MOBILITY AND COMPLETENESS IN PUBLISH/SUBSCRIBE TOPOLOGIES

Transcription

1 MOBILITY AND COMLETENE IN UBLIH/UBCRIBE TOOLOGIE asu Tarkoma and Jaakko Kangasharju Helsinki Institute for Information Technology.O. Box 9800, FIN HUT, Finland ABTRACT The event paradigm and publish/subscribe systems allow clients to asynchronously receive information that matches their interests. The requirements of mobile computing present new challenges pertaining to event delivery that need to be solved. In this paper, we formally examine several state transfer protocols for different pub/sub topologies. The new results of this paper are the cost functions for both subscriber and publisher mobility, and investigation and formulation of completeness of subscriptions and advertisements. The results show that rendezvous points are good for pub/sub mobility, handovers in incomplete topologies are more costly than in complete, and the brokers involved with mobility have no way of detecting completeness based on local information alone. KEY WORD Mobile Computing, Content Distribution Networking, Large-scale Monitoring and rovisioning ystems 1 Introduction Two important middleware services are event monitoring and event notification. Environment monitoring and notification are usually provided by the event or notification service, which allows software components to communicate asynchronously [1]. In this paper, we focus on contentbased routing in which event brokers forward notifications based on a routing configuration established by advertisement and subscription messages. Most research on event systems has focused on event dissemination in the fixednetwork, where clients are usually stationary and have reliable, low-latency, and high bandwidth communication links. Recently, mobility and wireless communication have become an active topic in many research projects working with event systems, such as iena [2] and Rebeca [3]. In subscription semantics subscription messages are propagated throughout the pub/sub network and notifications are sent on the reverse path of subscriptions. Advertisement semantics extends this model by flooding advertisements throughout the network and propagating subscriptions on the reverse path of advertisement messages. Content-based routing and advertisement semantics was investigated in the iena project, and most current event systems use advertisements, for example Hermes [8] and Rebeca, because they improve scalability. Techniques such as covering relations-based optimization and filter merging are used to reduce the size of routing tables at servers. A filter F 1 is said to cover another filter F 2 when F 1 matches all the notifications matched by F 2. Filter merging is a technique for combining two filters to a filter that is equivalent to the disjunction of the input filters. In content-based routing servers typically propagate either non-covered filters or merged filters to minimize the size of routing tables. In advertisement-based pub/sub networks a succesful activation of a subscription may require that an advertisement is first propagated through the network, and then a connecting subscription is propagated on the reverse path to the source. These mechanisms become challenging when the topology needs to be reconfigured with the introduction of mobile components. In this paper, we follow the general notification data model: a notification is a set of 3-tuples <name, type, value>. A filter is a set of constraints on notifications. We assume that the pub/system follows the weakly valid routing configuration [4], which means that all updates will eventually finish and notifications will match all subscriptions that are terminated. In this paper, we call such update procedures that have ended succesfully complete in the topology, and use completeness to characterize the properties of pub/sub mobility. Definition 1 An advertisement A is complete in a pub/sub system iff there does not exist a broker s with an overlapping subscription that has not processed A. imilarly, a subscription is complete in iff there does not exist a broker s such that s has an an advertisement that overlaps with and is not active on s. In this model false negatives that occur during topology reconfiguration caused by stationary components are tolerated. A false negative is a notification that is published, but not delivered to a client with an active subscription that matches the event. Based on this notion, we define a mobility-safe pub/sub system as follows: Definition 2 A pub/sub system is mobility-safe if starting from an initial configuration C 0 at time T 0 and ending in a final configuration C e at time T e handovers will not cause any false negatives. We also relax Definition 2 by not considering server failures or by assuming that faults may be masked. False

