Workload Characteristics Research Articles

Workload-optimized sensor data store for industrial IoT gateways

In industrial Internet of Things (IoT) environments, sensor devices continue to generate a stream of sensor data, and the management of an ever-growing amount of data is a vital feature for IoT gateways. However, our preliminary analysis of some popular key–value stores used in IoT gateways revealed that none of the systems exploits the distinctive characteristics of IoT workloads – append-only and immutable – thus the systems show limitations in managing sensor data. To address this issue, in this study, we propose Indexing-of-Indexes (IoI) and LogFlush-and-Append (LFA) to exploit such characteristics. IoI is an indexing and data organization scheme designed to eliminate the notorious compaction-induced write amplification observed in legacy key–value stores, and LFA is a data ingestion scheme intended to remove double data write issues in legacy write-ahead logging implementations. We implement a prototype key–value store, Sen-Store, which incorporates our proposals, and we evaluate its performance using synthetic workloads and the TPCx-IoT benchmark. The evaluation results show that Sen-Store achieves up to 17.6× and 2.1× higher IoTps than industry-leading RocksDB and state-of-the-art IoTDB, respectively.

The full story of 1000 cores

In our initial DaMoN paper, we set out the goal to revisit the results of “Starring into the Abyss [...] of Concurrency Control with [1000] Cores” (Yu in Proc. VLDB Endow 8: 209-220, 2014). Against their assumption, today we do not see single-socket CPUs with 1000 cores. Instead, multi-socket hardware is prevalent today and in fact offers over 1000 cores. Hence, we evaluated concurrency control (CC) schemes on a real (Intel-based) multi-socket platform. To our surprise, we made interesting findings opposing results of the original analysis that we discussed in our initial DaMoN paper. In this paper, we further broaden our analysis, detailing the effect of hardware and workload characteristics via additional real hardware platforms (IBM Power8 and 9) and the full TPC-C transaction mix. Among others, we identified clear connections between the performance of the CC schemes and hardware characteristics, especially concerning NUMA and CPU cache. Overall, we conclude that no CC scheme can efficiently make use of large multi-socket hardware in a robust manner and suggest several directions on how CC schemes and overall OLTP DBMS should evolve in future.

Automatic Stream Identification to Improve Flash Endurance in Data Centers

The demand for high performance I/O in Storage-as-a-Service (SaaS) is increasing day by day. To address this demand, NAND Flash-based Solid-state Drives (SSDs) are commonly used in data centers as cache- or top-tiers in the storage rack ascribe to their superior performance compared to traditional hard disk drives (HDDs). Meanwhile, with the capital expenditure of SSDs declining and the storage capacity of SSDs increasing, all-flash data centers are evolving to serve cloud services better than SSD-HDD hybrid data centers. During this transition, the biggest challenge is how to reduce the Write Amplification Factor (WAF) as well as to improve the endurance of SSD since this device has a limited program/erase cycles. A specified case is that storing data with different lifetimes (i.e., I/O streams with similar temporal fetching patterns such as reaccess frequency) in one single SSD can cause high WAF, reduce the endurance, and downgrade the performance of SSDs. Motivated by this, multi-stream SSDs have been developed to enable data with a different lifetime to be stored in different SSD regions. The logic behind this is to reduce the internal movement of data—when garbage collection is triggered, there are high chances of having data blocks with either all the pages being invalid or valid. However, the limitation of this technology is that the system needs to manually assign the same streamID to data with a similar lifetime. Unfortunately, when data arrives, it is not known how important this data is and how long this data will stay unmodified. Moreover, according to our observation, with different definitions of a lifetime (i.e., different calculation formulas based on selected features previously exhibited by data, such as sequentiality, and frequency), streamID identification may have varying impacts on the final WAF of multi-stream SSDs. Thus, in this article, we first develop a portable and adaptable framework to study the impacts of different workload features and their combinations on write amplification. We then propose a feature-based stream identification approach, which automatically co-relates the measurable workload attributes (such as I/O size, I/O rate, and so on.) with high-level workload features (such as frequency, sequentiality, and so on.) and determines a right combination of workload features for assigning streamIDs . Finally, we develop an adaptable stream assignment technique to assign streamID for changing workloads dynamically. Our evaluation results show that our automation approach of stream detection and separation can effectively reduce the WAF by using appropriate features for stream assignment with minimal implementation overhead.

