Throughput Scalability Research Articles

Immersion by rotation-based application of the matrix for fast and reproducible sample preparations and robust results in mass spectrometry imaging.

Automated matrix deposition for matrix-assisted laser desorption/ionization (MALDI) mass spectrometry imaging (MSI) is crucial for producing reproducible analyte ion signals. Here we report an innovative method employing an automated immersion apparatus, which enables a robust matrix deposition within 5 minutes and with scalable throughput by using MAPS matrix and non-polar solvents. MSI results received from mouse heart and rat brain tissues were qualitatively similar to those from nozzle sprayed samples with respect to peak number and quality of the ion images. Overall, the immersion-method enables a fast and careful matrix deposition and has the future potential for implementation in clinical tissue diagnostics.

Scalability of Sterilizing-Grade Filters in Different Filtration Modes.

Scalability of filter throughput in normal flow filtration runs is an important consideration in the development of biopharmaceutical downstream processes. Depending on the filtration mode used, filter device geometry can significantly affect scalability. In this study, scaling of different polyethersulfone sterilizing-grade filters was performed in two filtration modes-at constant flow and at constant pressure-using a particulate model solution as well as a cell-free mAb solution as a representative example. Both filtration methods were compared regarding their practicability as well as their scalability of the final filter throughput and the filtration time. The pressure-dependent filter fouling that occurred with the mAb solution and the model solution showed that using different pressures for small- and process-scale filtration runs could potentially influence the predicted filter capacity. Overall, good scalability of the final filter throughput was determined for filters ranging from small-scale flat disc filters (4.5 cm2) to large pleated filter assemblies (5.4 m2) in filtration runs at constant flow as well as for filter capsules (0.6 m2) of up to 10" for filtration runs at constant pressure. Moreover, at constant flow, the filtration time could be accurately predicted because it was determined by the adjusted flow rate. However, at constant pressure, potential resistances in process-scale devices can result in lower fluid fluxes and, hence, a longer unpredictable filtration time compared with small-scale filter elements. This paper introduces a novel scaling method performed at constant pressure that compensates for the pressure losses resulting from process-scale device resistances. Improved scalability regarding final filter throughput and filtration time are shown here with this scaling method compared with scaling at constant pressure. Therefore, this study provides information essential in decision-making to achieve optimal scaling within biopharmaceutical process development.

ColumnML

The ability to perform machine learning (ML) tasks in a database management system (DBMS) provides the data analyst with a powerful tool. Unfortunately, integration of ML into a DBMS is challenging for reasons varying from differences in execution model to data layout requirements. In this paper, we assume a column-store main-memory DBMS, optimized for online analytical processing, as our initial system. On this system, we explore the integration of coordinate-descent based methods working natively on columnar format to train generalized linear models. We use a cache-efficient, partitioned stochastic coordinate descent algorithm providing linear throughput scalability with the number of cores while preserving convergence quality, up to 14 cores in our experiments. Existing column oriented DBMS rely on compression and even encryption to store data in memory. When those features are considered, the performance of a CPU based solution suffers. Thus, in the paper we also show how to exploit hardware acceleration as part of a hybrid CPU+FPGA system to provide on-the-fly data transformation combined with an FPGA-based coordinate-descent engine. The resulting system is a column-store DBMS with its important features preserved (e.g., data compression) that offers high performance machine learning capabilities.

Identification of Compounds That Target Acute Myeloid Leukemia Stem Cells Using a Scalable Next-Generation Screening Platform