2 positives may be removed by using event identifiers so they are not considered. In distributed pub/sub systems it is evident that after issuing a subscription it may take several hops before the subscription is activated for all publishers. During this time several notifications may be missed. In our mobility-aware weakly valid routing configuration false negatives that occur during topology reconfiguration caused by subscriptions and advertisements from stationary components are tolerated. In our subsequent examination we do not count these notifications as false negatives. Mobile clients change the scenario, because the flow endpoints are mobile and brokers transfer end-points using a handover process. False negatives that occur during mobility are not tolerated. In subsequent examination, terms complete and incomplete refer to the completeness and incompleteness of the subscriptions and advertisements involved in mobility. In this paper, we formally examine a state transfer protocol for acyclic pub/sub networks based on advertisement semantics. The new results of this paper are: 1. establishing the communication cost of pub/sub mobility, and 2. comparing three pub/sub mobility mechanisms and topologies: generic mobility support, acyclic graphs, and rendezvous-based topologies. Due to brevity, we do not present mobility-safety proofs for the mobility protocols for complete topologies. It is evident that once the topology is complete it is possible to construct an efficient mobility protocol. This motivates this investigation of different pub/sub topologies. The main motivation for a pub/sub mobility protocol is the avoidance of triangle routing through a designated home broker, which may be inefficient. Experimental results show that home broker based approaches do not perform well [5, 6]. Mobility protocols are also needed for load balancing subscribers and advertisers between brokers. The rest of the paper is organized as follows. ection 2 presents a generic pub/sub mobility protocol, in ection 3 we examine subscriber and publisher mobility in networks based on acyclic graphs. ection 4 examines rendezvous-based pub/sub. Finally, ection 5 presents the conclusions. 2 Generic Mobility upport The iena event system was extended with generic mobility support, which uses existing pub/sub primitives: publish and subscribe [2]. The benefit of a generic protocol is that it may work on top of various pub/sub systems and requires no changes to the system AI. On the other hand, the performance of the mobility support decreases, because mobility specific optimizations are difficult to realize when the underlying topology is hidden by the AIs. Indeed, in this section we show that a general AI-based pub/sub mobility support may have a very high cost in terms of message exchanges. The use of advertisement semantics was not discussed for the iena mobility support service so we extend the generic model to cover also advertisements. The mobility protocol proceeds in four distinct phases: first, the target subscribes all events, then the target and source synchronize by sending ping and pong events in order to ensure that the subscriptions have taken effect, the source unsubscribes, and finally the events are relocated and merged. In addition, there may be further costs triggered by changes in the subscription tables of the intermediate routers. Let N denote the number of brokers in the network and n the number of brokers between the source and target of mobility. The cost structure for this procedure is presented in Table 1 as the number of brokers that will be updated during each phase. For advertisement semantics we assume that the publication of a ping message includes also the advertisement. The unsubscription cost depends also on other active subscriptions on the servers and is a worst-case estimate. For subscription semantics the basic problem with the ping-based synchronization protocol is that the ping message is guaranteed not to be subscribed by any other broker on the network and hence the ping subscription message will be introduced at every broker on the network. If uniqueness is not guaranteed simultaneous mobility by different clients becomes impossible. If the client relocates faster than the ping message is propagated it has to wait until the target receives the ping subscription. This requires that the pub/sub AI allows to query subscription status of the broker. The iena support service does not require AI support, because the ping/pong messages are continuously resent [2], but this kind of behaviour further burdens the network. The ping is published by the target along with the pong subscription and they will reach the source, which replies with the pong event. It is clear that when the pong reaches the target the subscription is established between the source and target. Advertisement semantics presents several problems for mobility, because the ping and pong messages need to be advertised before they may be subscribed. The source advertises pong and subscribes ping, whereas the target advertises ping and subscribes pong. This is problematic because the advertisement from source needs first to reach the target before target s subscription starts to propagate on the reverse path. We envisage that this kind of oscillation may be avoided by using virtual advertisements for the ping/pong messages so that they are propagated using subscription semantics. Another solution is to bundle the advertisement and notification into one, which we have proposed previously [5]. The pub/sub AI needs to support either solution. ublisher mobility for subscription semantics does not require any changes, but it is supported directly by the pub/sub system. ublisher mobility for advertisement semantics follows the ping/pong model and has a similar cost structure.

