Airbyte has announced availability of PyAirbyte, an open source Python library, intended to make it easy to move data across API sources and destinations by enabling Airbyte connector resources to be created and managed using code, rather than the user interface (UI). Python users (most existing data pipelines are written in Python) can add one command to their code and gain access to Airbyte’s more than 250 data connectors for moving data (all open source). Airbyte and LangChain have teamed to offer a new document-loading package for LLM-driven applications – making 250-plus data sources and more than 10,000 distinct datasets available for GenAI projects.
…
SaaS app data protector Alcion has unveiled a partner program, “Alcion for Partners,” tailored for Managed Service Providers (MSPs) and aiming to protect their Microsoft 365 customers’ environments. It says it has a specifically tailored engagement process and terms to fit the needs of the MSP market and enhance partners’ ability to grow their business. The Alcion partner portal provides, we’re told, an intuitive unified experience to monitor operations, manage configuration, and track licensing across all MSP customer tenants. More details here.
…
Cloud NoSQL database supplier Couchbase has added vector search to its Capella Database-as-a-Service (DBaaS offering and CouchBase server. It says it’s the first database platform to announce it will offer vector search optimized for running onsite, across clouds, to mobile and IoT devices at the edge. Vector search is a necessary part of Gen AI Large Language Model (LLM) processing and use cases include chatbots, recommendation systems and semantic search. It says its customers get similarity and hybrid search, combining text, vector, range and geospatial search capabilities in one. They have RAG to make AI-powered applications more accurate and enhanced performance because all search patterns can be supported within a single index to lower response latency. Couchbase is extending its AI partner ecosystem with LangChain and LlamaIndex support to further boost developer productivity.
…
A 2024-25 DCIG Enterprise Multi-site File Collaboration Solutions report names 5 top suppliers: 9 CTERA Enterprise File Services Platform, Nasuni File Data Platform, NetApp Cloud Volumes Edge Cache, Panzura CloudFS and Qumulo Scale Anywhere Platform. It evaluated a whole bunch of suppliers – see the image. Get a cope of the report completing the signup form here.
…
Dell PowerScale storage is being used by Subaru Lab to hold data for its AI-based EyeSight Driver Assist Technology (EDAT) developments. EDAT, built to monitor traffic movement, optimize cruise control and warn drivers if they sway outside their lane, is used in more than 5.5 million Subaru vehicles. It is based on stereo camera images, not using Lidar, and builds 3D images of the road around the moving vehicle. The aim is to reduce or eliminate vehicle collisions rather than develop full automated driving technology.
…
Reuters reports file collaborator Egnyte has hired bankers to facilitate an IPO with a potential $3 billion valuation. JPMorgan Chase is the lead banker along with UBS. Egnyte is profitable and has raised a total of $138 million in funding. It was valued at $460m million when it raised $75 million in 2018. Egnyte is located in a cloud file services market and overlaps with Box and Dropbox at the consumer and endpoint side of the market and with CTERA, Nasuni and Panzura in the filer replacement side.
…
GenAI Large language Model language processing unit (LPU) developer Groq has acquired Definitive Intelligence and set up a Groq Cloud business, led by its co-founder and CEO Sunny Madra. The aim is to significantly expand access to the LPU Inference Engine. GroqCloud is a developer playground with fully integrated documentation, code samples, and self-serve access, and available today at https://console.groq.com/. There will be a separate Groq Systems business unit, developing the LPU HW and SW, and serving the public sector and customers that require Groq hardware for AI compute centers.
…
Huawei’s Dr. Peter Zhou, President of the Data Storage product lines, presented at MWC24 in Barcelona talking about storage for AI, mentioning several existing products in his presentation: DME, OceanProtect E8000 and X9000, OceanStor A800, OceanFS, and OceanStor A310. He introduced a new product, the OceanStor Arctic magneto-electric storage system for cold data. There is very little information available about it. We are told it can reduce total connection cost by 20 percent compared to tape, and power consumption by 90 percent compared to hard drives. So, by implication, it does not use tape media or hard disk drives or, we assume, SSDs. That could mean it’s an optical disk device. The magneto-electric effect denotes any coupling between the magnetic and the electric properties of a material. That doesn’t sound very much like optical disk technology which relies on light and not magnetism. We have asked Huawei for more information about this OceanStore Arctic product.
…
A TechNewsSpace article claims Micron is set to use nanoprinting technology from Canon in some layering stages of DRAM chip production. Nanoprinting can achieve sub-nanometer resolutions and a nanoprinter could be five times less costly than EUV optical lithography equipment. A nanometer is one-billionth of a meter. Read a Canon article on nano-imprint lithography here.
…
ObjectiveFS is a distributed, log-structured, POSIX-compliant file system with an object store backed. The latest v7.2 release includes a new custom snapshot schedule for user-defined automatic snapshots, new mount options, performance improvements and more. New features and performance improvents include:
- Custom snapshot schedule for user-defined automatic snapshot schedules (learn more)
- New writedelay mount option to allow the kernel to delay writes to the filesystem
- New filehole mount option for user-defined maximum file hole size
- Increased read ahead in hpc mode to improve large read performance
- Improved Linux file attributes implementation to support newer FUSE versions
- Improved filesystem readiness for mounts from command line
- Added new regions for AWS and GCS
- Improved read performance with up to 5 percent speedup
…
Samsung has started testing the industry’s first 256 gigabyte (GB) SD Express microSD card, with maximum speeds of 800 MBps. It is scheduled to be ready for purchase later this year. The company has also started the mass production of its 1TB UHS-1 microSD card, set to launch Q3 2024.
…
Western Digital is selling 80 percent of its SanDisk Semiconductor Shanghai business unit, based in Shanghai, China, to partner JCET, according to Reuters. JCET is a Chinese test and assembly firm and is paying $624 million in cash for the flash drive’s business stake. The SanDisk unit will now be a joint-venture between JCET and minority partner Western Digital with its 20 percent stake.
…
Zilliz, supplier of the open-source Milvus vector database, has announce its Zilliz Cloud BYOC, with BYOC meaning Bring Your Own Cloud. Customers can host their vector embedding data within their private cloud networks. It uses a dual-VPC architecture. Private cloud hosting enables controlled data access permissions and can ensure compliance with regulatory standards. Deployment, upgrades, and maintenance are managed by Zilliz. The BYOC facility is available to customers on a scale of 128 CUs (Ziliz Computing Units) or more. Read a Zilliz blog to find out more. See Zilliz pricing plans here.
