Tài liệu miễn phí Tổ chức sự kiện
Download Tài liệu học tập miễn phí Tổ chức sự kiện
If a node calculates that it does not have enough storage
capacity for the table, it initiates the group formation algo-
rithm. To minimize the number of times an original tuple
must be transmitted to make it available to every member
of a group, we require that all nodes in the group are within
broadcast range of each other. A second required property
of a group is that it must have enough cumulative storage
capacity to accommodate the table of predicates. If these
requirements can not be met, the join classification (see
Section 3.2) is not intermediate but rather...
8/30/2018 2:40:28 AM +00:00
The group formation algorithm deals with message loss
by allowing every state in the finite state machine to time
out while having a minimal effect on other nodes. For ex-
ample, if a master node does not hear back from enough
neighbors, it will time out (shown as TO in Figure 2) and
transition back into the Need Group state. Nodes that had
responded to the master cannot respond to any other master
until they hear back from the current one. If they never hear
back, they time out and go back to the Need Group state.
The algorithm adds some...
8/30/2018 2:40:28 AM +00:00
For situations in which there are a very large number of
tuples in the join table, we can just disseminate information
that allows sensors to identify tuples that definitely do not
join with any predicates. Suppose we know that there are
no predicates on attribute a in the range a1 … a2. If we
transmit this range into the network, then a sensor tuple, t,
with value t.a inside a1 … a2 is guaranteed to not join with
any predicates and need not be...
8/30/2018 2:40:28 AM +00:00
The cache diffusion algorithm then works as follows.
Every time the root basestation receives a tuple that does
not join, it sends the maximal ERD which that tuple inter-
sects one hop in the direction that the tuple came from.
This node then checks its local value cache for tuples that
are contained within this ERD. If one is found, this value
and any other values that overlap with the ERD are re-
moved from the local value cache, and the ERD is added to
the ERD cache table with priority 1. If no match is found,
then the ERD...
8/30/2018 2:40:28 AM +00:00
Thus, for a node x of depth d, it takes d tuples that fall
within an ERD to be produced before the ERD reaches
node x. Note that these d tuple productions do not have to
be consecutive as long as the matching ERD that diffuses
to node x does not get removed from the ERD cache of its
ancestor nodes on its way. Further, note that despite the
fact that it takes d tuples before node x receives the ERD,
these tuples get forwarded fewer and fewer times while the
ERD gets closer and closer to x. In...
8/30/2018 2:40:28 AM +00:00
RAM is consumed by TinyDB and forwarding buffers in
the networking stack. We have also experimented with
several other types of join queries and found similar re-
sults: irrespective of the query, join-predicate selectivity
and average node depth have the largest effect on query
execution cost for the distributed join algorithm.
For all graphs showing results for the distributed join al-
gorithm, we show power utilization and result accuracy at
steady state, after groups have formed and nodes are per-
forming the join in-network. We do not include table dis-
tribution costs in the total transmission numbers. We
choose to do...
8/30/2018 2:40:28 AM +00:00
A widely cited application of sensor networks is event-
detection, where a large network of nodes is used to iden-
tify regions or resources that are experiencing some phe-
nomenon of particular concern to the user. Examples in-
clude condition-based maintenance in industrial plants
[14], where engineers are concerned with identifying ma-
chines or processes that are in need of repair or adjustment,
process compliance in food and drug manufacturing [25],
where strict regulatory requirements require companies to
certify that their products did not exceed certain environ-
mental parameters during processing, and applications
centered around homeland security, where shippers...
8/30/2018 2:40:28 AM +00:00
These methods can involve a social policy for resource
distribution. A policy is simply a set of rules for allo-
cation when resource demand exceeds resource supply.
One candidate policy is to seek efficient usage, which di-
rects amechanismto allocate resources to the set of users
who have the highest utility for the use of the resources.
Other social policies exist, such as those that favor small
experiments, or favor underrepresented stakeholders, or
(if money is involved) seek maximal revenue generation.
One can also implement a mixture of policies to meet a
complex social goal.
Past deployments of distributed system schedulers
(e.g. Condor [12]) focused on maximizing utilization,
and were not designed to...
8/30/2018 2:40:28 AM +00:00
Deploying a computationalmarket for resource alloca-
tion in the systems domain can benefit two research con-
stituencies. The first constituency, which will be ignored
for the rest of this paper, are the experimental economists
and economically-minded computer scientists. Rarely
are economists actually given the opportunity to deploy a
market or a whole economy, let alone several for compar-
ison. Computationalmechanismdesign [13] is an emerg-
ing topic partly because the results apply to many differ-
ent domains, and there is some merit in asking systems
researchers to be research subjects as they attempt to use
some market mechanism for their own work.
