Development of personalized mobile information service platform of university library based on multi-core processor and cloud computing platform

doi:10.21203/rs.3.rs-2702614/v1

Download PDF

Research Article

Development of personalized mobile information service platform of university library based on multi-core processor and cloud computing platform

https://doi.org/10.21203/rs.3.rs-2702614/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 23 May, 2023

Read the published version in Soft Computing →

You are reading this latest preprint version

In recent years, depending on information technology, knowledge and skills, university libraries have established a new "user centered" data service system, which is called personalized information service. Although this service has made some progress in recent years, these systems have only achieved a certain amount of personalized information services. Multi-core processor is the basic process unit of high-performance cloud computing node, which is an important basis to achieve system security, high performance, low power consumption and other high-performance requirements. Cloud computing node, as one of the core components in the high-performance computer system architecture, in addition to the acceleration node, master control unit, communication system, etc., its efficiency has constituted an important basic condition for the realization of the "high efficiency" goal of the system. The emergence of multi-core processors and cloud computing platforms has created conditions for the development of personalized mobile information service platforms in university libraries. Under the technical support of multi-core processors and cloud computing platforms, it is also an important goal of mobile libraries in the future to meet the needs of users to search library data information at any time and provide users with diversified data services. At present, the knowledge service platform of university library can not only actively select and recommend relevant materials for users according to their knowledge preferences, data utilization habits, and academic backgrounds, but also enable service users to quickly obtain the borrowing expiration prompt, independent consultation and retrieval and other personalized information functions, and carry out more in-depth personalized intelligent knowledge consulting services. Based on multi-core processor and cloud computing platform, this paper develops personalized mobile information service platform of university library.

Cloud computing technology can integrate software and hardware materials through virtualization technology to achieve more efficient computing technology [1–2]. Cloud computing also creates a pay as needed application mode for users, sharing computing power, storage and other virtual data on the cloud with users. This paper applies multi-core processor to the research of humanized mobile information service system of university library [3]. The personalized mobile information system of university library in this paper includes the automatic configuration system of operating system and application, and provides system authentication [4]. Personalized information system is also an important direction for the development of university mobile libraries. At present, the information content resources used by university mobile libraries have the characteristics of big data [5]. Therefore, based on the cloud computing platform, this paper has built a personalized information service system for university mobile libraries. Considering the information behavior of users, it collects users' information content resources through big data, and thus serves as the entrance to their personalized service information content, which can effectively improve the efficiency of personalized information services for university mobile libraries [6]. In order to verify the reliability of the automatic configuration mechanism and deployment system, this paper has carried out many performance tests and functional inspections [7]. The test results show that the personalized mobile information service system of the high library based on multi-core processors and cloud computing platform provided in this paper can help users of personalized mobile information service equipment of the high library to quickly and simply build an application environment [8]. It also reaches the high automation configuration level of application level and operation level, and has complete scalability and fault tolerance.

According to the literature, users will form more abundant big data information behavior data through their access, query, use and other big data behaviors on the information service platform of diversified activities such as university libraries, which contains not only structural content but also semi-structured and unstructured content [9]. The test content is not only large in information volume but also complex in types. At present, The information behavior data of the mobile information service platform of university library has met the four characteristics of "big data", namely, large amount of information, multiple types, high rate of information content increase, and high value of information data [10]. A processor technology Hydra integrating four MIPS83000 cores is presented in the literature. Hydra is a classic multi-core architecture, which uses the isomorphic and symmetric technology of sharing the second level cache, and also adds the multi-level communication technology of high-speed bus [11]. Since then, IBM, AMD, Intel and other large enterprises have released dual core, quad core and even higher core processors. The core system design of multi-core data processor mainly includes homogeneous structure and heterogeneous structure. Symmetrical design of homogeneous structure. All cores use the same architecture design [12]. Although it is simple in theory and easy for engineering design, it faces problems such as data consistency, stability impact on internal storage and I/O access, power consumption, and multi-core task coordination. However, most desktop multi-core processor technologies use homogeneous structure. The literature mainly describes several problems in the design and development process of cloud computing node system, as well as the hardware technology, including the overall architecture of the system and the design of interconnection with the processor, clock, power on and reset, memory, motherboard, and subsystems [13]. According to the literature, cloud computing is a new computing model, which provides computing power, data storage, applications and other resources to all users as public services through the Internet [14]. The practical application of cloud computing is mainly based on the idea that all internal capabilities are combined to be used by each member of the organization, who only needs to use these capabilities without studying them. The literature considers from the perspective of high-performance computing technology, not from the perspective of cloud computing technology [15]. In particular, there is no detailed conceptual model, detailed analysis and description of application technology on the logical architecture, interconnection mode, network technology and software of cloud computing node computers.