An SMDP-based approach to thermal-aware task scheduling in NoC-based MPSoC platforms

In this paper, we consider the operation of a thermal-aware task scheduler, dispatching tasks from an arrival queue as well as setting the voltage and frequency of the processing cores to optimize the mean temperature margin of the entire chip (i.e., cores as well as the NoC routers). We model the decision process of the task scheduler as a semi-Markov decision problem (SMDP) to account for the most common uncertainties prevalent in MPSoC systems (including: the stochastic nature of the workload inter-arrival times, time-varying workload characteristics, the uncertain chip thermal profile as well as random inter-task communications). SMDP is among the fairly general variants of continuous-time optimization frameworks from the stochastic control theory and is a much more efficient choice compared to discrete-time formalisms (which would lead to an increase in task waiting times and degraded system performance). To solve the formulated SMDP, we propose two reinforcement learning (RL) algorithms that are capable of computing the optimal task assignment policy without requiring the statistical knowledge of the stochastic dynamics underlying the system states. The proposed algorithms also rely on function approximation techniques to handle the infinite length of the task queue as well as the continuous nature of temperature readings. Compared to previous work, the simulations demonstrate nearly 6 Kelvin reduction in system average peak temperature and 66 milliseconds decrease in mean task service time.

Look-ahead energy efficient VM allocation approach for data centers

Energy efficiency is an important issue for reducing environmental dissipation. Energy efficient resource provisioning in cloud environments is a challenging problem because of its dynamic nature and varied application workload characteristics. In the literature, live migration of virtual machines (VMs) among servers is commonly proposed to reduce energy consumption and to optimize resource usage, although it comes with essential drawbacks, such as migration cost and performance degradation. Energy efficient provisioning is addressed at the data center level in this research. A novel efficient resource management algorithm for virtualized data centers that optimizes the number of servers to meet the requirements of dynamic workloads without migration is proposed in this paper. The proposed approach, named Look-ahead Energy Efficient VM Allocation (LAA), contains a Holt Winters-based prediction module. Energy efficiency and performance are inversely proportional. The energy-performance trade-off relies on periodic comparisons of the predicted and active numbers of servers. To evaluate the proposed algorithm, experiments are conducted with real-world workload traces from Google Cluster. LAA is compared with the best approach provided by CloudSim based on VM migration called Local Regression-Minimum Migration Time (LR-MMT). The experimental results show that the proposed algorithm leads to a consumption reduction of up to 45% to complete one workload compared with the LR-MMT.

Performance Investigation of Fiber Distributed Data Interface in Networking

Abstract: The extent of the ring, the number of stations on the ring, the number of stations waiting to communicate, and the frame size are all setup and workload characteristics that affect the performance of an FDDI LAN. In addition, the target token rotation time is a significant parameter that network management may regulate to optimise performance (TTRT). The influence of the TTRT on various performance indicators for different ring topologies was investigated using analytical modelling and simulation approaches. The performance of the FDDI network is the topic of this research. The functionality of the FDDI network was then assessed in terms of efficiency, reaction time, and network cost using various values of ring latency, fixed target token rotation time (TTRT), and number of stations. The entire project is done in MATLAB. Keywords: FDDI LAN, TTRT, Interface, Networking, MATLAB

Predicting physician burnout using clinical activity logs: Model performance and lessons learned