But systems researchers (the second constituency) are
much more interested in knowing...
8/30/2018 2:40:28 AM +00:00
Provide a vocabulary to describe complex resource
bundles. In any system, be it administrative or market-
based, users need a mechanism to express their resource
holdings and desires. Markets, which have been used for
decades to capture difficult resource allocation problems
(e.g. energy markets, wireless spectrum auctions, airline
landing slot exchanges), can also be used to capture the
intricacies of systems problems. Bidding languages have
been studied for their tradeoffs between expressivity and
compactness [15], and existing languages can be directly
applied to computer resources....
8/30/2018 2:40:28 AM +00:00
Link Cross-Testbed experimentation. Multiple
closed distributed systems that run in parallel can offer
unique resources such as access to specific scientific
equipment. One can imagine a physics researcher will-
ing to provide access to their Beowulf cluster [16] but
wishing to consume resources produced by data collec-
tors at a CERN [17] on a completely separate network.
Linked market-based mechanisms could be used to
quantify the value of the cluster time sold in one network
and the value of a CERN resource purchased in another
network in a manner similar to how real economies are
linked through a a currency exchange. Ongoing research
into exchange mechanisms for computational systems
could make this vision...
8/30/2018 2:40:28 AM +00:00
Pressing demand. Past market-based systems never
saw real field testing, and contention was often artifi-
cially generated. Today, a deployed market system could
have immediate usage and solve real resource conflicts.
Real usage data will help researchers calibrate and eval-
uate their market-based resource schedulers. Previous
mechanism designs were not able to take advantage of
user feedback to drive the mechanism design process.
Improved operating system infrastructure. Past sys-
tems had to deal with limitations in infrastructure, such
as a lack of user authentication or kernel-supported re-
source isolation. Today, systems research has produced
tools like BSD Jails, Xen, and Linux CKRM [21, 22],
which are already in use to provide resource isolation,
can...
8/30/2018 2:40:28 AM +00:00
Expressive market design. Previous work used bid-
ding languages that have been artificially limited in their
expressive power. During the past decade, tremendous
advances have been made in the theory and practice of
expressive market design. Current mechanisms can sup-
port combinatorial bidding, which more naturally cap-
tures resource needs. For instance, modern bidding lan-
guages can easily represent any logical combination of
goods, such as AND, OR, XOR, and CHOOSE. This ex-
pressive power did not exist in previous mechanism de-
ployments.
Scalable mechanisms. Solving large resource con-
tention problems has traditionally been computationally
expensive. Fortunately, significant advances have been
made in the theory of solving large-scale mixed-integer
optimization problems, which is an underlying...
8/30/2018 2:40:28 AM +00:00
The idea of using markets and pricing computer re-
sources is quite old. Pricing policies received consid-
erable attention at the dawn of modern multi-user time
sharing systems. Papers in the late 1960’s were dedi-
cated to automated pricing policies for computer time
[18, 19, 20]. As research, this work was short-lived.
The complexity of these schemes relative to their benefit,
combined with the environment of time-shared systems
(mostly cooperative, mostly controlled by a single en-
tity) quickly made pricing for shared resource allocation
a low priority. Shared resource allocation remained a hot
topic in operating systems, but the goal in this research
was maximizing utilization through clever scheduling. In
contrast, schedulers that...
8/30/2018 2:40:28 AM +00:00
In this section, we articulate the roadblocks that must
be addressed to make a market/systems integration suc-
cessful. In our opinion, these challenges are not in the
market details. Rather, we think that the biggest chal-
lenges to their adoption in systems will come fromunder-
standing, supporting, and using these mechanisms. After
presenting each challenge, we consider action items for
the general systems community, as well as for systems
market designers where appropriate. In our view, a mar-
kets/systems integration could fail if these challenges are
not overcome:
Allocation Policy Must be Explicit. One of the un-
comfortable realities of a market is that it forces user
communities to confront their social allocation...
8/30/2018 2:40:28 AM +00:00
Predicting Needs as a Buyer. It is difficult to describe
precisely the level of resources required to run an ex-
periment or job. Depending on the inputs to a program,
the ideal level of resource consumption can vary dramat-
ically.