3 Table 1. Cost structure for generic mobility hase ub semantics Adv semantics ource: ubsc. ing(id) N - Target: ubsc. Filter 0 x N 0 x N Target: ubsc. ong(id) N - Target: ub. ing(id) n N ource: ub. ong(id) n N ource: Unsub. ing(id) N n Target: Unsub. ong(id) N n Target: Unadv. ing(id) - N ource: Unadv. ong(id) - N 3.2 Mobile ubscribers Figure 1 presents an overview of the subscription handover both for complete and incomplete topologies. rior to the handover a set of brokers have advertised the events that are being relocated. Handover is performed between two brokers in the network: a, b V. The handover starts when b receives a message, typically from the client, that a set of subscriptions should be relocated from a to b. Now, b may optimize the operation if has been already subscribed by b. In this case b simply starts to buffer notifications for and retrieves the notifications buffered at a for the client. If the optimization cannot be applied, b issues a subscription message, which is propagated by the event system. The subscription message must include those parts that are not covered by existing subscriptions. 3 Acyclic Graphs with Advertisements Complete ubscription Topology Incomplete ubscription Topology 3.1 Overview a 3. M 2. a 0. ub 3. M 2. An outline of the protocol we describe was presented in [3, 7] with mobility restricted between border brokers. A mobility protocol for a hierarchical routing topology with more assumptions was examined in [6]. Formally, the network of application-level event brokers is an undirected acyclic graph G = (E, V ), where E is the set of edges and V is the set of vertices of the graph. We are interested in finding the theoretical cost for the handover protocol in terms of message exchanges. ome systems allow mobility only between border-brokers, the leaves of the routing tree, which limits mobility. We allow clients to roam between any two brokers. In subsequent examination we assume that the servers use reliable communication mechanism, G is connected, and messages are delivered in order. We assume that clients know the subscriptions and the address of the origin server (for direct communication). ome handover protocols use pub/sub routing to find the origin server; however, this approach may require the use of flooding to find the origin server, for example in scenarios where the relocated subscriptions are not advertised. We also assume that buffered events are delivered outside the event routing framework, since they have already been routed and it is not efficient to reroute them. Also, by transferring the notifications in one compressed message the overhead may be reduced considerably. In some existing systems buffered events are delivered using the event routing system. While this approach adheres to the pub/sub communication, it has been shown that this cost dominates the pub/sub signalling cost [6] in mobility scenarios. We consider the number of exchanged messages, because update processing and propagation causes the most stress for the pub/sub system b ub 1. Figure 1. ubscription handover with a complete and incompletecomplete topology. Advertisement Colored nodes Incomplete represent Advertisement the Topology inactive Topology path. 3. a M a ub In the figure the complete subscription ub topology update proceeds as follows: 1. b sends the update message, update message is propagated b towards a and will b meet a s subscriptions at node M (which can also be a ub itself), 3. M contacts a if applicable, 4. session is transferred and a may unsubscribe to M if necessary. For the incomplete case the procedure is similar, but the subscription update message may be sent to several directions as depicted in the 1. phase of the figure. In 2. subscriptions from a and b meet and the meeting point, M, notifies a in 3. It is clear that all subscriptions contained in must be covered at M. We view a path as consisting of the edges on the path, since that makes it more convenient to handle split paths. An edge between a and b is called active with respect to if has been sent on that edge. A path is active if all of its edges are active. An inactive path is one that is not active, i.e. contains at least one inactive edge. We know that a and b are connected since G is connected. Also the path between a and b is unique, since G is acyclic. If the covering optimization is not applicable we know that is not active on all edges on the path from a to b. Let a be the set of active edges on the path, and i the 4. b ub

