tweaks, turnips, and other things.
parent
a8289de6c4
commit
da958761f5
|
|
@ -68,6 +68,8 @@ Lukasz Ziarek (Univ. of Buffalo, Dept. of Comp. Sci. and Eng.)}
|
|||
\input{sections/5-priorresults}
|
||||
|
||||
\pagebreak
|
||||
\setcounter{page}{1}
|
||||
|
||||
{
|
||||
\bibliographystyle{nsf}
|
||||
\bibliography{main,geoffreychallen}
|
||||
|
|
|
|||
|
|
@ -7,7 +7,7 @@
|
|||
The world's 2~billion smartphones and 4~million apps have become a large part
|
||||
of most people's computing experiences.
|
||||
%
|
||||
A common requirement of apps is persisting structured data, a task frequently
|
||||
Most apps need to persist structured data, a task frequently
|
||||
performed using an \textit{embedded database} such as SQLite.
|
||||
%
|
||||
These are heavily used, with Android smartphones generating an average of
|
||||
|
|
@ -38,7 +38,7 @@ understood.
|
|||
|
||||
%%%%%%%%%%%%%%%%
|
||||
|
||||
To date, there have been some initial explorations of small-scale data
|
||||
To date, there have been some initial explorations of small-, personal-, or pocket-scale data
|
||||
management, both in academia and by industry:
|
||||
%
|
||||
\begin{itemize}
|
||||
|
|
@ -63,37 +63,36 @@ Unfortunately, unlike the largely homogeneous workloads and platforms that
|
|||
common to research on classical monolithic enterprise databases, \PocketData{}
|
||||
is far more diverse.
|
||||
%
|
||||
Data access patterns vary wildly by user, time of day, mix of installed apps,
|
||||
Data access patterns are extremely bursty and can vary wildly by user, time of day, mix of installed apps,
|
||||
network accessibility, and many other factors.
|
||||
%
|
||||
Platform properties such as RAM, persistent storage, CPU performance, and network
|
||||
bandwidth also exhibit extreme variations, sometimes by multiple orders of magnitude.
|
||||
bandwidth also exhibit extreme variations across phones, sometimes by multiple orders of magnitude.
|
||||
%
|
||||
Resource availability can also vary; Some users keep their phones constantly
|
||||
charged, while others go multiple days without charging.
|
||||
|
||||
%%%%%%%%%%%%%%%%
|
||||
|
||||
The heterogeneity of the \PocketData{} setting, make it challenging for researchers to
|
||||
The heterogeneity of the \PocketData{} setting makes it challenging for researchers to
|
||||
understand the tradeoffs and requirements of the setting.
|
||||
%
|
||||
This lack of clear high-level goals, in turn, makes it difficult to clearly identify successful
|
||||
research contributions and creates a daunting environment for new research efforts.
|
||||
%
|
||||
Lacking the resources necessary to better understand and adapt to \PocketData{} scale,
|
||||
Lacking the resources necessary to better understand and adapt to the \PocketData{} scale,
|
||||
research efforts in the area are presently limited.
|
||||
|
||||
%%%%%%%%%%%%%%%%
|
||||
|
||||
\textbf{Target Community}
|
||||
Research on mobile devices and the internet of things is cross cutting, intersecting communities
|
||||
Research on mobile devices and the more general space of the internet of things (IoT) is cross cutting, intersecting communities
|
||||
that work on data management systems, real-time and embedded devices, programming languages,
|
||||
and operating and mobile systems. We believe research involving \PocketData{} also lies at
|
||||
the intersection of these communities. Specialized databases systems for embedded devices is
|
||||
a growing topic in the database community. As embedded processors become more capable, with
|
||||
larger amounts of main memory available (e.g. Intel's Edison platform), there is a growing push from the embedded
|
||||
and also from the real-time communities to explore larger software capabilities, including database
|
||||
systems and query processing systems, in embedded deployments. The programming language
|
||||
the intersection of these communities. Specialized database systems for embedded devices are re-emerging as an interesting topic in the database community. As embedded processors become more capable, with
|
||||
larger amounts of main memory available (e.g. Intel's Edison platform), there is a growing push from the embedded systems
|
||||
and the real-time communities to explore larger software capabilities, including database
|
||||
systems and query processing systems in embedded hardware deployments. The programming language
|
||||
community is exploring domain specific languages for specialized query processing.
