Sneller is the fastest and easiest way to run real-time queries on your unstructured data in S3. Turn your event data into insights: CloudTrail, VPC Flow Logs, plus your own logs, metrics, and more! - Serverless and Speedy Query TBs per second using standard SQL. Pay only for what you use at $150 per PB scanned. Leverage Sneller’s handwritten SIMD/AVX-512 assembly. - Simple and Scalable Scale to petabyte-size tables on S3 object storage with ease. Keep your data in your own S3 buckets with no need for indexing. Low latency (~3 sec) direct ingest from S3 at zero cost. - Schemaless and Stateless Ingest any JSON, CSV, TSV, and more without defining schemas or ETL-ing your data. Never lose data due to full separation of storage and compute. True multi-tenant SaaS with no need to manage infrastructure. - Standard SQL Simple REST API for all your SQL queries. Robust support for SQL with many useful extensions. Navigate nested structures just as easily as in Javascript.
External link for Sneller
Los Altos, California 94022, US
Great work by Petr Kobalicek on speeding up integer arithmetic using #avx512
AVX-512: Division of 64-bit integers convertible to double precision without a precision loss is easy, but what about integers that are above 2^53? Even this is something we had to deal with at Sneller as 64-bit integers are first class citizens in our SQL engine. There is a lot of little things we had to deal with and this is one of them. See the blog post for more details: https://lnkd.in/depp9wB9
Fantastic acceleration of compression using #avx512 for fast data processing and large data using #sql
Another example of a great engineering we do at Sneller: AVX-512 optimized algorithm striking balance between lz4 and zstd! https://lnkd.in/eqHjBAjX
Some of you may be wondering what I've been up to. I have just published a blog post about some AVX-512 assembler code that we developed for an SQL query tool at Sneller. BLAZING FAST UNICODE-AWARE ILIKE IN AVX-512 We present a method to perform case-insensitive comparison of UTF-8 encoded strings using 16 parallel lanes and no branching. This method is used to implement the ILIKE operator for the Intel SkylakeX processor, written in AVX-512 assembly. https://lnkd.in/eHc7tcRe