3.1 Multi core processor technology

Multi-core processors are also called on-chip multiprocessors. Homogeneous frameworks usually adopt symmetrical design. All cores use the same framework to design and work. The basic principle is very simple and easy to implement. However, they are faced with problems such as data consistency, characteristics of kernel storage and use, power consumption, and task coordination among multiple cores.

At present, the main desktop multi-core processor technologies all use homogeneous frameworks. Heterogeneous architecture can combine cores of different structures to maximize processor performance and effectively reduce power consumption. However, heterogeneous implementation difficulties also arise in multi-core task allocation, organizational scalability and quantity control, and multiprocessor instruction system settings.

Multi-core processors have brought new solutions to many applications. currently, multi-core processors have become the main part of asymmetric embedded operating systems as a way to improve the stability of processor technology. In multi-core processors, the overall system throughput is one of the most commonly used indicators to quantify the overall multi-core system, and reducing the cache loss rate is another important way to increase the overall system throughput. Taking the overall performance of the system as the main objective is to optimize the total number of commands executed by the whole system in unit time, which is called the maximum IPC.

As the overall performance of the system, IPS is defined as:

$${\text{I}\text{P}\text{S}}_{\text{t}\text{o}\text{t}\text{a}\text{l}}=\sum _{\text{i}=0}^{\text{N}-1}{\text{I}\text{P}\text{S}}_{\text{i}}$$

Including:

$${\text{I}\text{P}\text{S}}_{\text{i}}=\frac{{\text{I}\text{n}\text{s}\text{t}}_{\text{i}}}{\text{t}}$$

Program performance function, which determines the program performance function according to the cache space of the corresponding number of paths allocated by the program:

$${\text{U}}_{\text{a}}^{\text{b}}={\text{m}\text{i}\text{s}\text{s}}_{\text{a}}-{\text{m}\text{i}\text{s}\text{s}}_{\text{b}}$$

Improved system fairness: The fairness oriented cache resource allocation strategy is to ensure that every program executed simultaneously on the same chip has the same performance change, and the rate of performance degradation caused by contention for shared cache resources among applications is basically the same, and the program execution time is not delayed for a long time due to contention for cache resources. The running time of multithreads in Cache also defines leveling as:

$$\frac{{\text{T}\text{s}\text{h}\text{r}}_{\text{i}}}{{\text{T}\text{d}\text{e}\text{d}}_{\text{i}}}=\frac{{\text{T}\text{s}\text{h}\text{r}}_{2}}{{\text{T}\text{d}\text{e}\text{d}}_{2}}=\dots =\frac{{\text{T}\text{s}\text{h}\text{r}}_{\text{n}}}{{\text{T}\text{d}\text{e}\text{d}}_{\text{n}}}$$

On most multi-core processors, the core group usually contains two multi-core cores and a shared end cache. Therefore, in this system, the mapping relationship between the execution thread and the kernel can be found directly through the matching method of the above formula. However, the cache structure of some multi-core processors does not directly use the above matching methods to find the best mapping processing scheme. In order to solve this problem, this paper can decide the final conclusion by gradually merging nodes. This paper reconstructs the communication matrix, which is used to mark the communication volume between thread pairs and thread pairs, and then performs these calculations again using this matrix as input. A reconstructed calculation matrix marking the communication between thread pairs is established by using heuristic variables:

$${\text{H}}_{\left(\text{x},\text{y}\right),\left(\text{z},\text{k}\right)}={\text{M}}_{(\text{x},\text{z})}+{\text{M}}_{(\text{x},\text{k})}+{\text{M}}_{(\text{y},\text{z})}+{\text{M}}_{(\text{y},\text{k})}$$

Although this method can not determine whether an optimal mapping scheme can be obtained, it can obtain an optimal solution. At the same time, because its time complexity and space complexity are very low, it can more reasonably control the consumption of mapping between programs.

