Information Technology Reference
In-Depth Information
paper, we focus on faults that propagate from participants to composite services. We
refer to these faults as
bottom-up
. Two examples of bottom-up faults are (1) a shut-
down scheduled by a participant's provider for maintenance; and (2) an update to a
participant's policy (e.g., new message parameters added to a WSDL specification)
that may affect the way that participant is consumed. Hence, composite services must
rapidly detect and handle faults in their participants to avoid run-time failures and
maintain consistency.
Web services generally use HTTP as the underlying message transport. Hence,
they are either guaranteed message delivery or notified if a message was not delivered
(e.g., because of a server unavailability). In the latter case, composite services be-
come aware of a fault only at the time they interact with their participants
not
at the
time that fault occurred. This may decrease the availability of composite services.
Besides, users' requests are pending as long as the composite service did not recover
from the fault (e.g., by replacing the faulty participant with an equivalent one). This
calls for a framework in which composite services are able to detect and handle bot-
tom-up faults as soon as those faults occur in their participants.
In this paper, we introduce a framework for managing bottom-up faults in com-
posite services. The proposed framework uses
soft-state
signaling to propagate faults
from participants to composite services. Soft state denotes a type of protocols where
state (e.g., whether a server is alive) is constantly refreshed by periodic messages;
state which is not refreshed in time expires [8]. This is in contrast to hard-state where
installed state remains installed unless explicitly removed by the receipt of a state-
teardown message. Advantages of the soft-state approach include implicit error re-
covery and easier fault management, resulting in high availability [16]. Soft state was
introduced in the late 1980s and has been widely used in various Internet protocols
(e.g., RSVP). However, to the best of our knowledge, this work is the first to use
soft-state for fault management in composite services. The major contributions of this
paper are summarized below:
•
We introduce a bottom-up fault model for composite services. The model in-
cludes a taxonomy of bottom-up faults, a definition of (composite) service, and
peer-peer topology for fault management.
•
We propose a soft-state protocol for bottom-up fault propagation.
•
We conduct experiments to assess the performance of the proposed framework.
The rest of this paper is organized as follows. In Section 2, we describe the bottom-
up fault model. In Section 3, we propose the soft-state protocol for bottom-up fault
propagation. In Section 4, we present experiments to assess the performance of the
proposed approach. In Section 5, we give a brief survey of related work. We finally
provide concluding remarks in Section 6.
2 Fault Model
In this section, we describe our model for bottom-up fault management. We first
provide a categorization of bottom-up faults. Then, we define the notion of partici-
pant's state. Finally, we introduce a peer-to-peer topology for bottom-up fault
management.