BackgroundBurnout is a significant public health concern affecting more than half of the healthcare workforce; however, passive screening tools to detect burnout are lacking. We investigated the ability of machine learning (ML) techniques to identify burnout using passively collected electronic health record (EHR)-based audit log data. MethodPhysician trainees participated in a longitudinal study where they completed monthly burnout surveys and provided access to their EHR-based audit logs. Using the monthly burnout scores as the target outcome, we trained ML models using combinations of features derived from audit log data—aggregate measures of clinical workload, time series-based temporal measures of EHR use, and the baseline burnout score. Five ML models were constructed to predict burnout as a continuous score: penalized linear regression, support vector machine, neural network, random forest, and gradient boosting machine. Results88 trainee physicians participated and completed 416 surveys; greater than10 million audit log actions were collected (Mean [Standard Deviation] = 25,691 [14,331] actions per month, per physician). The workload feature set predicted burnout score with a mean absolute error (MAE) of 0.602 (95% Confidence Interval (CI), 0.412–0.826), and was able to predict burnout status with an average AUROC of 0.595 (95% CI 0.355–0.808) and average accuracy 0.567 (95% CI 0.393–0.742). The temporal feature set had a similar performance, with MAE 0.596 (95% CI 0.391–0.826), and AUROC 0.581 (95% CI 0.343–0.790). The addition of the baseline burnout score to the workload features improved the model performance to a mean AUROC of 0.829 (95% CI 0.607–0.996) and mean accuracy of 0.781 (95% CI 0.587–0.936); however, this performance was not meaningfully different than using the baseline burnout score alone. ConclusionsCurrent findings illustrate the complexities of predicting burnout exclusively based on clinical work activities as captured in the EHR, highlighting its multi-factorial and individualized nature. Future prediction studies of burnout should account for individual factors (e.g., resilience, physiological measurements such as sleep) and associated system-level factors (e.g., leadership).

Pattern-Based Prefetching with Adaptive Cache Management Inside of Solid-State Drives

This article proposes a pattern-based prefetching scheme with the support of adaptive cache management, at the flash translation layer of solid-state drives ( SSDs ). It works inside of SSDs and has features of OS dependence and uses transparency. Specifically, it first mines frequent block access patterns that reflect the correlation among the occurred I/O requests. Then, it compares the requests in the current time window with the identified patterns to direct prefetching data into the cache of SSDs. More importantly, to maximize the cache use efficiency, we build a mathematical model to adaptively determine the cache partition on the basis of I/O workload characteristics, for separately buffering the prefetched data and the written data. Experimental results show that our proposal can yield improvements on average read latency by 1.8 %– 36.5 % without noticeably increasing the write latency, in contrast to conventional SSD-inside prefetching schemes.

The Concurrent Learned Indexes for Multicore Data Storage

We present XIndex, which is a concurrent index library and designed for fast queries. It includes a concurrent ordered index (XIndex-R) and a concurrent hash index (XIndex-H). Similar to a recent proposal of the learned index, the indexes in XIndex use learned models to optimize index efficiency. Compared with the learned index, for the ordered index, XIndex-R is able to handle concurrent writes effectively and adapts its structure according to runtime workload characteristics. For the hash index, XIndex-H is able to avoid the resize operation blocking concurrent writes. Furthermore, the indexes in XIndex can index string keys much more efficiently than the learned index. We demonstrate the advantages of XIndex with YCSB, TPC-C (KV), which is a TPC-C-inspired benchmark for key-value stores, and micro-benchmarks. Compared with ordered indexes of Masstree and Wormhole, XIndex-R achieves up to 3.2× and 4.4× performance improvement on a 24-core machine. Compared with hash indexes of Intel TBB HashMap, XIndex-H achieves up to 3.1× speedup. The performance further improves by 91% after adding the optimizations on indexing string keys. The library is open-sourced. 1

Storage Type and Hot Partition Aware Page Reclamation for NVM Swap in Smartphones

With the rapid advances in mobile app technologies, new activities using smartphones emerge every day including social network and location-based services. However, smartphones experience problems in handling high priority tasks, and often close apps without the user’s agreement when there is no available memory space. To cope with this situation, supporting swap with fast NVM storage has been suggested. Although swap in smartphones incurs serious slowing-down problems in I/O operations during saving and restoring the context of apps, NVM has been shown to resolve this problem due to its fast I/O features. Unlike previous studies that only focused on the management of NVM swap itself, this article discusses how the memory management system of smartphones can be further improved with NVM swap. Specifically, we design a new page reclamation algorithm for smartphone memory systems, which considers the following: (1) storage types of each partition (i.e., file system for flash storage and swap for NVM), and (2) access hotness of each partition including operation types and workload characteristics. By considering asymmetric I/O cost and access density for each partition, our algorithm improves the I/O performance of smartphones significantly. Specifically, it improves the I/O time by 15.0% on average and by up to 35.1% compared to the well-known CLOCK algorithm.