The design focus of asymmetric embedded operating system is to achieve different tasks through different cores, so as to give full play to the advantages of different cores or different operating systems.

When the system with 16 CPUs (64 cores) and the matrix order are increased from 10000 to 48000, the performance of Linpack floating-point operation is shown in Fig. 1. When 64 cores are used, the performance reaches 44.5 Gflop.

3.2 Cloud computing platform

Different distributed computing architectures can be generated according to various business needs, and different cloud computing models are formed accordingly. At present, the main computing platforms include cloud platforms based on MapRednce computing architecture, event driven cloud platforms, and highly flexible cloud platforms. However, these cloud computing platforms face great difficulties in implementing a scheduling computing that can take full advantage of the advantages of clusters. MapReduce applies the idea of "divide and conquer" and comes to the final conclusion after calculation.

MapReduce cloud service has attracted extensive attention from academia and industry because of its advantages in security and simplicity, and has conducted in-depth research in optimizing MapReduce's computing architecture, management, task scheduling and other fields. Improving cluster efficiency has always been the goal of everyone, but data localization is the key factor affecting the efficiency. For tasks without data locality, information must be read or input remotely to extend the task execution period. Therefore, the fewer tasks without data locality, the better the overall performance.

MapReduce computing architecture on HDFS. Closely integrate HDFS and MapReduce architecture to store files for MapReduce. If the last completion time of MapReduce task is j, the task will be completed in tj time in the kth scheduling. Then the completion time of MapReduce task can be described as:

$$\text{T}=\left(\text{k}-1\right)\text{*}\varDelta \text{T}+{\text{t}}_{\text{j}}$$

Set the average execution time of task 1 as melon, as shown in formula (7):

$${\text{T}}_{\text{i}}={\text{T}}_{\text{c}\text{p}\text{u}}\left(\text{i}\right)+{\text{T}}_{\text{c}\text{o}\text{m}\text{m}}\left(\text{i}\right)$$

Set the time cost of data block transmission as Ttran, as shown in Formula (8):

$${\text{T}}_{\text{c}\text{o}\text{m}\text{m}}=\text{d}+\left(1-{\lambda }\right)\text{*}{\text{T}}_{\text{t}\text{r}\text{a}\text{n}}$$

Where d is the communication cost when the task has data locality. Without losing generality, this paper combines d and CPU running time into Tcpu (i), as shown in the formula:

$${\text{T}}_{\text{i}}={\text{T}}_{\text{c}\text{p}\text{u}}\left(\text{i}\right)+\left(1-{\lambda }\right)\text{*}{\text{T}}_{\text{t}\text{r}\text{a}\text{n}}\left(\text{i}\right)$$

Domino is an event driven design structure written in JAVA language for iterative and incremental use. It is implemented using a plug-in of HBase. Domino uses set triggers to share information, and uses the operation functions of set triggers to implement application logic.

It can be seen from Fig. 2 that since Domino's performance has greatly exceeded MapReduceo's, based on Domino's high security and open source features, Domino is selected as the configuration target for the scheduling methods used in this article.

Assuming that the working time of each trigger is equal, it is recorded as △ T. The working time limit of the trigger is t (i, j), which can be expressed as:

$$\text{T}\left(\text{i},\text{j}\right)=\sum _{\text{k}=1}^{\text{a}(\text{i},\text{j})}{\Delta }\text{t}=\text{a}(\text{i},\text{j}){\Delta }\text{t}$$

Then the completion time of task scheduling step j is T_j, and the formula is:

$${\text{T}}_{\text{j}}=\text{m}\text{a}\text{x}\left\{\text{a}\right(\text{i},\text{j}\left)\right\}{\Delta }\text{t}$$