|
||||
The mobile community is continually exploring how to push the envelope on smartphone based computing,
|
||||
whether via power aware mechanisms, or through more adaptive systems. Many of these solutions use mobile databases
|
||||
|
|
@ -118,19 +117,18 @@ During this planning grant we will focus our efforts in three key areas:
|
|||
\begin{enumerate}
|
||||
|
||||
\item \textbf{Growth of the Mobile Embedded Database community}: We have established an
|
||||
initial community of interested CISE researchers for \PocketData{} consisting of both
|
||||
academic and industrial researchers. We believe that this community
|
||||
initial community of interested CISE researchers for \PocketData{} from both
|
||||
academia and industry. We believe that this community
|
||||
shows that there is sufficient interest in CISE to pursue our proposed \PocketData{} infrastructure.
|
||||
However, for long term success we would like to expand this community to ensure that the infrastructure
|
||||
meets the needs of the broader community and not just niche.
|
||||
meets the needs of the broader community and not just a specific research niche.
|
||||
|
||||
\item \textbf{Expansion to IoT}: Our current efforts have focused primarily exploring questions
|
||||
relating to \PocketData{} in the mobile domain, specifically Android. Although a \PocketData{}
|
||||
infrastructure based solely in Android is valuable, we believe a more comprehensive infrastructure
|
||||
must take into account recent developments in IoT. There are similarities between how mobile
|
||||
applications leverage embedded databases and how proposed IoT application would use embedded
|
||||
databases, specifically in the area of personal health care devices and smart city deployments where
|
||||
small devices process data before sending \emph{relevant} data for more centralized big data processing.
|
||||
applications leverage embedded databases and how proposed IoT applications would use embedded
|
||||
databases, specifically in the areas of personal health care devices that aggregate and summarize a user's personal data and smart city deployments where small devices process data before sending \emph{relevant} data for more centralized big data processing.
|
||||
We propose to expand and modify our \PocketData{} infrastructure to meet the needs of IoT community.
|
||||
|
||||
\item \textbf{Workshops and Tutorials}:
|
||||
|
|
@ -140,14 +138,14 @@ major conferences in the database, systems, real-time systems, and programming l
|
|||
communities.
|
||||
Our budget includes funding for travel to such conferences to host workshops and
|
||||
tutorials. This will also enable the PIs to receive valuable feedback on the needs of the community in
|
||||
structuring the \PocketData{} infrastructure.
|
||||
designing and building out the \PocketData{} infrastructure.
|
||||
|
||||
|
||||
\end{enumerate}
|
||||
|
||||
|
||||
A successful planning grant will enable us to proceed with the development of a full
|
||||
\PocketData{} infrastructure.
|
||||
\PocketData{} infrastructure proposal.
|
||||
Concretely, the following three resources will be developed as part of the full infrastructure
|
||||
proposal:
|
||||
\begin{enumerate}
|
||||
|
|
@ -165,22 +163,18 @@ summarizing those datasets.
|
|||
\item \textbf{Standards and Benchmarks}:
|
||||
We will create a toolkit to establish a set of standards for evaluating research efforts on
|
||||
\PocketData{} for both Android and IoT.
|
||||
First, significant parts of the Android platform have been locked down for reasons of
|
||||
security and intellectual property, making properties like process scheduling and power usage
|
||||
difficult to measure reliably.
|
||||
The \PocketData{} setting requires unique metrics that can be difficult to reliably measure and attribute on the Android platform.
|
||||
The toolkit will include instrumentation for Android that will make it easier for researchers
|
||||
to measure the performance of their \PocketData{} tools.
|
||||
to measure the performance through as-yet-uncommon measures like availability of idle time, thread scheduling, power consumption, and other metrics that can be hard to measure reliably on the Android platform like CPU and memory usage for specific libraries.
|
||||
Second, to standardize comparisons across different research efforts, the toolkit will
|
||||
include a benchmark suite.
|
||||
This benchmark will create clearly defined metrics for evaluating success. Moreover, by
|
||||
making it extensible, the benchmark will act a clearinghouse for app behaviors discovered
|
||||
in the wild and changing database requirements.
|
||||
\item \textbf{Visualization}: We will create a data visualization tool and associated queries
|
||||
to help researchers understand and navigate the data. The raw traces gather are very
|
||||
larger and the bulk of the data may not be useful for answering a specific question a
|
||||
given researcher may which to explore. Through visualization, filtering, data navigation,
|
||||
and specialized queries, we will enable researchers to more quickly and accurately explore
|
||||
relevant characteristics of \PocketData{}.