Moreover, there is a tangible penalty formisestimating
resource need, since these bids are made in advance of
when the resources will actually be available. In order to
match enough buyers with sellers, current market-based
resource allocation schemes batch allocations into blocks
of time. The time scale of this batch system can be min-
utes or days ahead of when the resources will actually
be made available. This means that users must predict
their...
8/30/2018 2:40:28 AM +00:00
Community Action Items: The general systems com-
munity should think more about building tools to help
users estimate their resource needs. Perhaps users in
a shared environment will have access to a best-effort
staging ground where they will be able to gauge their re-
source usage. One can imagine future research tools (ei-
ther modeling or analysis) that attempt to capture the re-
source profile of a wide-area application. Such tools are
an open area for ongoing and future research [10]. Sys-
tems Market Designer Action Items: While there is on-
going research into online market mechanisms—making
an allocation decision before seeing all bid activity—
designers should develop markets that are less...
8/30/2018 2:40:28 AM +00:00
Well-Defined Currency. Almost all previously de-
ployed computational markets have used a virtual cur-
rencies instead of real cash. The low barrier to utiliza-
tion and low stakes in case of deployment error make
simple closed virtual currencies attractive to developers.
In these scenarios, it is all too easy to skip the monetary
policy considerations that make currencies work.
For all of their bootstrapping advantages, virtual cur-
rencies require initial thought and ongoing care to func-
tion properly. Virtual currencies often suffer from a lack
of liquidity, making it difficult to convert into or out of
the virtual currency. As a result, these ersatz currencies
are quite limited; certain usersmight be willing...
8/30/2018 2:40:28 AM +00:00
Calculating and Expressing Valuation. It can be dif-
ficult for a user to accurately value their ideal resource
bundles. There needs to be a simple and effective way
for people to express their resource need and calculate
its value. To stress this point, imagine a market inter-
face that asked the user for their valuation, one ques-
tion at a time, over the entire space of good combina-
tions. This painful approach would require the user to
think about their valuation for a whole slew of bundles,
a time-consuming and sometimes difficult task. An area
of market design that has received almost no attention
for computer resources is in the user...
8/30/2018 2:40:28 AM +00:00
In this paper, we present a distributed location discovery frame-
work, called Sextant, that extracts geographic constraints fromalready-
present wireless radios and uses these constraints to infer node and
event location with high accuracy. Sextant operates by setting up
a system of relative geographic constraints among the network par-
ticipants based on network connectivity and solving this system in
a distributed and efficient manner with the aid of absolute position
information provided by a small number of landmarks. A landmark
is a node whose absolute position is known; Sextant landmarks can
be cheap static nodes whose positions are fixed, or they may be mo-
bile nodes equipped with dedicated hardware,...
8/30/2018 2:40:28 AM +00:00
The rest of the paper is structured as follows. The next section
discusses related work and expands on Sextant’s contributions. In
sections 3 through 6, we discuss the basic operation of Sextant,
including its area representation, its extraction of constraints from
wireless radios and sensors, and its distributed solution techniques
for node and event localization. Section 7 describes how the inter-
action between node and event localization can be used to refine
position estimates. Section 8 describes the network protocol used
to obtain and combine position estimates. Section 9 outlines the
structure and complexity of the Sextant implementation. Section 10
provides results from our simulations and physical experiments and
Section 11...
8/30/2018 2:40:28 AM +00:00
There has been extensive past work on node localization as well
as event tracking in sensor networks (see [23] for a survey). These
systems differ in the way they obtain range measurements, propa-
gate location estimates transitively, utilize positive versus negative
information, and represent potential node locations.
Range measurements can be obtained through simple connectiv-
ity, signal strength, time of arrival, time difference of arrival or an-
gle of arrival measurements. Recent work has examined heuristics
for performing range measurements via hop counts [13]. Sextant
is agnostic to the choice of range measurements, and assumes the
simplest form of range measurements based on connectivity, which
is available from any wireless radio....
8/30/2018 2:40:28 AM +00:00
A common approach to estimating node positions is direct mea-
surement or triangulation against landmarks in the immediate one-
hop vicinity. Active Badge [24] relies on the closest infrared re-
ceiver to locate specialized beacons carried by tracked assets. RADAR [19]
relies on a centralized database of signal fingerprints from land-
marks obtained at all locations and orientations to localize a node.