Then the complete execution time T of this group of trigger tasks can be expressed as:

$$\text{T}=\sum _{\text{j}=1}^{\text{k}}{\text{T}}_{\text{j}}=\sum _{\text{j}=1}^{\text{k}}\text{m}\text{a}\text{x}\left\{\text{a}\right(\text{i},\text{j}\left)\right\}{\Delta }\text{t}$$

Furthermore, when completing this set of trigger tasks, the cluster throughput T_P can be expressed by a formula.

$${\text{T}}_{\text{p}}=\frac{\sum _{\text{j}=1}^{\text{k}}{\text{T}}_{\text{p}}\left(\text{j}\right)}{\text{k}}$$

4.1 Development of personalized mobile information service of university library based on cloud computing platform

Cloud computing platform has laid a very important technical foundation for the realization of humanized mobile information services in libraries. Cloud computing node is the main core component in the high-performance computer system architecture, except for the acceleration node, master control unit, network, etc. The efficiency of computer nodes has become the key issue to achieve the overall goal of "high efficiency" of the whole machine.

First of all, some large high-performance computer systems in research and use are very concerned about the breakthrough of computing node efficiency. Secondly, in terms of data processor nodes, due to the emergence of multi-core processors in recent years, the main efficient computing nodes have generally used multi-channel multi-core technology. In the current technology, the main routes are 4 and 8, while the higher density routes are relatively lacking. Furthermore, in terms of the selection of data processors, the current node computers of high efficiency cloud computing technology mainly choose foreign data processors.

Table 1

Comparison of cloud computing node machines
Cloud computing node machine	Number and type of CPUs	CPU clock frequency	Node power consumption	Node volume	Multi core CPU
Cloud computing node with multi-core CPU	16 way, 4-core Loongson 3A	1.0GHz	About 300W	1 U Rack	Y
IBM Blnde JS20 Server	2 way, 1BM PowerPC970	2.2GHz	About 395W	About 1/2 U Blade	N
IBM BladeCenter JS22 Experiences server	4-way, POWER6	4.0GHz	About 350W	About 1/2 U Blade	N
TYAN GT62B8230-L server	2-way AMD 12-Core Opteron6100	2.1GHz	About 350W	1 U Rack	N

In general, the more new historical browsing tags are used, the more attention they will have to the document. Therefore, when using historical browsing tags to judge the "attention" of browsing documents, different link weights should be given to each historical browsing data according to its time stage.

The number of time cycles in the past is represented by T, and FILE represents the number of requested views of file f in the T time cycle. Then the "heat" of document f can be calculated:

$$\text{P}\text{o}\text{p}\text{u}\text{l}\text{a}\text{r}\text{i}\text{t}\text{y}\left(\text{f}\right)=\sum _{\text{t}=1}^{\text{T}}({2}^{(\text{t}-\text{T})}\times {\text{A}}_{\text{f}}^{\text{t}})$$

Then the popular file p that needs to create a copy is:

$$\text{p}=\text{m}\text{a}\text{x}\left[\text{P}\text{o}\text{p}\text{u}\text{l}\text{a}\text{r}\text{i}\text{t}\text{y}\right(\text{f}\left)\right]$$

Next, you can use the average browsing "heat" of the relative popular file p and the average "heat" of all documents in the collection FILE to determine the number of copies made by all popular files p.

The average heat of all files in the collection FILE is:

$$\stackrel{-}{\text{P}\text{o}\text{p}\text{u}\text{l}\text{a}\text{r}\text{i}\text{t}\text{y}\left(\text{f}\right)}=\frac{\sum _{\text{f}\in \text{F}\text{I}\text{L}\text{E}}\text{P}\text{o}\text{p}\text{u}\text{l}\text{a}\text{r}\text{i}\text{t}\text{y}\left(\text{f}\right)}{\left|\text{F}\text{I}\text{L}\text{E}\right|}$$