Efficient Stack Distance Approximation Based on Workload Characteristics

A stack distance of a reference is the depth from which the reference must be extracted from a stack. It has been widely applied to a variety of applications utilizing temporal locality information. However, calculating an exact stack distance requires significant computational complexity that is impractical for online production systems. This paper proposes an efficient algorithm that approximates a stack distance based on workload characteristics. The proposed algorithm utilizes two simple reference metrics: inter-reference distance and a quasi-unique reference count. More importantly, its approximation algorithm requires surprisingly simple calculation. Consequently, it significantly reduces computational overhead while exhibiting exceptional accuracy and runs extremely fast. Moreover, it allows configurable parameters for performance optimization by making optimal use of workload characteristics. Our extensive experiments with diverse realistic workloads demonstrate our stack distance approximation algorithm outperforms the current state-of-the-art algorithm with higher accuracy (3.7×), lower memory footprint (3.05×) and fast execution (less than 0.05 seconds) on average. In addition, with configuration parameters, our refined algorithm can achieve a 28% average additional accuracy improvement.

Three-Way Ensemble Prediction for Workload in the Data Center

Accurate prediction of data center workload, an important technology of cloud computing, is particular in improving resource utilization and reducing energy consumption. However, the workload presents a quasi-volatile, is challenging to obtain accurate results in cloud resource management. In this paper, the three-way ensemble prediction for workload in the data center is first proposed to improve the accuracy of the prediction. Moreover, we first defined the workload as the stable period, the volatility period, and the jitter period and adopted a simulated annealing algorithm to learn the optimal threshold to divide the workload. Then, according to the basic idea of the three-way decision, we assigned various prediction models based on workload characteristics and a priori error prediction to improve the prediction accuracy further. Finally, all the experimental results carried on the CPU load monitoring logs from Google cluster trace and compared with ARIMA, NN, and DMASVR-3WD, TWD-RCPM improve the accuracy of workload prediction by 69.0%, 68.6% and 72.6%, respectively.

A Novel Workload-Aware and Optimized Write Cycles in NVRAM

With the emergence of the Internet of things (IoT), embedded systems have now changed its dimensionality and it is applied in various domains such as healthcare, home automation and mainly Industry 4.0. These Embedded IoT devices are mostly battery-driven. It has been analyzed that usage of Dynamic Random-Access Memory (DRAM) centered core memory is considered the most significant source of high energy utility in Embedded IoT devices. For achieving the low power consumption in these devices, Non-volatile memory (NVM) devices such as Parameter Random Access Memory (PRAM) and Spin-Transfer Torque Magnetic Random-Access Memory (STT-RAM) are becoming popular among main memory alternatives in embedded IoT devices because of their features such as high thickness, byte addressability, high scalability and low power intake. Additionally, Non-volatile Random-Access Memory (NVRAM) is widely adopted to save the data in the embedded IoT devices. NVM, flash memories have a limited lifetime, so it is mandatory to adopt intelligent optimization in managing the NVRAM-based embedded devices using an intelligent controller while considering the endurance issue. To address this challenge, the paper proposes a powerful, lightweight machine learning-based workload-adaptive write schemes of the NVRAM, which can increase the lifetime and reduce the energy consumption of the processors. The proposed system consists of three phases like Workload Characterization, Intelligent Compression and Memory Allocators. These phases are used for distributing the write-cycles to NVRAM, following the energy-time consumption and number of data bytes. The extensive experimentations are carried out using the IoMT (Internet of Medical things) benchmark in which the different endurance factors such as application delay, energy and write-time factors were evaluated and compared with the different existing algorithms.