▪The main barrier to curing acute myeloid leukemia (AML) is disease relapse, which occurs due to therapy resistance and persistence of leukemic stem cells (LSCs) after conventional induction chemotherapy. LSCs possess stem cell properties such as quiescence and self-renewal, as well as a transcriptional signature that resembles that of normal hematopoietic stem and progenitor cells. While patient-derived xenograft (PDX) models are the most stringent way to examine the properties of LSCs in human AML, this labor-intensive assay is not amenable to high throughput approaches to identify druggable vulnerabilities.Although current therapies effectively reduce tumor burden, achieving lasting event-free survival remains a challenge in the management of patients with AML. Current drug discovery efforts are focused on finding agents that eradicate bulk tumor, but not necessarily LSCs. Here, we report a next-generation screening strategy as a novel paradigm that allows the direct elucidation of molecules which antagonize LSC properties. This screening workflow is based on two scalable models of LSCs. The first is the gene expression profile of a core set of 104 genes (LSC104) that are differentially expressed between LSC+ and LSC‒ fractions obtained from primary AML samples and validated in xenotransplantation assays (Ng et al, Nature 2016). This LSC gene expression signature is strongly associated with overall survival and chemotherapy response in independent cohorts of AML patients comprising all subtypes. The second is a continuous AML culture system derived from a patient with relapsed AML that maintains, and allows the prospective enrichment of, a rare population of quiescent, self-renewing LSCs restricted to the CD34+CD38‒ fraction. These LSCs, but not the bulk tumor cells in this system, demonstrate leukemia-initiating capacity both in vitro and in PDX models. Similarly, the LSC fraction of this system, but not the bulk cells, has a LSC104 expression profile that correlates strongly with that of LSC-enriched fractions from patient samples. Together, these components form a clinically-relevant scalable model system to identify compounds with activity against LSC stemness properties, based on alterations in the proportion of LSCs in the continuous AML culture system as well as in the core LSC104 gene expression profile. To this end, we assembled a collection of 1200 curated bioactive small molecules, including 150 metabolic inhibitors and 35 targeted epigenetic probes, and performed a high-dimensional throughput screen using flow-cytometry to examine the effect of these compounds on the LSC-enriched CD34+CD38‒ fraction in our AML culture system. Candidate hits were then tested for their effects on the LSC104 profile. Rigorous analyses unveiled a number of candidate compounds with the potential to antagonize LSC properties, including several already in clinical use or in clinical trials against AML, such as cytarabine, and inhibitors of FLT3, CDK, PLK1, and aurora kinase, as well as several classes of compounds not previously described against AML, targeting NAMPT, BRPF1B, CHK1, and KSP, among others.In conclusion, we describe a next-generation scalable throughput approach that integrates stem cell biology to uncover modulators of stemness in AML. This multi-parametric screening strategy serves as a platform for the deeper understanding of druggable vulnerabilities of LSCs, and serves as a starting point to achieve lasting event-free survival in AML patients. DisclosuresChan:AbbVie: Research Funding; Celgene: Research Funding; Genentech: Research Funding.

Throughput Scalability Analysis of Fork-Join Queueing Networks

Parallel and distributed processing systems have expanded in size as technology advances in cloud computing and big data analytics. A critical issue concerns throughput scalability: whether throughput decreases to zero as the systems scale in size and capabilities. We model parallel and distributed processing systems as fork and join queueing networks with blocking (FJQN/Bs). Such networks can have arbitrary topology, arbitrary initial state, and generally distributed service times. We propose a key topological concept, called the “minimum level,” that determines the throughput scalability of FJQN/Bs. We construct throughput bounds as functions of minimum level, network degree, buffer sizes, and processing speed, and we present necessary and/or sufficient conditions to guarantee throughput scalability of arbitrary size and topology FJQN/Bs. The e-companion is available at https://doi.org/10.1287/opre.2018.1748 .

Quantifying the Cost of Distrust: Comparing Blockchain and Cloud Services for Business Process Execution

Blockchain is of rising importance as a technology for engineering applications in cross-organizational settings, avoiding reliance on central trusted third-parties. The use of blockchain, instead of traditional databases or services, is an architectural choice in the development of a software system. Architecture impacts the non-functional qualities of systems, creating design trade-offs between these qualities. The costs of execution and storage are important non-functional qualities, but as yet little is known about them for blockchain-based systems. How expensive is it to use blockchains compared to conventional execution and storage infrastructure? We investigate this question using business process execution as a lens. Specifically, we compare the cost for computation and storage of business process execution on blockchain vs. a popular cloud service. Besides monetary cost, blockchains like Ethereum limit the complexity of new blocks by capping costs through network-defined limits. For applications using such blockchains, the limit per block, thus, translates into an upper bound on throughput scalability. First, we implement and measure the cost of business process execution on blockchain and cloud services for a business process model from a large-scale industrial dataset and an example from literature. We observe two orders of magnitude difference in this cost. Second, we illustrate how cost models can be used to project the impact of different workload assumptions. Finally, we discuss throughput scalability limits as well as trade-offs between cost and other non-functional qualities in the design of blockchain-based systems.