Where, | FILE | represents the number of files in the collection FILE, the number of copies to be created for popular FILE p is:

$${\text{N}}^{\text{p}}=\left[\frac{\text{P}\text{o}\text{p}\text{u}\text{l}\text{a}\text{r}\text{i}\text{t}\text{y}\left(\text{p}\right)}{\stackrel{-}{\text{P}\text{o}\text{p}\text{u}\text{l}\text{a}\text{r}\text{i}\text{t}\text{y}\left(\text{f}\right)}}\right]$$

The number of times P is requested and used in different regions is also different, that is, the contribution to its "popularity" is also different in different regions. By collecting relevant data sent from different nodes, different summary points control the historical use of all data files. Therefore, according to the historical use data of p controlled by the summary point, the "heat" of p use in different regions can be calculated separately:

$${\text{P}\text{o}\text{p}\text{u}\text{l}\text{a}\text{r}\text{i}\text{t}\text{y}}_{\text{r}}\left(\text{p}\right)=\sum _{\text{t}=1}^{\text{T}}\left({2}^{\left(\text{t}-\text{T}\right)}\times {\text{A}}_{\text{p}}^{\text{t}}\right),\text{r}=\text{1,2},\dots ,\text{R}$$

In the case of meeting the constraints such as the QoS standard of the task scheduling system, the cloud service provider's fees are mainly computing costs. Assuming that the application domain target ti will be allocated to the computing resource ci and the matrix element xi is configured, the computing cost to be paid by the node cj to achieve the information scheduling function of the operation target ti is:

$$\text{C}\text{o}\text{m}\text{p}\_{\text{c}\text{o}\text{s}\text{t}}_{\text{j}}^{\text{i}}=\text{I}\text{n}\text{s}\text{t}\text{r}\text{u}\_{\text{c}\text{o}\text{s}\text{t}}_{\text{j}}\times \text{I}\text{n}\text{s}\text{t}\text{r}\text{u}\_{\text{c}\text{o}\text{u}\text{n}\text{t}}_{\text{i}}$$

The overhead cost and the number of unit instructions contained in the application task t. Compared with the above calculation overhead cost, the service revenue that the node cj can obtain by completing the scheduling request of the application task ti is:

$${\text{R}\text{e}\text{v}\text{e}\text{n}\text{u}\text{e}}_{\text{j}}^{\text{i}}={\text{B}\text{u}\text{d}\text{g}\text{e}\text{t}}_{\text{i}}-{\text{F}\text{i}\text{n}\text{e}}_{\text{j}}^{\text{i}}$$

Further, the total service revenue that the entire cloud computing system can obtain by completing the scheduling operation of all tasks in the application task set to be scheduled is:

$$\text{R}\text{e}\text{v}\text{e}\text{n}\text{u}\text{e}=\sum _{\text{j}=1}^{\text{n}}\sum _{\text{i}=1}^{\text{m}}({\text{R}\text{e}\text{v}\text{e}\text{n}\text{u}\text{e}}_{\text{j}}^{\text{i}}\times {\text{x}}_{\text{i},\text{j}})$$

Cloud computing system is a commercial public service. On the premise of meeting the constraint requirements of task scheduling goals, it will spend as little as possible to obtain as much business profits as possible; Compared with fixed computing costs, the more total service profits cloud services expect to obtain, the better.

In recent years, some fine-grained resource scheduling strategies have also emerged on the MapReduce platform. Among them, the Through put Scheduler is designed for heterogeneous MapReduce clusters. It uses machine learning methods to predict the resource requirements of tasks and the computing capacity of MapReduce cluster nodes, and then formulates the optimal task allocation strategy. Polo adjusts the number of task slot slots of MapReduce cluster nodes according to job resource information, and formulates task allocation strategies to maximize the utilization of cluster resources. However, these two scheduling algorithms are still aimed at slot one-dimensional resource scheduling.

The delay scheduling strategy is also applied to the cloud platform of elastic computing, which also uses a distributed file system, such as HDFS, but also faces the problem of data locality. This method uses the delay configuration information for local targets without data to prevent remote reading of targets.

The Genexalized Fairness on Jobs policy fairness measure is based on the number of jobs allocated to each user.

