CXL memory disaggregation
Researchers at the Korea Advanced Institute of Science and Technology (KAIST) have developed a Compute Express Link (CXL) solution for directly accessible, high-performance memory disaggregation that they believe will improve performance compared to existing Remote Direct Memory Access (RDMA)-based memory significantly improves disaggregation.
RDMA allows one host to directly access another host’s storage over the data center’s InfiniBand network protocol. However, the researchers pointed out that this method requires an additional CPU and causes longer access latencies due to redundant data copies and software fabric interventions for RDMA-based memory disaggregation.
The new CXL-based storage disaggregation framework includes CXL-enabled custom CPUs, CXL devices, CXL switches, and CXL-enabled OS modules. The team’s CXL device is a purely passive and directly accessible memory node that contains multiple DRAM DIMMs and a CXL memory controller. Because the CXL storage controller supports the storage in the CXL device, a host can use the storage node without processor or software intervention.
The CXL switch enables scaling of a host’s storage capacity by hierarchically connecting multiple CXL devices to the CXL switch, allowing for more than hundreds of devices. On the switches and devices, the CXL-enabled operating system removes the redundant data copy and protocol conversion that traditional RDMA exhibits, which can significantly reduce access latency to the storage nodes.
“Our CXL-based memory disaggregation framework breaks away from traditional RDMA-based memory disaggregation and can provide high scalability and performance for various data centers and cloud service infrastructures,” said Myoungsoo Jung, Professor at KAIST’s Computer Architecture and Memory Systems Laboratory .
In a test comparing 64B data (cacheline) loading from memory pooling devices, CXL-based memory disaggregation showed 8.2 times higher data loading performance than RDMA-based memory disaggregation and even a similar performance as the local DRAM memory. In the team’s evaluations for a big data benchmark such as a machine learning-based test, the CXL-based memory disaggregation technology also showed a maximum of 3.7 times better performance than previous RDMA-based memory disaggregation technologies.
“Our CXL-based storage disaggregation research will yield a new paradigm for storage solutions that will spearhead the era of big data,” added Jung.
Optoelectronic logic gates
Researchers from Korea Institute of Science and Technology (KIST) and Gwangju Institute of Science and Technology developed ultra-fast, high-efficiency optoelectronic logic gates (OELGs) using organic-inorganic perovskite photodiodes.
The optoelectronic logic gate uses light as an input signal, resulting in low energy loss and can only work on light energy without electrical power supply. In the device, two layers of perovskite thin films are stacked vertically like a sandwich. A binary logic operation is possible by inputting two light beams of different wavelengths and intensities.
The optoelectronic perovskite logic gate can freely change the photocurrent polarity using light, making it possible to perform more than one logic gate operation result for the same input value.
This means that compared to the existing logic gate which can only perform one logical operation on one device, the newly developed one can implement all five different basic logic operations: AND, OR, NAND, NOR and NOT. The team said that since one logic gate can function like five logic gates, this enables the development of optical processors with high spatial efficiency and integration.
“Perovskite optoelectronic logic gates, which perform multiple logic operations in response to optical inputs, are expected to be used for ultra-small and low-power general-purpose optical sensor platforms in the future,” said Yusin Pak of the Sensor System Research Center at KIST. Researchers expect it will be useful for applications such as next-generation optical communications, optical networking and healthcare.
Purely optical random bit generator
Researchers from Taiyuan University of Technology, Guangdong University of Technology, Northwestern Polytechnical University, Institute of Southwestern Communication and Bangor University propose a method for real-time physical random bit generation by combining broadband photonic entropy sources with all-optical signal processing techniques.
Researchers found that optical chaos is a reliable way to generate fast and real-time random bits due to its high bandwidth and large amplitude fluctuations. However, most optical chaos-based random bit generators do their quantization in the electrical domain using electrical analog-to-digital converters, creating a bottleneck.
In the team’s all-optical random bit generation method, chaotic pulses in the optical domain are quantized into a physical random bit stream using a length of highly nonlinear fiber. In the proof-of-concept experiment, they successfully generated a random 10 Gb/s bitstream in a single channel.
The team notes that the current 10Gb/s rate time is only limited by the assumed chaos bandwidth. Your scheme can potentially operate at rates much higher than 100 Gb/s if the bandwidth of the chaotic entropy source is sufficient.
Jesse Allen is the Knowledge Center Administrator and Senior Editor at Semiconductor Engineering.
#Research #Bits #July #18th