Database Reference
In-Depth Information
Protocol Library.
Smile offers a protocol library providing the scheduler de-
veloper with pre-cooked scheduling queries (e.g., for SS2PL). These scheduling
queries can be used out of the box or as a starting point to develop domain-
specific protocols. We expect developers to extend this library over time with
their own protocol modules.
Runtime Statistics Collection.
We let Smile gather statistics about the be-
haviour of its operations at runtime such as the cardinalities of
and
the execution times of the scheduling queries. In future work, we plan to expose
this information to the scheduler developer for the use in the scheduling queries
and let her provide policies for scheduling the execution of the Smile threads.
We developed strategies deciding when to pause a thread ensuring an e
cient
resource usage. E.g., running the Executor while
R
,
H
and
E
E
is empty wastes resources.
2.3 Example: Use Case Implementation
We sketch the protocol implementation of the use case to illustrate the simplicity
and conciseness of our approach. Scheduling queries are given as domain rela-
tional calculus (DRC) expressions. A simplified DRC formulation of
Q
Schedule
implementing the use case constraints is:
Q
Schedule
=
{S, C | is
2
PL
(
S, C
)
∧
∃C
2
(
C
2
=
premium
∧
(
is
2
PL
(
,C
2
)
⇒ C
=
premium
))
}
We use a declarative implementation of SS2PL to realize constraint C1. Predicate
is
2
PL
(
S, C
)uses
to determine all requests
S
with their customer class
C
that
can be executed without violating the SS2PL constraints that have to hold for
the generated schedule (requests in relation
R
). We use
S
as a shorthand for
the request related attributes of
is
2
PL
(transaction ID etc.). Using Oshiya, we
can implement scheduling constraint C2 as follows: If there exists at least one
request of a premium customer (
C
2
=
premium
∧ is
2
PL
(
,C
2
)), then only
premium requests are selected by
Q
Schedule
(
C
=
premium
).
H
References
1. Alvaro, P., Condie, T., Conway, N., Elmeleegy, K., Hellerstein, J.M., Sears, R.:
Boom Analytics: Exploring Data-Centric, Declarative Programming for the Cloud.
In: EuroSys, pp. 223-236 (2010)
2. Cooper, B.F., Ramakrishnan, R., Srivastava, U., Silberstein, A., Bohannon, P.,
Jacobsen, H.-A., Puz, N., Weaver, D., Yerneni, R.: PNUTS: Yahoo!'s Hosted Data
Serving Platform. PVLDB 1(2), 1277-1288 (2008)
3. Tilgner, C.: Declarative Scheduling in Highly Scalable Systems. In: EDBT/ICDT
Workshops, pp. 41:1-41:6 (2010)
4. Tilgner, C., Glavic, B., Boehlen, M., Kanne, C.-C.: Correctness Proof of the Declar-
ative SS2PL Protocol Implementation. Technical Report IFI-2010.0008, University
of Zurich, Department of Informatics (2010)
5. Vogels, W.: Data Access Patterns in The Amazon.com Technology Platform. In:
VLDB, vol. 1 (2007)