Table 2

Comparison of Typical Scheduling Algorithms
	FIFO	Capacity	Fair	DRF	Asset	CEEI
Design idea	Submission time sorting	Resource occupancy	Single resource equity	Equity of main resources	Equity of total resources	Maximize resource consumption
Multidimensional resources	no	no	no	yes	yes	yes
Big homework	Incremental mechanism	Incremental mechanism	no	no	no	no

The existing typical scheduling algorithms are shown in Table 2.

4.2 Development of personalized mobile information service of university library based on multi-core processor

Interact with the school educational administration system, and submit relevant information about the current learning content to the user at the appropriate time through the school curriculum. The functions of the personalized mobile information service system of university library mainly include: (1) converting the user's information behavior statistics into the user's information content needs; (2) Collect user information behavior statistics; (3) Generate recommendation information and submit information content service through calculation. With the help of multi-core processors and cloud computing platforms, resources can be integrated. As shown in Table 3, in multi-core processors, each data processor has a buffer that directly corresponds to each other, but they have two conflicting parameters: efficiency and volume. If we only consider the overall hit time of the processor, we must reduce the amount of cache space to reduce access costs; However, if we fully consider increasing the amount of cache space, we must weaken the hit time, which also leads to the emergence of multi-level cache design.

Table 3

AMD X4 processor cache hit time
Cache Level	size	Degree of association	Hit Time
L1	64KB	2-way	3 Clock cycles
L2	512KB	16-way	≥ 9 clock cycles
L3	2MB	32-way	About 38 clock cycles

Although the performance of each level of cache is similar, its structure and performance are different. Generally speaking, all the information in the high-level cache structure includes the low-level cache structure, while there are many differences in the multi-level cache structure. Because the multi-level cache structure should consider the characteristics of multi-level exclusion, and all the information in the high and low caches should not intersect with each other. Therefore, if an error is requested in the high-level cache but hit in the low-level cache, the information in the high and low caches can be exchanged with each other. The advantage of implementing multi-level exclusion is that you can give the low-level cache as much information as possible to save space.

For the detection of quick sort algorithm, unordered data blocks with different information scales are generally used. From the running results, it can be found that the most time-consuming functions generally include the information input module, quick sort algorithm recursion and interval analysis module. The information input module generally includes the reading process of a large number of data files, which has nothing to do with the quick sort algorithm itself, Therefore, the function analysis focuses on the quick sort module and the region analysis module.

As shown in Fig. 3, by analyzing the data of the same model on different data scales, it can be seen that the execution time of the quick sort algorithm accounts for 5% of the total execution time of the calculation, while the execution time of the interval partition model accounts for 90% of the total execution time of the calculation. Through this comparison, it is not difficult to see that the call relationship between the modules in the quick sort algorithm is relatively stable, The bottleneck problem of the algorithm is generally focused on the interval division module and recursive quick sorting module.

Table 4

Multi core synchronization effect test
	Nuclear 0	Nuclear 1	Nuclear 2	Nuclear 3	Nuclear 4	Nuclear 5	Nuclear 6	Nuclear 7
Number of instruction cycles after 50 synchronizations of the first type	640355	640640	640679	640363	640787	640853	640398	640166
Number of instruction cycles after 50 synchronizations of the second type	653340	638002	653439	653432	653440	653209	637654	653120
Number of instruction cycles after 50 synchronizations of the third type	653319	653239	639000	653702	653290	653118	653820	653833

According to Table 4, the multi-core communication network with AMP structure can meet the requirements of millisecond data transmission for transient interference protection. When multiple cores are the same, there are three options: one is to get the semaphore before the algorithm module, and the algorithm module then releases the semaphore to achieve synchronization; Second, it does not need semaphores at all, but synchronizes by setting flag bits in shared memory; Third, when the algorithm module is completed, you can use only the semaphore obtained or only the semaphore released to judge, thus realizing step.

With the support of multi-core processors, the personalized mobile information service platform of university library on cloud computing platform transmits text faster. The multi-core processor can also use the service number to provide the template mode of fast push information notification. On the premise of user approval, the multi-core processor can actively send the user information such as the activity information that the user is borrowing books, the information that the books are about to expire, and the information about the emergency closure of the library. The active push message service mode will provide greater convenience for users.

