• Journal of IKEEE
• /
• v.25 no.3
• /
• pp.473-478
• /
• 2021

Although CNTFETs have attracted great attention due to their ability to increase semiconductor device performance by about 13 times, the commercialization of CNTFETs has been challenging because of the immature deposition process of CNTs. To overcome these difficulties, circuit design method considering the known limitations of the CNTFET manufacturing process is receiving increasing attention. SRAM is a major element constituting microprocessor and is regularly and repeatedly positioned in the cache memory; so, it has the advantage that CNTs can be more easily and densely deposited in SRAM than other circuit blocks. In order to take these advantages, this paper presents a circuit design method for SRAM cells considering CNT density and then evaluates its performance improvement using HSPICE simulation. As a result of simulation, it is found that when CNTFET is applied to SRAM, the gate width can be reduced by about 1.7 times and the read speed also can be improved by about 2 times when the CNT density was increased in the same gate width.

• Journal of IKEEE
• /
• v.25 no.4
• /
• pp.664-671
• /
• 2021

This paper proposes a GPGPU structure that can reduce the number of operations and memory access by effectively applying a data reuse method to a convolutional neural network(CNN). Convolution is a two-dimensional operation using kernel and input data, and the operation is performed by sliding the kernel. In this case, a reuse method using an internal register is proposed instead of loading kernel from a cache memory until the convolution operation is completed. The serial operation method was applied to the convolution to increase the effect of data reuse by using the principle of GPGPU in which instructions are executed by the SIMT method. In this paper, for register-based data reuse, the kernel was fixed at 4×4 and GPGPU was designed considering the warp size and register bank to effectively support it. To verify the performance of the designed GPGPU on the CNN, we implemented it as an FPGA and then ran LeNet and measured the performance on AlexNet by comparison using TensorFlow. As a result of the measurement, 1-iteration learning speed based on AlexNet is 0.468sec and the inference speed is 0.135sec.

• Journal of IKEEE
• /
• v.28 no.3
• /
• pp.337-341
• /
• 2024

As the production process for silicon-based integrated circuits approaches physical limits, a lot of attention is focused on the new semiconductor materials to overcome these problems. Carbon NanoTubes(CNTs) are attracting a lot of interest as one of the most competitive materials with excellent electrical transport and scaling properties, and CNTFETs using CNTs are gaining popularity as next-generation semiconductor devices. However, since the technology to place CNTs in a certain direction and interval on the wafer is not yet mature enough, it is difficult to construct all necessary circuits with CNTFET only. So, there is increasing interest in a hybrid configuration using MOSFET and CNTFET together. Because SRAM plays a role as a cache in microprocessors and is a critical circuit block influencing microprocessor performance, research to implement existing SRAM in a hybrid form is steadily progressing. Therefore, in this paper, we will explain the design method of hybrid 8T SRAM based on the existing hybrid 6T SRAM and discuss the performance difference between the two circuits.

• Journal of KIISE:Computer Systems and Theory
• /
• v.34 no.9
• /
• pp.407-415
• /
• 2007

In parallel cluster computing systems, the efficiency of communication between computing nodes is one of important factors that decide overall system performance. Accordingly, many researchers have studied on high-performance inter-node communication. The recently launched multi-core processor, however. increases the importance of intra-node communication as well because the more the number of cores in a node, the more the number of parallel processes running in the same node. Though there have been studies on intra-node communications, these have limited considerations on the state-of-the-art systems. In this paper, we propose a Linux kernel module that minimizes the number of data copy by exploiting the memory mapping mechanism for high-performance intra-node communication. The proposed kernel module supports the Linux kernel version 2.6. The performance measurements over a multi-core system present that the proposed kernel module can achieve lower latency up to 62% and higher throughput up to 144% than an existing kernel module approach. In addition, the measurements reveal that the performance of intra-node communication can vary significantly based on whether the cores that run the communication processes are belong to the same processor package (i.e., sharing the L2 cache).

• The KIPS Transactions:PartD
• /
• v.19D no.2
• /
• pp.161-186
• /
• 2012

Recently, the flash memory consistently increases the storage capacity while the price of the memory is being cheap. This makes the mass storage SSD(Solid State Drive) popular. The flash memory, however, has a lot of defects. In order that these defects should be complimented, it is needed to use the FTL(Flash Translation Layer) as a special layer. To operate restrictions of the hardware efficiently, the FTL that is essential to work plays a role of transferring from the logical sector number of file systems to the physical sector number of the flash memory. Especially, the poor performance is attributed to Erase-Before-Write among the flash memory's restrictions, and even if there are lots of studies based on the log block, a few problems still exists in order for the mass storage flash memory to be operated. If the FAST based on Log Block-Based Flash often is generated in the wide locality causing the random writing, the merge operation will be occur as the sectors is not used in the data block. In other words, the block thrashing which is not effective occurs and then, the flash memory's performance get worse. If the log-block makes the overwriting caused, the log-block is executed like a cache and this technique contributes to developing the flash memory performance improvement. This study for the improvement of the random writing demonstrates that the log block is operated like not only the cache but also the entire flash memory so that the merge operation and the erase operation are diminished as there are a distinct mapping table called as the offset mapping table for the operation. The new FTL is to be defined as the XAST(extensively-Associative Sector Translation). The XAST manages the offset mapping table with efficiency based on the spatial locality and temporal locality.