On the scalability of dynamic Flex-Grid/SDM optical core networks

Future optical core networks realizing Flex-Grid and Space Division Multiplexing (Flex-Grid/SDM) will be key to effectively cope with the rapid growth of the Internet traffic, thanks to their superior capacity and spectrum utilization flexibility. Specifically about SDM, the spatial dimension can be considered either fixed or flexible. The goal of this work is two-fold. Firstly, it aims at quantifying the capacity (in terms of throughput) of Flex-Grid/SDM networks assuming either a fixed or a flexible spatial dimension. To this end, we scale up the spatial multiplicity from 7 to 30 spatial channels, considering Multi-Fiber (MF) and Multi-Core Fiber (MCF) solutions. Secondly, it also aims at analyzing the throughput scalability in these networks versus their hardware requirements, especially from the transponder requirements point of view. The obtained results show that, for a 4.3× spatial multiplicity scaling factor, the throughput of spatially fixed Flex-Grid/SDM networks scales up to 3.5×. Conversely, when the spatial dimension is considered flexible, throughput can be pushed up to a 5× factor for both MF and MCF solutions. However, a long-term realization of spectrally-spatially flexible optical networks (SS-FONs) requires up to 9.6× more laser sources and 2.7× electronic complexity (per required connection) in the transponders compared to spatially fixed Flex-Grid/SDM networks. Furthermore, SS-FONs enable different options for assigning the spectral and spatial resources to traffic demands. In this sense, another contribution of this work is the introduction of a novel methodology that trades network throughput for transponder (the major network equipment cost in SDM networks) complexity reduction.

Concurrent hash tables on multicore machines: Comparison, evaluation and implications

Concurrent hash table has been an area of active research in recent years, and a wide variety of fast and efficient concurrent hash tables (CHTs) have been proposed to exploit the advantages of modern parallel computer architectures such as today’s mainstream multi-core systems. As one of the fundamental data structures widely used in software systems, existing works on CHTs focus on either algorithmic improvements, or hardware-oriented optimizations, or application-specific designs. However, there is a lack of a comprehensive and comparative study on different implementations. In this paper, we conduct an experimental study on the state-of-the-art, and our goal is to critically review existing CHTs from wider aspects and with more detailed analysis. Concretely, we have conducted extensive evaluations of five CHTs using a unified testing framework on four multi-core hardware platforms, and implemented our HTM-based concurrent hash table. A variety of metrics such as throughput scalability, latency, impact of memory hierarchy, thread pinning strategies, synchronization mechanisms, and memory consumption, are measured in order to obtain the deep insights about performance impediments and good design choices. With this study, we hope to identify potential issues and pinpoint promising directions for future research of CHTs.

Implementation of Database Massively Parallel Processing System to Build Scalability on Process Data Warehouse

The problem with telecommunications companies today is that transactional data is more extensive than existing source tables. This makes business reporting less efficient and overwhelms query processing results in data warehouses so that they do not meet business requirements. The fast and complex evolution of the digital world must be scalable to the data warehouse process, so that the authors implement it in the data warehouse using massive parallel processing (MPP) with the Greenplum database, so that business users can get reports faster and more optimally. This case study explains how the MPP system implements and measures the performance of the Greenplum database by performing complex queries in the data warehouse with parallel processing. Therefore, this case study analyzes whether the use of MPP systems can measure the scalability of throughput and the response time in the data warehouse so that system performance in the Greenplum database remains stable for daily, weekly, and monthly operations.

Continuous and Quantitative Purification of T-Cell Subsets for Cell Therapy Manufacturing Using Magnetic Ratcheting Cytometry.

T-cell-based immunotherapies represent a growing medical paradigm that has the potential to revolutionize contemporary cancer treatments. However, manufacturing bottlenecks related to the enrichment of therapeutically optimal T-cell subpopulations from leukopak samples impede scale-up and scale-out efforts. This is mainly attributed to the challenges that current cell purification platforms face in balancing the quantitative sorting capacity needed to isolate specific T-cell subsets with the scalability to meet manufacturing throughputs. In this work, we report a continuous-flow, quantitative cell enrichment platform based on a technique known as ratcheting cytometry that can perform complex, multicomponent purification targeting various subpopulations of magnetically labeled T cells directly from apheresis or peripheral blood mononuclear cell (PBMC) samples. The integrated ratcheting cytometry instrument and cartridge demonstrated enrichment of T cells directly from concentrated apheresis samples with a 97% purity and an 85% recovery of magnetically tagged cells. Magnetic sorting of different T-cell subpopulations was also accomplished on chip by multiplexing cell surface targets onto particles with differing magnetic strengths. We believe that ratcheting cytometry's quantitative capacity and throughput scalability represents an excellent technology candidate to alleviate cell therapy manufacturing bottlenecks.

