@lak_luster@OctoML We have around a 100 hardware options to pick for deployment, & I selected 11 of them, to show the wide range of inference times depending on the hardware. We didn't pick just the 11 fastest targets as those would also likely be the 11 most costly so not ideal for all use cases.
@lak_luster@OctoML Hey @lak_luster! That Azure instance was not the fastest one and it was somewhat arbitrarily chosen, mainly to show that we have coverage across all 3 clouds, along with some edge devices that we host ourselves like the Jetson + Cortex A72.
Are you planning on running @ApacheSpark streaming pipelines in @googlecloud via Dataproc or @databricks?
Today we released a Spark Structured Streaming connector for Pub/Sub Lite, a new exceptionally low-cost scalable messaging service.
Details in blog:
https://t.co/k2Z6l17aZq
You can now use Dataflow Flex Templates in @GCPcloud to turn any streaming or batch Dataflow pipeline into a reusable template that anyone can run.
https://t.co/4dsYhcywgi
Another excellent video using math to explain the world from @3blue1brown today, this time #coronavirus "Exponential growth and epidemics" edition. Cases outside China are growing 10x every 16 days and 1 billion people could be infected by May 26.
https://t.co/PueRSNwHL0
If you run #Hadoop in Google Cloud, you'll want to upgrade to the new GCS Connector 2.0. This blog covers the improvements we added, in collaboration with @TwitterEng.
https://t.co/liDVWIgPjs
Want to learn the latest about #BigQuery, #DataWarehousing best practices in @googlecloud, & BigQuery ML?
Tomorrow (May 15) at 9am PDT, I'll be hosting a free, half-day BigQuery event online.
Sign up here:
https://t.co/9HCMLSwPK8
Want to learn how Twitter is implementing one of the largest #Hadoop data migrations on the planet? Join us at #GoogleNext19 next week when @twitterhadoop's strategy and lots of new technical details will be revealed...
The Makali'i was Hawaii's first research submarine. Fabricated in 1966, it was used to harvest corals in the 70s and to visit the site of the first hydrogen bomb crater at Eniwetok Atoll in the 80s. Now on display at @MauiOceanCenter
Google worked with NOAA to train a neural network (CNN) to identify humpback whale songs in170,000 hours of underwater audio recordings from the Pacific Ocean. Anyone wanna take a crack at decoding their language next?
https://t.co/gzwYyvzMdx
https://t.co/2AuPAMAIso
Twitter just released a 363 GB dataset of 10 million tweets from 4,611 Russian/Iranian accounts. "Independent analysis of this activity by researchers is a key step toward promoting shared understanding of these threats."
https://t.co/w3Ws2H0Jns
Appreciate the conversation today. This reflects a lot of our thinking. We’ve changed Twitter a lot over time, but one thing we haven’t touched is what the service implicitly incentivizes people to do. That’s the level we need to reconsider.
In case you missed it at #GoogleNext18, the video for "How @TwitterHadoop Chose Google Cloud" to migrate their #Hadoop clusters (>500K cores & >300PB) is now online:
https://t.co/TxpZrtUak5