, communication, collation, call style, membership, and failure handling. Termination semantics. Termination semantics specify the guarantees that are given about the termination of a call. These micro protocols ....
....to the stream of request and ordering messages. Servers must communicate reliably or communication would halt if an ordering message was lost. This functionality is provided by CRET. Formal dependency and configuration graphs for a similar suite of group RPC micro protocols can be found in [HS95a, Hil96] 4.3 Performance Measurements Tests consisted of one or more clients sending a 4 byte integer to one or more servers, which respond with an integer. Each test makes 1000 RPC calls and was run 10 times. The round trip times are the average of the 10 test runs. A version of Sun RPC ....
M. Hiltunen and R. Schlichting. Constructing a configurable group RPC service. In Proceedings of the 15th IEEE Conference on Distributed Computing Systems, pages 288--295, Vancouver, BC, May 1995.
....by allowing internal structuring of x kernel protocols as collections of micro protocols. CactusRT has evolved from a previous system called Coyote, which has been used to construct highlycustomizable versions of communication services without real time constraints, including group RPC [HS95, BS95] membership [HS98] and atomic multicast [GBB 95] Although CactusRT is linked with the x kernel model, the concept of micro protocols and the execution model supported by the system can be implemented using any number of different vehicles. For example, a second version of Cactus ....
M. Hiltunen and R. Schlichting. Constructing a configurable group RPC service. In Proceedings of the 15th International Conference on Distributed Computing Systems, pages 288--295, Vancouver, BC, Canada, May 1995.
....servers, or message ordering options needed for more complex applications. Additional micro protocols are currently being implemented, including other reply ordering policies, two different policies for handling orphan computations, and provisions for atomic execution of requests by the server. [21] elaborates further on various abstract properties of RPC and describes the different combinations that are possible using our micro protocol approach. The second set of performance figures are from a null composite protocol designed to measure the event mechanism. Each of the tests measured ....
....construction of customized services with properties tailored to the needs of a given application. It also encourages experimentation with different communication substrates for a given application. The topic of when and how micro protocol variants can be configured into a system is addressed in [19, 21, 22] for different types of network services, including membership and group RPC. The use of this approach for constructing a customized atomic multicast protocol for a version of the Linda coordination language with fault tolerance extensions [2] is described in [17] The prototype implementation ....
M. Hiltunen and R. D. Schlichting. Constructing a configurable group RPC service. In Proceedings of the 15th IEEE Conf. on Distr. Computing Syst., Vancouver, BC, May 1995. To appear.
....of the model is currently underway using the OSF RI MK 7.2 real time operating system [Rey95] on a cluster of Pentium PCs. An earlier version of the system has been used to develop highly configurable versions of communication services without real time constraints, including group RPC [HS95, BS95] membership [HS97] and atomic multicast [GBB 95] The rest of the paper is organized as follows. Section 2 describes our model for configurable real time services and outlines an implementation of the model designed for OSF RI MK 7.2 and the x kernel [HP91] We also outline the ....
M. A. Hiltunen and R. D. Schlichting. Constructing a configurable group RPC service. In Proceedings of the 15th International Conference on Distributed Computing Systems, pages 288--295, Vancouver, BC, Canada, May 1995.
....[CHS98] These protocols are not configurable in the same sense as our approach, however. In particular, rather than customizing one protocol, the user must select from a related collection of protocols. Other work dealing with multiple failure models is that on hybrid fault models [TP88, LR93, WHS95] The basic idea is to detect a range of different types of failures at the same time by using multiple failure detection techniques. This approach allows the system to tolerate a larger number of failures than traditional Byzantine algorithms, since simpler failures that require less redundancy ....
....techniques. This approach allows the system to tolerate a larger number of failures than traditional Byzantine algorithms, since simpler failures that require less redundancy can be distinguished from true Byzantine failures. Some algorithms designed for hybrid fault models, in particular [WHS95] allow specification of the maximum number of faults to be tolerated for each different fault model. Although such an algorithm could exhibit comparable flexibility to our approach with respect to choice of failure model, our approach allows greater optimization of the algorithm used in each ....
[Article contains additional citation context not shown here]
M. Hiltunen and R. Schlichting. Constructing a configurable group RPC service. In Proceedings of the 15th InternationalConference on Distributed ComputingSystems, pages 288--295, Vancouver, BC, Canada, May 1995.
....Micro Protocols In this section, we outline the micro protocols for the configurable RPC service. All major micro protocols are described, with pseudo code for several shorter micro protocols given to illustrate the programming style. Detailed pseudo code for all micro protocols can be found in [HS94b] The section starts with a base micro protocol RPCMain followed by collection of micro protocols for various aspects of remote procedure calls ranging from ones for user thread managements to dealing with orphans. 162 7.3.1 RPCMain The RPCMain micro protocol handles the main control flow of ....
M. A. Hiltunen and R. D. Schlichting. Constructing a configurable group RPC service. Technical Report 94-28, Department of Computer Science, University of Arizona, Tucson, AZ, Oct 1994.
....or non blocking, where the invoker continues execution without waiting. Event registration, detection, and invocation are implemented by a standard runtime or framework that is linked with the micro protocols. The framework also supports 2 Text for this sub section appeared previously in [10] shared data (e.g. messages) that can be accessed by the micro protocols configured into the framework. The object formed by the linking of a collection of micro protocols and associated framework is known as a composite protocol. Once created, such a composite protocol can be composed in a ....
Matti A. Hiltunen and Richard D. Schlichting. Constructing a configurable group rpc service. Submitted to 15th Conference on Distributed Computing Systems, 1995.
....in this area have focused on two lines of investigation. The first is identifying and classifying properties of various fault tolerant services, including the numerous variants that have been developed. Services for which substantial progress has been made include membership [Hilt95c] group RPC [Hilt95b], and atomic multicast [Gued95] preliminary work has also been done on atomic transactions and protocols for mobile computing, an area closely related to fault tolerance. As just one example of the range of customization this approach engenders, we have identified over 20 properties and property ....
M. Hiltunen and R. Schlichting. Constructing a Configurable Group RPC Service. Proc. 15th Conf. on Distr. Computing Syst., Vancouver, Canada (May 1995), 288-295.
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