Wireless Multihop Device-to-Device Caching Networks

We consider a wireless device-to-device network, where $n$ nodes are uniformly distributed at random over the network area. We let each node caches $M$ files from a library of size $m\geq M$ . Each node in the network requests a file from the library independently at random, according to a popularity distribution, and is served by other nodes having the requested file in their local cache via (possibly) multihop transmissions. Under the classical “protocol model” of wireless networks, we characterize the optimal per-node capacity scaling law for a broad class of heavy-tailed popularity distributions, including Zipf distributions with exponent less than one. In the parameter regime of interest, i.e., $m=o(nM)$ , we show that a decentralized random caching strategy with uniform probability over the library yields the optimal per-node capacity scaling of $\Theta (\sqrt {M/m})$ for heavy-tailed popularity distributions. This scaling is constant with $n$ , thus yielding throughput scalability with the network size. Furthermore, the multihop capacity scaling can be significantly better than for the case of single-hop caching networks, for which the per-node capacity is $\Theta (M/m)$ . The multihop capacity scaling law can be further improved for a Zipf distribution with exponent larger than some threshold > 1, by using a decentralized random caching uniformly across a subset of most popular files in the library. Namely, ignoring a subset of less popular files (i.e., effectively reducing the size of the library) can significantly improve the throughput scaling while guaranteeing that all nodes will be served with high probability as $n$ increases.

Low-Latency, Low-Area, and Scalable Systolic-Like Modular Multipliers for $GF(2^{m})$ Based on Irreducible All-One Polynomials

In this paper, an efficient recursive formulation is suggested for systolic implementation of canonical basis finite field multiplication over GF(2 m ) based on irreducible AOP. We have derived a recursive algorithm for the multiplication, and used that to design a regular and localized bit-level dependence graph (DG) for systolic computation. The bit-level regular DG is converted into a fine-grained DG by node-splitting, and mapped that into a parallel systolic architecture. Unlike most of the existing structures, it does not involve any global communications for modular reduction. The proposed bit-parallel systolic structure has the same cycle time as that of the best existing bit-parallel systolic structure, but involves significantly less number of registers. The proposed bit-parallel design has a scalable latency of l + ⌈log 2 s⌉ +1 cycles which is considerably low compared with those of existing systolic designs. Moreover, the proposed time-multiplexed structure is designed specifically for scalability of throughput and hardware-complexity to meet the area-time trade-off in resource-constrained applications while maintaining or reducing the overall latency. The ASIC synthesis report shows that the proposed bit-parallel structures offers nearly 30% saving of area and nearly 38% saving of power consumption over the best of the existing AOP-based systolic finite field multiplier.

Reply to the 'Comment on "Robust scalable high throughput production of monodisperse drops"' by M. Nakajima, Lab Chip, 2017, 17, DOI: 10.1039/C7LC00181A.

This reply to the comment by Nakajima on our article that appeared in Lab on a Chip (E. Amstad, M. Chemama, M. Eggersdorfer, L. R. Arriaga, M. Brenner and D. A. Weitz, Lab Chip, 2016, 16, 4163-4172) highlights the differences between the microchannel step emulsification devices developed by the Nakajima group and the millipede device reported by us in Lab on a Chip.

Comment on "Robust scalable high throughput production of monodisperse drops" by E. Amstad, M. Chemama, M. Eggersdorfer, L. R. Arriaga, M. P. Brenner and D. A. Weitz, Lab Chip, 2016, 16, 4163.

This comment on an article that appeared in Lab on a Chip (Amstad et al., Lab Chip, 2016, 16, 4163-4172) provides information on the performance of microchannel (step) emulsification devices developed by the Nakajima Group.

Quality-Aware Streaming and Scheduling for Device-to-Device Video Delivery

On-demand video streaming is becoming a killer application for wireless networks. Recent information-theoretic results have shown that a combination of caching on the users’ devices and device-to-device (D2D) communications yields throughput scalability for very dense networks, which represent critical bottlenecks for conventional cellular and wireless local area network (WLAN) technologies. In this paper, we consider the implementation of such caching D2D systems where each device pre-caches a subset of video files from a library, and users requesting a file that is not already in their library obtain it from neighboring devices through D2D communication. We develop centralized and distributed algorithms for the delivery phase, encompassing a link scheduling and a streaming component. The centralized scheduling is based on the max-weighted independent set (MWIS) principle and uses message-passing to determine max-weight independent sets. The distributed scheduling is based on a variant of the FlashLinQ link scheduling algorithm, enhanced by introducing video-streaming specific weights. In both cases, the streaming component is based on a quality-aware stochastic optimization approach, reminiscent of current Dynamic Adaptive Streaming over HTTP (DASH) technology, for which users sequentially request video “chunks” by choosing adaptively their quality level. The streaming and the scheduling components are coupled by the length of the users’ request queues. Through extensive system simulation, the proposed approaches are shown to provide sizeable gains with respect to baseline schemes formed by the concatenation of off-the-shelf FlashLinQ with proportional fair link scheduling and DASH at the application layer.

