Shankar Malik’s Post

Come sit in on our 4 Tech Talks at this years #OCP2024 and let's schedule a meeting to define how #ScaleFlux can efficiently and reliably #Accelerate and provide better #TCO to your #Enterprise, #DataCenter, #AI environments

#Arasan’s eMMC5.1 PHY is a fully compliant #PHY layer for #JEDEC eMMC5.1 and #eMMC5.1 JESD84-B50 specification.  https://lnkd.in/gUDbzng

There are two types of interface between user-space and kernel-space in the WiFi AP system. WEXT(Wireless Extensions) and NL80211/CFG80211. WEXT is ioctl based operations on a network socket file descriptor, iwpriv, iwlist and iwconfig as the representative tools. NL80211/CFG80211 subscribe to some network-related notifications from the kernel using netlink. Support the multicast communications between the kernel and multiple user-space processes. CFG80211 is a thin layer between user space and driver. iw, cfg80211tool and wifistats as the representative tools Qualcomm migrates all WEXT based configuration tools like iwpriv, iwconfig, proprietary applications to cfg80211 based systems, starting with QC WLAN driver 11.0. #WiFi #CLI #CFG80211 #WEXT

The JCS UltraBlock™ maintains product integrity and sterility on either side of the barrier between the UHT and the Aseptic Surge Tank. This maximizes production runs, ensures the separation of incompatible products, coordinates all functions with the filler, and can be fed to multiple destinations. #beverageprocessing #jcsprocess #jcsultrablock https://hubs.la/Q02pb1j20

Product Update: EHG9512 Series 12-Port IEC 61850-3 Certified Managed Gigabit Switch : Why us? ATOP Technologies, Inc. continues to innovate with the introduction of the EHG9512 Series, setting new standards in network management and reliability. Designed to meet the rigorous demands of industrial environments, ATOP's EHG9512 Series offers unparalleled performance, reliability, and security. Features: IEC61850-3 Certified: Ensures reliability and performance in critical industrial applications. Gigabit Switch: 8 x 10/100/1000 BASE-T(X) ports and 4 x 1000 BASE-X SFP ports for high-speed data transmission. IEEE1588v2 Precision Time Protocol Support: Achieve nanosecond-accurate time synchronization for precise operations. Hardware-based E2E Transparent Clock and Software Boundary Clock: Ensures accurate timekeeping and synchronization across the network. Redundancy Support: ERPS and Compatible Ring ensure rapid recovery (<20ms @ 40 switches), alongside STP/RSTP/MSTP/MRP (Client) for network redundancy. Remote Management: Access and manage the switch remotely via Web browser, Telnet, Serial console, or Utility, providing flexibility and convenience. Security Features: Compliant with IEC62443-4-2, offering robust security measures to safeguard critical infrastructure. Specifications: Priority Queues: 8 VLAN Table: 4096 MAC-Based VLAN: 512 Trunk Group: 4 MAC Table Size: 16K Packet Buffer Size: 1.5 MB Jumbo Frame: 9216 Byte Ethernet Standards: IEEE 802.3 for 10BASE-T, IEEE 802.3u for 100BASE-T(X), IEEE 802.3ab for 1000BASE-T, IEEE 802.3z for 1000BASE-X, and more. Applications: Industrial Automation: Ideal for industrial automation applications requiring high-speed, reliable network communication. Power Substations: Suitable for power substations and other critical infrastructure where reliability and precision are paramount. Transportation Systems: Ensures seamless communication in transportation systems such as railways and highways. Smart Grids: Facilitates efficient and reliable communication in smart grid deployments. Benefits: Reliability: IEC61850-3 certification and redundancy support ensure uninterrupted operation in critical environments. Precision: IEEE1588v2 support provides nanosecond-accurate time synchronization for precise operations. Security: Compliance with IEC62443-4-2 ensures robust security measures to protect against cyber threats. Flexibility: Remote management capabilities enable convenient access and control from anywhere. Scalability: Designed for scalability to accommodate growing network demands. For more information, visit ATOP Technologies, Inc. | by BlackBear TechHive or contact [email protected].

While HTMX's popularity really took off starting mid 2023 (doubling in 2024), it's actually been around for quite awhile. Intercooler (HTMX's predecessor) has been around for over a decade. (HTMX is "Intercooler 2.0") * HTMX first commit in 2020 * Intercooler first commit in 2013 The State of HTMX (2024): https://lnkd.in/e6hmuC-e #htmx

Large language models (LLMs) depend on intricate internal mechanisms that necessitate a greater amount of memory than what is often accessible for functioning on conventional devices. A method that can be used for this purpose is the key-value (KV) cache. This cache is responsible for storing and retrieving data that has been computed before, allowing the model to create replies more efficiently by avoiding the need to recalculate information that has already been processed. This approach consumes a significant amount of memory as it stores a huge quantity of data in an easily accessible manner to improve the speed and efficiency of the model. As a result, the KV cache might become excessively big as the activities become more sophisticated, occasionally necessitating up to 320 GB for a single operation. In response to this issue, we have created FastGen, an innovative approach designed to decrease the memory requirements for Language Models (LLMs).

DTLP v3 is coming soon! About one month ago I created the v1 of DTLP (Data Transmission Laser Protocol). It was just sending data bit-by-bit from laser to photoresistor. The post about v1 is here: https://lnkd.in/eR63Tte9 Just now I got the first results of v3. Now it transmits the data byte-by-byte (x8 to speed). However there still some improvements to do (you can see it in the end of the video when receiving wrong bytes) So I need to ask QAs not to blame me 😀 P.S. v2 is ready too. It's enabling the encryption, so the data is encrypted in transmitter end and decrypted in receiver end. P.S.S. it's totally open-source and who is interested can support here: https://lnkd.in/eNPDUGPr

For VCF 9 the idea is to also provide a Federated Storage View, so that it will be easier to understand what you have available and where. We are even planning on visualizing vSAB stretched cluster configurations. Watch this video for more details! https://lnkd.in/eiZtdCG7

LLMs rely on memory-intensive mechanisms like the key-value (KV) cache to store and quickly retrieve data. FastGen optimizes KV cache usage, reducing LLM memory demands by up to 50% while maintaining performance. Learn more here and at #ICLR2024. https://msft.it/6049YVzqd