اشترك بالحزمة الذهبية واحصل على وصول غير محدود شمرا أكاديميا
تسجيل مستخدم جديدImproving the performance of heterogeneous data centers through redundancy
110
0
0.0
(
0
)
اسأل ChatGPT حول البحث
ﻻ يوجد ملخص باللغة العربية
We analyze the performance of redundancy in a multi-type job and multi-type server system. We assume the job dispatcher is unaware of the servers capacities, and we set out to study under which circumstances redundancy improves the performance. With redundancy an arriving job dispatches redundant copies to all its compatible servers, and departs as soon as one of its copies completes service. As a benchmark comparison, we take the non-redundant system in which a job arrival is routed to only one randomly selected compatible server. Service times are generally distributed and all copies of a job are identical, i.e., have the same service requirement. In our first main result, we characterize the sufficient and necessary stability conditions of the redundancy system. This condition coincides with that of a system where each job type only dispatches copies into its least-loaded servers, and those copies need to be fully served. In our second result, we compare the stability regions of the system under redundancy to that of no redundancy. We show that if the servers capacities are sufficiently heterogeneous, the stability region under redundancy can be much larger than that without redundancy. We apply the general solution to particular classes of systems, including redundancy-d and nested models, to derive simple conditions on the degree of heterogeneity required for redundancy to improve the stability. As such, our result is the first in showing that redundancy can improve the stability and hence performance of a system when copies are non-i.i.d..
قيم البحث
اقرأ أيضاً
The applications being developed within the U.S. Exascale Computing Project (ECP) to run on imminent Exascale computers will generate scientific results with unprecedented fidelity and record turn-around time. Many of these codes are based on particl
e-mesh methods and use advanced algorithms, especially dynamic load-balancing and mesh-refinement, to achieve high performance on Exascale machines. Yet, as such algorithms improve parallel application efficiency, they raise new challenges for I/O logic due to their irregular and dynamic data distributions. Thus, while the enormous data rates of Exascale simulations already challenge existing file system write strategies, the need for efficient read and processing of generated data introduces additional constraints on the data layout strategies that can be used when writing data to secondary storage. We review these I/O challenges and introduce two online data layout reorganization approaches for achieving good tradeoffs between read and write performance. We demonstrate the benefits of using these two approaches for the ECP particle-in-cell simulation WarpX, which serves as a motif for a large class of important Exascale applications. We show that by understanding application I/O patterns and carefully designing data layouts we can increase read performance by more than 80%.
In this paper, we study the stability of light traffic achieved by a scheduling algorithm which is suitable for heterogeneous traffic networks. Since analyzing a scheduling algorithm is intractable using the conventional mathematical tool, our goal i
s to minimize the largest queue-overflow probability achieved by the algorithm. In the large deviation setting, this problem is equivalent to maximizing the asymptotic decay rate of the largest queue-overflow probability. We first derive an upper bound on the decay rate of the queue overflow probability as the queue overflow threshold approaches infinity. Then, we study several structural properties of the minimum-cost-path to overflow of the queue with the largest length, which is basically equivalent to the decay rate of the largest queue-overflow probability. Given these properties, we prove that the queue with the largest length follows a sample path with linear increment. For certain parameter value, the scheduling algorithm is asymptotically optimal in reducing the largest queue length. Through numerical results, we have shown the large deviation properties of the queue length typically used in practice while varying one parameter of the algorithm.
Mechanical resonances are used in a wide variety of devices; from smart phone accelerometers to computer clocks and from wireless communication filters to atomic force microscope sensors. Frequency stability, a critical performance metric, is general
ly assumed to be tantamount to resonance quality factor (the inverse of the linewidth and of the damping). Here we show that frequency stability of resonant nanomechanical sensors can generally be made independent of quality factor. At high bandwidths, we show that quality factor reduction is completely mitigated by increases in signal to noise ratio. At low bandwidths, strikingly, increased damping leads to better stability and sensor resolution, with improvement proportional to damping. We confirm the findings by demonstrating temperature resolution of 50 mu K at 200 Hz bandwidth. These results open the door for high performance ultrasensitive resonant sensors in gaseous or liquid environments, single cell nanocalorimetry, nanoscale gas chromatography, and atmospheric pressure nanoscale mass spectrometry.
With more applications moving to the cloud, cloud providers need to diagnose performance problems in a timely manner. Offline processing of logs is slow and inefficient, and instrumenting the end-host network stack would violate the tenants rights to
manage their own virtual machines (VMs). Instead, our Dapper system analyzes TCP performance in real time near the end-hosts (e.g., at the hypervisor, NIC, or top-of-rack switch). Dapper determines whether a connection is limited by the sender (e.g., a slow server competing for shared resources), the network (e.g., congestion), or the receiver (e.g., small receive buffer). Emerging edge devices now offer flexible packet processing at high speed on commodity hardware, making it possible to monitor TCP performance in the data plane, at line rate. We use P4 to prototype Dapper and evaluate our design on real and synthetic traffic. To reduce the data-plane state requirements, we perform lightweight detection for all connections, followed by heavier-weight diagnosis just for the troubled connections.
We present a new paradigm for computation of radiation spectra in the non-linear regime of operation of inverse Compton sources characterized by high laser intensities. The resulting simulations show an unprecedented level of agreement with the exper
iments. Increasing the laser intensity changes the longitudinal velocity of the electrons during their collision, leading to considerable non-linear broadening in the scattered radiation spectra. The effects of such ponderomotive broadening are so deleterious that most inverse Compton sources either remain at low laser intensities or pay a steep price to operate at a small fraction of the physically possible peak spectral output. This ponderomotive broadening can be reduced by a suitable frequency modulation (also referred to as chirping, which is not necessarily linear) of the incident laser pulse, thereby drastically increasing the peak spectral density. This frequency modulation, included in the new code as an optional functionality, is used in simulations to motivate the experimental implementation of this transformative technique.
سجل دخول لتتمكن من نشر تعليقات
التعليقات
جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