A Necessary and Sufficient Condition for Throughput Scalability of Fork and Join Networks with Blocking

Due to emerging applications such as cloud computing and big data analytics, modern information processing systems are growing increasingly large and complex. A critical issue concerns the throughput performance as the system grows in size. This paper models distributed information processing systems as fork and join queueing networks with blocking. We identify necessary and sufficient conditions for throughput scalability of such fork and join networks as they grow in size. Previous studies have either focused on special structured networks such as tandem or tree networks, or provided only necessary conditions for throughput scalability. In this paper, we show that such necessary conditions are not sufficient. We present a key topological concept called ``minimum level" of the underlying graph, and develop lower and upper bounds for the throughput of arbitrary FJQN/Bs. The bounds depend on network degree, minimum level, deterministic cycle time, buffer sizes, and service time distributions, but not on network size. We show that level-boundedness and degree-boundedness are necessary and sufficient conditions to guarantee that the throughput of an FJQN/B is bounded away from zero as network size goes to infinity.

63 - Performance of the Genomic DNA Assay for the Agilent 4200 TapeStation System

The new Agilent 4200 TapeStation system provides an automated system for fast and reliable DNA and RNA electrophoresis with scalable throughput from 1 to 96 samples. The Agilent Genomic DNA ScreenTape assay has been developed for the separation and analysis of gDNA from 200 up to greater than 60,000 base pairs. It provides accurate quantification data, sizing and an automated numerical assessment of gDNA quality based on the DNA integrity number (DIN)1. The Agilent 4200 TapeStation system and the Genomic DNA ScreenTape assay are ideal QC tools for next generation sequencing (NGS) and array comparative genomic hybridization (aCGH) workflows.

An FPGA-Based SIMD Architecture for Video Compression with Scalable Throughput

Data processing using spatial parallelism from single instruction, multiple data (SIMD) machines can be used in image processing to improve system performance regarding data throughput. The spatial parallelism is an alternative for increasing the device operating frequency, generally limited by hardware constraints. In parallel architectures for video processing with high requirements in terms of data throughput, different processing units (PUs) simultaneously operate on pixels belonging to nonoverlapping regions of the image frame.

A game theory‐inspired channel allocation policy for multi‐radio pervasive ad hoc communications

AbstractThe impressive increase of innovative wireless communication technologies and applications represents nowadays a key approach to enable pervasive communications environments. In particular, the emerging paradigms of the Internet of Things and capillary networks offer effective ways to make devices connected. This has triggered the development of effective methodologies and procedures to allow an unlimited number of devices to exchange information mainly in an autonomous mode. However, it is a well‐known concept that wireless networks capacity usually decreases with the number of nodes. In particular, in order to improve the throughput scalability, a promising approach is that of resorting to the use of multiple radio interfaces at each node, in order to exploit spatial reuse of frequencies. Towards this end, game theory methodologies offer efficient approaches to solve the complex radio interfaces selection and allocation problems. This paper proposes a game theory‐inspired approach to efficiently select the number of radio interfaces to be used at each node site in order to lower the energy consumption and maximise the end‐to‐end throughput of any communication on which the node is involved in. The good behaviour of the proposed approach is validated by provided theoretical framework and numerical results derived by considering different data packets wireless forwarding schemes. Copyright © 2015 John Wiley & Sons, Ltd.

Wireless Backhaul Networks: Capacity Bound, Scalability Analysis and Design Guidelines

This paper studies the scalability of a wireless backhaul network modeled as a random extended network with multiantenna base stations (BSs), where the number of antennas per BS is allowed to scale as a function of the network size. The antenna scaling is justified by the current trend toward the use of higher carrier frequencies, which allows packing a large number of antennas in small form factors. The main goal is to study the per-BS antenna requirement that ensures scalability of this network, i.e., its ability to deliver nonvanishing rate to each source–destination pair. We first derive an information theoretic upper bound on the capacity of this network under a general propagation model, which provides a lower bound on the per-BS antenna requirement. Then, we characterize the scalability requirements for two competing strategies of interest: 1) long hop : each source–destination pair minimizes the number of hops by sacrificing multiplexing gain while achieving full beamforming (power) gain over each hop; and 2) short hop : each source–destination pair communicates through a series of short hops, each achieving full multiplexing gain. While long hop may seem more intuitive in the context of massive multiple-input–multiple-output transmission, we show that the short hop strategy is significantly more efficient in terms of per-BS antenna requirement for throughput scalability. As a part of the proof, we construct a scalable short hop strategy and show that it does not violate any fundamental limits on the spatial degrees of freedom.