4 set of inactive edges on the path. roposition 1 states that active and inactive paths meet at node M. The topology update cost is the cost of updating i. roposition 1 There exists a node M on the path from a to b such that the path from a to M consists of exactly the edges in a (and then i is the path from b to M). Given that there are simultaneous unadvertisements roposition 1 is not necessarily valid. Consider a scenario in which during the handover of a subscriber the subscribed advertisement is removed from the system. In this case, M may not exist or may disappear during the handover. The propagation of the update message from b will reach neither a nor M. The protocol must be able to cope with this and signal to a that the update is complete. imultaneous advertisements do not cause such a problem. It also follows for complete topologies that if an update is required it is sent by b to exactly one output interface. If this were not the case there would be a publisher on an interface not connected with the subscriptions from a, which would violate the assumption that the topology is complete. When reaches M the subscription has taken effect. After this a may unsubscribe if necessary. M can do that for a, but since the a s state may have changed after the update was sent by b, M contacts a. The problem with covering and merging is that M does not know the exact interface from which a is reachable. Therefore M needs to forward the completion message to all interfaces that have covering subscriptions. In essence, this is content-based flooding. Now, this phase may be optimized in complete topologies by allowing M to directly contact a, which is feasible since the topology has already been updated for the path from a to b. After this a may unsubscribe if necessary Communication Cost Assuming completeness, the total cost for the mobility in terms of exchanged messages for the subgraph defined by the path from a to b is i messages for the path, one message from M to a, the cost of transferring the buffered events, and then the cost of unsubscription C unsub. The worst case size of i is n 1, where n is the number of brokers in the routing topology. Therefore for the complete topology the update cost for subscriptions is at most n + C unsub. If is already covered at b then the cost is 1 + C unsub Handovers on Incomplete Topology It may also be the case that the set of subscriptions,, is still being propagated and the subscription path is not totally active at the time of the handover. This happens when the handover happens while the system propagates and processes subscription messages or update messages. This kind of behaviour may occur because the handover is triggered by out-of-band mechanisms, for example by terminal mobility. If mobility is activated by sending a message using the pub/sub system this problem may be avoided. rotocols that forward the update message to interfaces that have overlapping advertisements work in this scenario. Clearly, a does not know when the topology is complete. Node b may detect the local incompleteness of the subgraph defined by the path from a to b; however, b cannot detect the completeness of the subgraph based on local information. For example, consider the scenario where there is one broker between a and b that has not yet processed the subscription from a and is covered at b. Now, since is covered at b, b will ask a to send the session eventhough the subscriptions may not be complete on the path at the time. This incompleteness of the path means that the handover is not necessarily mobility-safe. This may be solved by synchronizing a and b using a message passed through the path. The update message is needed even when is covered at b and requires at least a i messages. The problem is that b does not know where a is located so the update is flooded on the reverse-path of covering subscriptions. Incompleteness may also cause flooding when is not covered at b and the protocol needs to cope with the existence of multiple meeting points M. 3.3 Mobile ublishers Related work has typically considered only the protocol for relocating subscriptions. ince it is probable that subscriptions are relocated more often this is reasonable. The problem with mobile publishers is that advertisements are propagated throughout the routing network. This means that the removal of an advertisement may have a high cost. We follow a similar approach as with mobile subscribers. The protocol proceeds in four phases: 1. b sends an update message towards a and overlapping subscriptions are sent towards b, 2. Existing advertisements and subscriptions meet the advertisement sent by b and a is notified by M that the topology has been updated. 3. a sends an update message to b to ensure completeness. Finally, 4. a unadvertises if necessary. If we assume that the advertisement topology is complete we find that the set A of advertisements to be relocated from a to b is already advertised to b. The advertisement may be more general due to covering and merging. The problem is that the advertisements point to a, which needs to be changed. We have the two paths, the active path A a, and the inactive path A i and M at their intersection. If there are no publishers that advertise covering advertisements on the path A a A i it follows that A a =. When b receives A, it does not need to propagate A to those interfaces where A has already been forwarded. Those parts of the routing network are already updated. If all interfaces advertise A at b then the handover is complete. If all interfaces do not advertise A at b, they need to be updated. If there are more than one interface that does not advertise a set of advertisements that covers A then the topology is incomplete. (The one is the interface that re-