Lorincz andWelsh [14] propose a similar RF fingerprint-based node
localization technique that relies on strength signatures and a dis-
tributed database. Cricket [25] relies on time difference of arrival
between radio and ultrasound signals to measure distances to ded-
icated beacons. VORBA [26] uses angle of arrival measurements
from...
8/30/2018 2:40:28 AM +00:00
Using market mechanisms for resource allocation in dis-
tributed systems is not a new idea, nor is it one that
has caught on in practice or with a large body of com-
puter science research. Yet, projects that use mar-
kets for distributed resource allocation recur every few
years [1, 2, 3], and a new generation of research is
exploring market-based resource allocation mechanisms
[4, 5, 6, 7, 8] for distributed environments such as Planet-
lab, Netbed, and computational grids.
This paper has three goals. The first goal is to ex-
plore why markets can be appropriate to use for allo-
cation, when simpler allocation mechanisms exist. The
second goal is to...
8/30/2018 2:40:28 AM +00:00
However, those runners who have the patience to develop speed for shorter
distances will realise even greater benefits when they decide to take the step up
the marathon. Building speed before endurance is always the best method of
enhancing long-term performance. This article will give you a taste of speed
training as it includes an eight-week training program designed to develop your
running so you can complete a 10k race in your desired time.
The program assumes a basic level of fitness, allowing you to complete the first
week relatively comfortably. If you feel as though you might...
8/30/2018 2:40:28 AM +00:00
For optimal results in a 10km you will need to run at an intensity slightly above
your anaerobic threshold. This is quite intense in terms of physiological demand,
particularly for those wishing to achieve the distance in between 40 and 60
minute. The primary difference between the 5km and 10km in terms of demand
is that the 10km uses a greater amount of aerobic energy, and therefore strength
and endurance are also very important.
For these reasons, the primary focus of any 10km program should be boosting
anaerobic threshold, improving aerobic endurance, and developing strength to
minimise...
8/30/2018 2:40:28 AM +00:00
The speed portion of the session is conducted as a series of high-speed efforts
ranging from 200-300m in length. These efforts are aimed at improving your
maximal speed and running economy. This improved running economy will filter
down to slower speeds as well, such as your 10km race speed. Each speed
repetition is conducted in a fresh state, to allow to you hold good posture, and
achieve high speeds. While these efforts are done at a high speed, they should
not be a maximal sprint; focus on being fast, tall and in control of your technique.
...
8/30/2018 2:40:28 AM +00:00
The estimation of these quantites arises in many research areas such as in physics
and engineering problems. In network analysis such as in advanced telecommunication
systems studies X traditionally represents the length of service centers
in an open/closed queueing network processing jobs. In this context these two
quantities (1.1) represent repectively the probability of buer-over
ows and the
distribution of the queueing process in this over
ow regime.
Several numerical methods have been proposed in the literature to estimate
the entrance probability into a rare set. We refer the reader to the excellent paper
Glasserman et al. (1999) which contains a precise review on these methods as
well as a detailed...
8/30/2018 2:40:27 AM +00:00
An estimated 41,000 central line-associated bloodstream infections (CLABSI) occur in U.S. hospitals each year.1 These infections are usually serious infections typically causing a prolongation of hospital stay and increased cost and risk of mortality.
CLABSI can be prevented through proper insertion techniques and management of the central line. These techniques are addressed in the CDC’s Healthcare Infection Control Practices Advisory Committee (CDC/HIPAC) Guidelines for the Prevention of Intravascular Catheter-Related Infections, 2011.2
Settings: Surveillance will occur in any inpatient location where denominator data can be collected, which may include critical/intensive care units (ICU), specialty care areas (SCA), neonatal units including neonatal intensive care...
8/30/2018 2:40:27 AM +00:00
In 2002, an estimated 250,000 healthcare-associated pneumonias developed in U.S. hospitals and 36,000 of these were associated with deaths.1 Patients with mechanically-assisted ventilation have a high risk of developing healthcare-associated pneumonia. For the year 2011, NHSN facilities reported more than 3,525 VAPs and the incidence for various types of hospital units ranged from 0.0-4.9 per 1,000 ventilator days.2
Prevention and control of healthcare-associated pneumonia is discussed in the CDC/HICPAC document, Guidelines for Prevention of Healthcare-Associated Pneumonia, 20033. The Guideline strongly recommends that surveillance be conducted for bacterial pneumonia in ICU patients who are mechanically ventilated to facilitate identification of trends and...
8/30/2018 2:40:27 AM +00:00