Comparative performance evaluation of suboptimal binary search trees

Three relevant types of suboptimal binary search trees are comparatively evaluated in this paper: two well-known representatives of height-balanced approaches (the AVL and red-black trees) and a popular self-adjusting splay tree. After a brief theoretical background, an evaluation method was described that employs a suitable synthetic workload method capable of producing diverse desired workload characteristics (different distributions and ranges of key values, varying input sequence lengths, etc.). Evaluation analysis was conducted for search, insert, and delete operations separately for each particular type and in appropriate combinations. Experimental results for an average operation cost as well as for tree maintenance cost are comparatively presented and carefully discussed. Finally, the suggested favorable conditions for application of each tree type are summarized.

DeepCM: Deep neural networks to improve accuracy prediction of database cost models

AbstractA major challenge for many database management tasks including admission control, query scheduling, progress monitoring and self‐driving data storage systems is to enhance queries performances which are based on computational models known as database cost models. One of the most challenging aspects of developing accurate database cost models is identifying their parameters and capturing their relationships, consequently we can derive the query execution cost on the basis of a specific database hosted on a given platform. Furthermore, the highly dynamic workload (i.e., a set of queries) and the query execution variation lead to performance degradation risk, therefore cost models need to be improved by considering newer software configuration and future workload characteristics. In this article, we propose a framework called DeepCM that is based on a min–max optimization for building robust database cost model against uncertainty parameters. Furthermore, our framework is based on Robust Deep Neural Networks to build database cost models that guarantee a high accuracy regardless of variations from software configuration and workload characteristics. Several experiments have been done to evaluate the robustness of produced cost models and findings show that DeepCM provides a high cost model prediction accuracy and stable performance.

Database workload characterization with query plan encoders

Smart databases are adopting artificial intelligence (AI) technologies to achieve instance optimality , and in the future, databases will come with prepackaged AI models within their core components. The reason is that every database runs on different workloads, demands specific resources, and settings to achieve optimal performance. It prompts the necessity to understand workloads running in the system along with their features comprehensively, which we dub as workload characterization. To address this workload characterization problem, we propose our query plan encoders that learn essential features and their correlations from query plans. Our pretrained encoders captures the structural and the computational performance of queries independently. We show that our pretrained encoders are adaptable to workloads that expedites the transfer learning process. We performed independent assessments of structural encoder and performance encoders with multiple downstream tasks. For the overall evaluation of our query plan encoders, we architect two downstream tasks (i) query latency prediction and (ii) query classification. These tasks show the importance of feature-based workload characterization. We also performed extensive experiments on individual encoders to verify the effectiveness of representation learning, and domain adaptability.

Identifying the Subtypes and Characteristics of Mental Workload Among Chinese Physicians in Outpatient Practice: A Latent Profile Analysis.

Objective: The purpose of this study is to investigate the mental workload level of physicians in outpatient practice since the normalization of prevention and control of the COVID-19 pandemic in China and explore the subtypes of physicians regarding their mental workload.Methods: A cross-sectional survey of 1,934 physicians primarily in 24 hospitals in 6 provinces in Eastern, Central, and Western China was conducted from November 2020 to February 2021. A latent profile analysis was performed to identify clusters based on the six subscales of the Chinese version of physician mental workload scale developed by our research team. Chi-square tests were performed to explore the differences in demographic characteristics of the subtypes among the subgroups, and multinomial logistic regression analysis was further conducted to identify the determinants of the subtypes of physicians.Results: Overall, the participating physicians reported high levels of task load but with high self-assessed performance (68.01 ± 14.25) while performing communication work tasks characterized by direct patient interaction in outpatient clinics. About 33.8% of the participating physicians were identified as “high workload and high self-assessment” subtype, compared to 49.7% “medium workload and medium self-assessment” subtype and 16.4% “low workload and low self-assessment” subtype. Physicians in “high workload and high self-assessment” subtype had the highest mean mental workload score. Physicians who were female, younger, married, worse health status, those who had lower educational level and an average monthly income of 5,001–10,000 RMB, those who worked in tertiary A hospitals, more hours per week and more than 40 h per week in outpatient clinics, and those who saw more outpatients per day, and spent more time per patient but with higher outpatient satisfaction were more likely to belong to “high workload and high self-assessment” subtype.Conclusion: Our findings can help provide a solid foundation for developing targeted interventions for individual differences across physicians regarding their mental workload. We suggest the hospital managers should pay more attention to those physicians with characteristics of the “high workload and high self-assessment” subtype and strengthen the management of the workload of this subtype of physicians to reduce the risks of their mental health, and to maintain their high work performance in outpatient clinics.