5 ceived the advertisement). By definition each router advertises a received advertisement to all other interfaces. Now, b sends the update message to all interfaces that do not cover A and advertises A from b. This means that subscriptions on are hop-by-hop connected with b. Here we have the inactive path i towards a, which is the path from b to subscribers that are not yet connected. In order to avoid false negatives a and b must wait that the subscriptions are connected. In order to avoid false negatives, a sends a ping message to b to indicate when the subscriptions on the path have been updated. For complete topologies the ping message can be sent from M, if completeness cannot be assumed it is sent by a. In some cases, subscribers are already connected after the first advertisement update sent by b. ince b does not know if there are other subscriptions that overlap with A later in the path, the whole path from b to a needs to be checked. It is important that the protocol is ended properly in order to ensure that published events are properly disseminated to subscribers, and that the next handover may be started from a complete configuration. If the advertisements are not successfully terminated in G the publisher handover protocol ensures that upon completion of the handover the subgraph defined by the path from a to b is complete for both advertisements and subscriptions. In this case, it may be necessary to use flooding to find M and a Communication Cost The total cost for the mobility in terms of exchanged messages for the subgraph defined by the path from a to b is A i messages for the advertisement update. The subscription update requires at least i for the ping message. One message is needed to notify a that the subscriptions are complete. This happens after b receives the ping message. Finally, we have the cost of unadvertisement C unadv. Assuming completeness this gives the total cost of 1 + A i + i + C unadv. The worst case size for both A i and i is n 1, where n is the number of brokers in the routing topology. Therefore the handover cost for relocating advertisements from a to b is at most 2n 1 + C unadv. If A is already covered at b then the cost is 1 + C unadv. R of an event type is obtained by hashing the event type to the flat addressing space of the overlay. Most current overlays are based on flat addressing spaces. This means that rendezvous points are uniformly distributed over the addressing space [8]. Figure 2 illustrates R-based notification: 1. ublisher advertises an event type with a filter, 2. Advertisement is forwarded to the R, 3. ubscriber subscribes an event of the same type with a filter, 4. The subscription message is not covered at any intermediate broker and forwarded to the R, 5. Another subscriber subscribes, 6. ubscription is forwarded to the R. After step 6. in the type/attribute-based routing supported by Hermes, the R sends the subscriptions on the reverse path of advertisements. Any events conforming to the advertisement from the publisher are sent on the reverse path of subscriptions. The model is the advertisement semantics model with three key differences. First, all subs/advs are sent towards the R. Thus routing topology is constrained by the R. econd, advertisements are introduced only on the path from the advertiser to R. Third, subscriptions are introduced on the path from the subscriber to R and on any nodes with overlapping advertisements. These differences are interesting, because advertisement becomes a local property of a branch of the multicast tree rooted at R. This may modelled using virtual advertisements: a R has virtual advertisements for all events of the event type managed by the R and hence subscriptions are sent towards it. In the following examination we assume that the overlay topology is static a dynamic topology requires more complex investigation. 3. AC AC3 4. B2 R 2. A AC2 1. A Figure 2. Example of the Hermes model 4 Rendezvous-oint Models Hermes is peer-to-peer event system based on an overlay called an that supports a variant of the advertisement semantics. ystems such as Hermes [8] leverage the features of the underlying overlay system for message routing, scalability, and improved fault-tolerance. Hermes supports the basic pub/sub operations introduced previously. Each event flow is rooted at a designated server, the Rendezvous oint (R), and messages are propagated toward that node. The R manages the event type and Hermes supports chaining Rs into type hierarchies. In general, the 4.1 Incompleteness The R may be used to guarantee completeness of advertisements and subscriptions by requiring an acknowledgement from the R. We propose to solve the problems posed by incomplete topologies using two mechanisms: R completeness checking both at the origin and destination of mobility.

