@rseroter I hope there are exponential back offs with the retries.
English
Anna Berenberg
164 posts
Anna Berenberg
@kniga
Engineering Fellow at Google. Services, Load balancing, reliable distributed systems. Mother of grown up children. Knitting. Tweets are my own.
Katılım Şubat 2009
189 Takip Edilen1.8K Takipçiler
Great new doc from a friend of mine!
"Every LLM call, every tool invocation, and every step of the agentic loop is persisted by Temporal. If the process crashes, the network drops, or an API times out, Temporal automatically retries and resumes from the last completed step."
Temporal@temporalio
1/ Temporal is now in the Google Gemini API docs. We partnered with the Gemini team to publish a tutorial: building a durable ReAct-style agent with Gemini and Temporal. Three screenshots tell the story better than any benchmark. 🧵 ai.google.dev/gemini-api/doc…
English
Anna Berenberg retweetledi
Google’s Cloud IDP Could Replace Platform Engineering thenewstack.io/googles-cloud-… < this @thenewstack piece features yours truly, along with the much wiser @kniga
English
It was exiting to share One Network architecture with the world! Aside from the technical complexity, sheer size of rearchitecture included 12+ teams and 125+ projects over last 4.5 years! Thank you to #QConSF for the opportuniry to bring One Network public! #envoyproxy #grpc
QCon San Francisco Software Development Conference@QConSF
🚀@kniga, Engineering Fellow @googlecloud, discusses One Network, a cloud-agnostic, service-oriented network architecture that unifies policy enforcement across public/private clouds, runtimes, and traffic patterns—simplifying deployment and enhancing security. #QConSF
English
Anna Berenberg retweetledi
To achieve the best end-user experience for #generativeAI apps and to gain efficient use of limited and costly GPU and TPU resources, we announced several new networking capabilities that optimize traffic for #AI applications.
Take a deeper look → goo.gle/3Wi0DbI
English
@mattklein123 Would be interesting to think how logs play into ability of LLMs to predict abnormal events based on previously saved logs. How many logs from the past is enough to predict the future ? Is the cost of keeping large volume of logs justified in this case ?
English
New post on what happens after you add a control plane and local storage for observability: you get 1000x the telemetry when it matters.
mattklein123.dev/2024/04/17/100…
English
The future is application-centric. Today we launched App Hub GA - this is a baby step and the sky is the limit in the journey of speaking the language of business instead of language of cloud infrastructure. cloud.google.com/blog/products/…
English
@QuinnyPig @DuckbillGroup That’s the point. They may continue to have them, but who will care if you customer will be able to answer the questions such as “how much my application cost per day, per month ? “ or what is the priciest workload in my application ? Natural language will change products
English
@kniga @DuckbillGroup "Thousands." As of 10AM PDT today, AWS has exactly 1,215,100 SKUs in their pricing API.
English
Looking forward to hearing about all of the useful things Google Cloud is helping customers with instead of a bunch of AI hype! Surely they won't disappoint me here...
(Atomic Liquors drinkup tomorrow evening at 7PM; come one come all...)
English
@QuinnyPig @DuckbillGroup Too early. Not enough tools. Not enough products to use. Give it a time. If you have UI or programmability or customization, it will radically evolve. Billing will definitely change - cloud providers will not get away with thousands of SKUs.
English
@kniga That's kinda my point; take The @DuckbillGroup. Our GenAI experiments have been slight improvements to existing processes at best; there's no slam dunk "this evolves our business" play here that I've been able to find. What'm I missing?
(The question is sincere.)
English
@QuinnyPig There are also customers who run custom gen ai products - either they own inference stack or they build infrastructure for gen ai app developers. We need to solve for them also
English
@QuinnyPig There are 2 questions I want to ask every customer
1. Are you planning to evolve business with gen AI and if yes, how ? Answer is possible No if they don’t have customer support
2. How do you think genAI will help you to consume GCP ? I believe here possibilities are endless
English
All Google traffic is load balanced by global algorithms and traffic management policies. A few days ago, we released some of these policies to GCP LBs and Traffic Director to sprinkle a bit of Google magic on Cloud.
cloud.google.com/load-balancing…
@GoogleCloudTech
English
The cat is out of the bag : Duet AI is here. It was quite a steep and fast learning of genAI components to build a serving stack for Duet AI. #GenAI #GoogleCloudNext #GoogleCloud
Brad Calder@CalderBrad
Here is a summary of the Duet AI in #GoogleCloud features announced today. Duet AI nows provides AI assistance across dev, sre, ops, data practitioners, and cybersecurity professionals. #GenAI #DuetAI cloud.google.com/blog/products/…
English
@loriberenberg Teenagers have selective hearing problem. Penny is 9, it is new teen - so welcome to teen parenthood :-)
English
it’s not that i want to use baby voice, it’s that if i don’t talk like elmo, my dog will literally pretend she can’t hear me
English
I never thought I'd play a part in the movie. And yet here it is - the documentary about the history of Envoy proxy - envoyprojectdocumentary.com from @Speakeasy_Films premiers at KubCon in April.
English
one of my favorite parts of introducing myself is the inevitable "oh my aunt's name is lori!"
English
Anna Berenberg retweetledi
We're looking for woman founders to present at Demo, Don't Pitch (lu.ma/DDP1) on the 24th at 6:30 PM. 2 spots are being reserved and we'd love to get a couple more applicants at tally.so/r/wM1zvX
English
@kelseyhightower I saw on Twitter someone describing SRE as Sidecar Restart Engineer, so I guess they understood explicit choice
English