|
||||
to help researchers understand and navigate the data. The raw traces we plan to offer researchers are very
|
||||
large and the rich structure and variability of SQL queries generated by smartphone apps do not admit traditional indexing strategies for analytics. By providing database-driven tools that aid in the analysis and visualization of the resulting queries, we will enable researchers to more quickly and accurately explore
|
||||
relevant characteristics of real-world \PocketData{} workloads.
|
||||
|
||||
\end{enumerate}
|
||||
|
||||
|
|
@ -190,7 +184,8 @@ expertise from three of the main communities our proposed \PocketData{} infrast
|
|||
All three PIs already have a record of successful collaboration in the past~\cite{pocketdata,Challen:2015:MWE:2699343.2699361}, and PIs Kennedy and Ziarek have been working together for the last three and a half years on adaptive indexing~\cite{techreport,agarwal2013monadic,kennedy2015just}.
|
||||
PI Kennedy's expertise covers databases, incremental computation~\cite{Ahmad:2012:DHD:2336664.2336670,kennedy2011dbtoaster,koch2013dbtoaster}, uncertain data management~\cite{Kennedy:2011:JEO:1989323.1989410,5447879,Yang:2015:LOA:2824032.2824055}, online aggregation~\cite{4812533,Kennedy:2011:FPP:1989323.1989482}, and compiler design~\cite{kennedy2011dbtoaster,Ahmad:2012:DHD:2336664.2336670,koch2013dbtoaster}.
|
||||
PI Ziarek's expertise covers programming languages~\cite{Ziarek:2011:CAE:1993498.1993572,Ziarek:2010:LCC:1852977.1852979}, real-time systems~\cite{Blanton:2013:NIC:2512989.2512994,Yan:2013:RDR:2512989.2512990}, virtual machines~\cite{Pizlo:2010:HPE:1755913.1755922,Pizlo:2010:SFR:1806596.1806615}, and compiler design~\cite{Sivaramakrishnan:2012:ERB:2258996.2259005,Ziarek:2008:FTS:1466762.1466777}.
|
||||
PI Challen's expertise covers \todo{Geoff... a quick blurb}.
|
||||
%Geoff... validate this if you could.
|
||||
PI Challen's expertise covers systems and networking~\cite{hotwireless2015-sharing,infocom2016-scans,hotnets2014-pocketsniffer}, mobile systems~\cite{iiswc2015-agility}, and smartphones~\cite{mobicase2015-jouler,hotmobile2015-numerator,mobicase2014-pocketmocker}
|
||||
|
||||
\subsection{Datasets}
|
||||
\input{sections/1-1-metrics.tex}
|
||||
|
|
|
|||
Binary file not shown.
|
|
@ -27,7 +27,7 @@ Computer Science and Engineering (CSE) data storage facilities include vulcan, a
|
|||
|
||||
CSE faculty compute systems include castor, a Sun Blade 1000; citrix[1-3], a load-balanced Citrix farm of Dell PowerEdge 2650 servers; the-who, a Sun Fire V20z desktop virtualization server; benatar, a virtualized general compute server; and the underground cluster, a 4-node compute cluster comprised of Dell 1425s. CSE Faculty also have use of all CSE student systems (below).
|
||||
|
||||
CSE student compute systems include timberlake, a Dell PowerEdge R600 compute server; metallica, a Dell PowerEdge R500 compute server; pollux, a Sun Sparc enterprise T5220 compute server; coldplay, a Sun Fire V20z compute server; fork, a Sun Fire V20z dedicated to the Operating Systems course; nickelback, a Dell PowerEdge 1950 desktop virtualization server; dragonforce, a Dell PowerEdge R720 desktop virtualization server; styx, a Dell PowerEdge R400 desktop virtualization server.
|
||||
CSE student compute systems include a Dell PowerEdge R600 compute server; a Dell PowerEdge R500 compute server; a Sun Sparc enterprise T5220 compute server; a Sun Fire V20z compute server; a Sun Fire V20z dedicated to the Operating Systems course; a Dell PowerEdge 1950 desktop virtualization server; a Dell PowerEdge R720 desktop virtualization server; and a Dell PowerEdge R400 desktop virtualization server.
|
||||
|
||||
CSE research groups occupy 6628 square feet of research lab space ranging from secure, monitored, temperature-controlled data centers to specialized experimental facilities. CSE instructional labs occupy 4096 square feet, each configured to serve the characteristic needs of the courses they host. The Patricia Eberlein is the CSE general student computing lab which occupies 1056 square feet.
|
||||
|
||||
|
|
|
|||
Loading…
Reference in New Issue