HUBUNGAN ANTARA KARAKTERISTIK DEMOGRAFI PEKERJA, BEBAN KERJA MENTAL, DAN GAYA KEPEMIMPINAN TERHADAP BURNOUT PADA STAFF ADMINISTRASI FKM UNDIP

Burnout is physical, mental, and emotional exhaustion that occurs continuously which causes a person to lose motivation at work until appears the behavior of withdrawing from work to the desire to stop working, burnout is also caused by a mismatch between job demands and the rewards received. In carrying out work, it is not uncommon for workers to experience physical or mental exhaustion to experience prolonged stress. This prolonged stress is also known as burnout. Burnout sufferers are often found in public service workers such as teachers, police, nurses, and administrative workers are no exception. Administrative staffs are quite vulnerable to burnout because administrative staffs are required to work calmly, carefully, thoroughly, and able to work with a team. The factors that influence burnout consist of internal and external factors. Internal factors consist of demographic characteristics (age, gender, marital status, education, and years of service), personal characteristics (personality characteristics), and the attitude of how a person responds to his work (job attitude). Meanwhile, external factors consist of job characteristics, occupational characteristics, and organizational characteristics. The purpose of this study was to determine the relationship between the demographic characteristics of workers (age, gender, marital status, education, years of service), mental workload, and leadership style on burnout. This research is a quantitative correlation study using a cross sectional design. The population of this study is the administrative staffs of Faculty of Public Health Diponegoro University with the sampling technique used is total sampling and the data analysis used is univariate and bivariate analysis. The result of this study is that there are no significant relationship between age, gender, marital status, education, years of service, mental workload, and leadership style with burnout.

Design analysis and Work load characteristics of a Micro data Center

Data centers are large-scale data storage and processing systems. It is made up of a number of servers that must be capable of handling large amount of data. As a result, data centers generate a significant quantity of heat, which must be cooled and kept at an optimal temperature to avoid overheating. To address this problem, thermal analysis of the data center is carried out using numerical methods. The CFD model consists of a micro data center, where conjugate heat transfer effects are studied. A micro data center consists of servers aligned with air gaps alternatively and cooling air is passed between the air gaps to remove heat. In the present work, the design of data center rack is made in such a way that the cold air is in close proximity to servers. The temperature and airflow in the data center are estimated using the model. The air gap is optimally designed for the cooling unit. Temperature distribution of various load configurations is studied. The objective of the study is to find a favorable loading configuration of the micro data center for various loads and effectiveness of distribution of load among the servers.

Detection method of tunnel lining voids based on guided anchoring mechanism

In tunnel construction engineering, the form of tunnel void diseases are complex and easily affected by the geographical environment. The traditional manual interpretation of image data has the characteristics of heavy workload, high probability of missing, and misjudgment. This paper constructs a convolution neural network that integrates the mechanism of guiding anchoring to detect tunnel voids. The network is composed of four parts: Feature extraction network extracts disease features from the enriched samples; Region proposal by guided anchoring join the generalized intersection over union (GIoU) evaluation criteria, and predict the shape of the anchor point through learning; The obtained feature maps are fixed in the region of interest pooling; Finally, the disease features are classified and bounding box regression. Compared with the existing target detection algorithm, the experimental results show that the improved network achieves an average classification accuracy of 92.74%, and the trained model has good generalization ability and robustness.