• Journal of Internet Computing and Services
• /
• v.6 no.2
• /
• pp.69-84
• /
• 2005

As the Internet users are increasing rapidly, a cluster web server system is attracted by many researchers and Internet service providers. The cluster web server has been developed to efficiently support a larger number of users as well as to provide high scalable and available system. In order to provide the high performance in the cluster web server, efficient load distribution is important, and recently many content-aware request distribution techniques have been proposed. In this paper, we propose a new content-aware load balancing technique that can evenly distribute the workload to each node in the cluster web server. The proposed technique is based on the hash histogram transformation, in which each URL entry of the web log file is hashed, and the access frequency and file size are accumulated as a histogram. Each user request is assigned into a node by mapping of (hashed value-server node) in the histogram transformation. In the proposed technique, the histogram is updated periodically and then the even distribution of user requests can be maintained continuously. In addition to the load balancing, our technique can exploit the cache effect to improve the performance. The simulation results show that the performance of our technique is quite better than that of the traditional round-robin method and we can improve the performance more than $10\%$ compared with the existing workload-aware load balancing(WARD) method.

• Journal of KIISE:Databases
• /
• v.32 no.1
• /
• pp.56-70
• /
• 2005

The results of location-dependent queries(LDQ) generally depend on the current locations of query issuers. Many mechanisms, e.g. broadcast scheme, hoarding, or racking policy, have been developed to improve the system peformance and provide better services, which are specialized for LDQs. Considering geographical adjacency of data and characteristics oi target area, caching policy and broadcast scheme affect the overall performance in LDQ. For this reason, we propose both the caching policy and broadcast scheme, which these features are reflected in. Based on the adjacency of data in LDQ, our broadcast scheme use Hilbert curve to cluster data. Moreover, in order to develop the caching policy suitable for LDQ on urban area, we apply the moving distance of a MH(Mobile Host) to our caching policy We evaluate the performance of the caching policy measuring the workload of MHs and the correctness of LDQ results and the performance of the broadcast scheme measuring the average setup-time of MHs in our experiments. Finally, we expect that our caching policy Provides more correct answers when executing LDQ in focal cache and leads significant improvement of the performance of MHs. It also seems quite probable that our broadcast scheme leads improvement of battery life of the MH.

• KIPS Transactions on Computer and Communication Systems
• /
• v.8 no.6
• /
• pp.127-138
• /
• 2019

The data response speed is a critical issue of cloud services because it directly related to the user experience. As such, the in-memory database is widely adopted in many cloud-based applications for achieving fast data response. However, the current implementation of the in-memory database is mostly based on the linked list-based hash table which cannot guarantee the constant data response time. Thus, cuckoo hashing was introduced as an alternative solution, however, there is a disadvantage that only half of the allocated memory can be used for storing data. Subsequently, bucketized cuckoo hashing (BCH) improved the performance of cuckoo hashing in terms of memory efficiency but still cannot overcome the limitation that the insert overhead. In this paper, we propose a data management solution called Wall Cuckoo which aims to improve not only the insert performance but also lookup performance of BCH. The key idea of Wall Cuckoo is that separates the data among a bucket according to the different hash function be used. By doing so, the searching range among the bucket is narrowed down, thereby the amount of slot accesses required for the data lookup can be reduced. At the same time, the insert performance will be improved because the insert is following up the operation of the lookup. According to analysis, the expected value of slot access required for our Wall Cuckoo is less than that of BCH. We conducted experiments to show that Wall Cuckoo outperforms the BCH and Sorting Cuckoo in terms of the amount of slot access in lookup and insert operations and in different load factor (i.e., 10%-95%).

• Journal of Korea Multimedia Society
• /
• v.16 no.8
• /
• pp.954-961
• /
• 2013

To enhance the efficiency of network, content-centric networking (CCN), one of future Internet architectures, allows network nodes to temporally cache transmitted contents and then to directly respond to request messages which are relevant to previously cached contents. Also, since CCN uses a hierarchical content-name, not a host identity like source/destination IP address, for request/response packet routing and CCN request message does not include requester's information for privacy protection, contents-providers/ network nodes can not identify practical requesters sending request messages. So to send back relevant contents, network nodes in CCN records both a request message and its incoming interfaces on Pending Interest Table (PIT). Then the devices refer PIT to return back a response message. If PIT is exhausted, the device can not normally handle request/response messages anymore. Hence, it is needed to detect/react attack to exhaust PIT. Hence, in this paper, we propose improved detection/reaction schemes against attacks to exhaust PIT. In practice, for fine-grained control, this proposal is applied to each incoming interface. Also, we propose the message framework to control attack traffic and evaluate the performance of our proposal.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.7 no.10
• /
• pp.2430-2447
• /
• 2013

In recent years, HTTP adaptive streaming (HAS) has attracted considerable attention as the state-of-the-art technology for video transport. HAS dynamically adjusts the quality of video streaming according to the network bandwidth and device capability of users. Content-Centric Networking (CCN) has also emerged as a future Internet architecture, which is a novel communication paradigm that integrates content delivery as a native network primitive. These trends have led to the new research issue of harmonizing HAS with the in-network caching provided by CCN routers. Previous research has shown that the performance of HAS can be improved by using the H.264/SVC(scalable video codec) in the in-network caching environments. However, the previous study did not address the misbehavior that causes video freeze when overestimating the available network bandwidth, which is attributable to the high cache hit rate. Thus, we propose a new SVC-based adaptation algorithm that utilizes a drop timer. Our approach aims to stop the downloading of additional enhancement layers that are not cached in the local CCN routers in a timely manner, thereby preventing excessive consumption of the video buffer. We implemented our algorithm in the SVC-HAS client and deployed a testbed that could run Smooth-Streaming, which is one of the most popular HAS solutions, over CCNx, which is the reference implementation of CCN. Our experimental results showed that the proposed scheme (SLA) could avoid video freeze in an effective manner, but without reducing the high hit rate on the CCN routers or affecting the high video quality on the SVC-HAS client.