The personalized mobile information service platform of university library using multi-core processor and cloud computing platform has laid a reliable foundation for the construction of information resources of the library, and is also the premise for the library to provide personalized information services to users. This paper makes use of the high utilization rate of cache resources in the multi-core processor platform to virtualize the personalized mobile information service desk of university library into a large shared cache that can be used by all processor cores to promote the content dissemination in the library. Based on cloud computing technology, the user's behavior is analyzed with big data, so as to master its real information resources, and take this as input to improve the effectiveness of the service system, which can more effectively improve the efficiency of personalized service of university libraries, and help users to achieve more efficient use of library information resources while effectively obtaining the information resources they need.

Conflict of interest

The authors declare that they have no conflict of interests

Ethical approval

This article does not contain any studies with human participants performed by any of the authors.

Data Availability

Data will be made available on request.

Shackleford JA, Kandasamy N, Sharp GC (2010) On developing B-spline registration algorithms for multi-core processors. Phys Med Biol 55(21):6329
Attaran M, Woods J (2019) Cloud computing technology: improving small business performance using the Internet. J Small Bus Entrepreneurship 31(6):495–519
Shackleford JA, Kandasamy N, Sharp GC (2010) On developing B-spline registration algorithms for multi-core processors. Phys Med Biol 55(21):6329
Choi O, Kim K, Wang D et al (2012) Personalized mobile information retrieval system. Int J Adv Rob Syst 9(1):11
Tsvetkov VY, Matchin VT (2014) Information conversion into information resources. Eur J Technol Des 4(2):92
Zhao Y, Deng S, Gao T, Zhou R (2016) Research on user needs for mobile information services in Chinese university libraries: Comparison between existing user and potential users. Electron Libr 34(4):617–635
Bunch C, Drawert B, Chohan N et al (2012) Language and runtime support for automatic configuration and deployment of scientific computing software over cloud fabrics. J Grid Comput 10(1):23–46
Yu H, Dan Y, Jing L, Mu Z (2013) Research on Personalized Recommender System for Tourism Information Service. Comput Eng Intell Syst 4:32–45
Cai H, Da Xu L, Xu B et al (2014) IoT-based configurable information service platform for product lifecycle management. IEEE Trans Industr Inf 10(2):1558–1567
Nikolova HD, Inman JJ (2015) Healthy choice: the effect of simplified point-of-sale nutritional information on consumer food choice behavior. J Mark Res 52(6):817–835
Pricopi M, Mitra T, “Bahurupi (2012) A polymorphic heterogeneous multi-core architecture. ACM Trans Archit Code Optim (TACO) 8(4):1–21
Jongerius R, Anghel A, Dittmann G et al (2017) Analytic multi-core processor model for fast design-space exploration. IEEE Trans Comput 67(6):755–770
Patel A, Taghavi M, Bakhtiyari K, Júnior JC (2013) An intrusion detection and prevention system in cloud computing: A systematic review. J Netw Comput Appl 36(1):25–41
Wei Y, Blake MB (2010) Service-oriented computing and cloud computing: Challenges and opportunities. IEEE Internet Comput 14(6):72–75
Ercan T (2010) Effective use of cloud computing in educational institutions. Procedia-Social and Behavioral Sciences 2(2):938–942

Download PDF

Journal Publication

published 23 May, 2023

Read the published version in Soft Computing →

Reviewers agreed at journal
01 Apr, 2023
Editor assigned by journal
18 Mar, 2023
First submitted to journal
16 Mar, 2023

You are reading this latest preprint version

Development of personalized mobile information service platform of university library based on multi-core processor and cloud computing platform

Status:

Journal Publication

Version 1

Abstract

Figures

1. Introduction

2. Related Work

3. Multi Core Processor And Cloud Computing Platform

3.1 Multi core processor technology

3.2 Cloud computing platform

4. Platform Development

4.1 Development of personalized mobile information service of university library based on cloud computing platform

4.2 Development of personalized mobile information service of university library based on multi-core processor

5. Conclusion

Declarations

References

Status:

Journal Publication

Version 1