6 reventing content-based flooding using the overlayaddress, which allows to determine in which direction a node is located. This property may also be used to prevent the content-based flooding at M when forwarding the mobility update messages within the event topology. The problems of incompleteness with the regular sub/adv semantics can be avoided with the R model by ensuring that each subscription and advertisement is complete to R. This incurs additional cost, but the number of acknowledgement messages may be minimized by pushing the acknowledgement generation away from the R to nodes that have covering subscriptions that are complete to R. The two central problems that are solved are: false negatives due to incompleteness of the path and a ping between source and destination is not needed, and by using the overlay address to locate the source content-based flooding does not need to be used. 4.2 Communication Cost The R completeness simplifies subscriber and producer mobility protocols and the communication costs. Within one branch the update message from b to a eventually meets the subscription from a to R. Between branches the cost is the distance from b to the R. Hence, the worst-case cost for topology update if R completeness is assumed is given by R max, where R max is the maximum number of nodes between any node and a R. One message is needed to notify a and one is needed to transfer buffered events. The case when a relocated subscription is covered is handled as before. roducer mobility is also simpler than before and for a new branch it is sufficient to synchronize b with the R or Rs in question. Although Rs are good for mobility, the cost of a handover grows with the number of R updates for cyclic overlay routing. Rendezvous points models with acyclic overlay routing have the simplifying features mentioned above and the upper bound cost cannot be greater than for the general acyclic graphs. 5 Conclusions In this paper, we examined the cost of pub/sub mobility using three mobility mechanisms and topologies: generic mobility support, acyclic graphs, and rendezvous-based topologies. We also discussed the impact of completeness and incompleteness of the pub/sub topology on the cost of mobility. The generic mechanism has a high cost for mobility. The other two mobility mechanisms have considerably smaller cost. If an acyclic graph-based routing topology is incomplete content-based flooding may or must be used to complete the handover succesfully. This means that the optimizations discussed for complete topologies may not be applied for incomplete topologies. ince it is not possible to detect completeness and many systems are inherently incomplete, the optimizations that avoid content-based flooding may not be applied in practise. Mobility-safety cannot be guaranteed if protocols engineered with the completeness assumption are used for incomplete topologies. Rendezvous point - based topologies were identified to be good for pub/sub mobility support, because they may be extended to guarantee the completeness of subscriptions and advertisements. In addition, the R models with advertisements do not require the flooding of the whole network with advertisement messages. References [1]. Eugster,. Felber, R. Guerraoui, and A. Kermarrec, The many faces of publish/subscribe, ACM Computing urveys, (2): , [2] M. Caporuscio, A. Carzaniga, and A. L. Wolf, An experience in evaluating publish/subscribe services in a wireles network, In Third International Workshop on oftware and erformance, Rome, Italy, July [3] A. Zeidler and L. Fiege, Mobility support with Rebeca, In roceedings of the ICDC Workshop on Mobile Computing Middleware, [4] G. Mühl, Large-scale content-based publish/subscribe systems, h.d. thesis, University of Darmstadt, [5]. Tarkoma, J. Kangasharju, and K. Raatikainen, Client mobility in rendezvous-notify, In Intl. Workshop on Distributed Event-Based ystems (DEB 03), [6] I. Burcea, H-A. Jacobsen, E. de Lara, V. Muthusamy, and M. etrovic, Disconnected operation in publish/subscribe middleware, In IEEE International Conference on Mobile Data Management, [7] G. Mühl, A. Ulbrich, K. Herrmann, and T. Weis, Disseminating information to mobile clients using publish/subscribe, IEEE Internet Computing, pages 46 53, May [8]. ietzuch, Hermes: A calable Event-Based Middleware, h.d. thesis, Computer Laboratory, Queens College, University of Cambridge, February 2004.