Benjamin Treynor Sloss

The Site Reliability Engineering: How Google Runs Production System book is arguably one of the most prevalent and well-known SRE resources out there, but there are many other great options for folks looking to expand on their learning. We recently spoke to Google's Reliability Advocate, Steve McGhee, in our Humans of Reliability interview series. In addition to his interesting anecdotes on the early days of SRE at Google, and his journey to becoming a Reliability Advocate, he also shared a handful of his favorite SRE resources, which we compiled here into a list: https://lnkd.in/gpUVhyRa

16

Folks I've spoken to about Amazon have likely heard me talk about the importance of the PR/FAQ process, a.k.a. "the working backwards process". This blog posts gives a very nice example of a PR/FAQ. https://lnkd.in/edTeuVRQ. Note that the blog doesn't capture the presumably 100s of reviews/discussions/rewrites that this doc represents, which is a part of the real value. When the process is done it all looks so simple, but getting a large organization to make a decision to something simple and do it in a BIG way is the challenge the fun of it all.

63

3 Comments

Just in time for #Climateweek2024! Thank you to the The New York Times for spending some time with me and Pure Storage to discuss how the global hunger for #AI is driving a need for new ways of recording, storing and managing data. It is all going to come down to #efficiency and using better science to solve hard problems. "As the leaders of business in this new era, we have a responsibility to build A.I. in a way that is sustainable, responsible and scalable to support our future." #ESG #DataPlatform https://lnkd.in/gBBZ3Q7p

33

1 Comment

➰ APIContext's latest insights on cloud-based API performance have been featured on Nordic APIs! In the 2024 report, we delve into crucial aspects affecting API efficiency: 🔐 Authentication is key. Ensuring robust authentication mechanisms in APIs is crucial, yet offloading this function can introduce complex challenges. 📖 DNS configurations matter. How you handle DNS settings can significantly impact API performance, particularly when tied to authentication processes. 🌐 The role of third parties. Critical third-party services play a significant role in 🔗 API transactions, underscoring the need for careful integration and monitoring. ⛅ Not all clouds are created equal. Differences in cloud services can drastically affect API performance, an essential consideration as APIs account for over 80% of all internet traffic. #APIs #CloudComputing #TechInsights #APIContext

15

1 Comment

AWS has rolled out its Trainium2 chips for general availability and is already teasing Trainium3, but it’s still early in the game, and some analysts question whether the massive compute Trainium supports is even necessary — or relevant — for many enterprises. GAI Insights Alvin Nguyen

4

Earlier this year, I shared the launch of Anthropic’s Claude 3 family of vision-enabled models on Amazon Bedrock. I am pleased to announce that Claude 3.5 Sonnet is now available to AWS customers; outperforming and operating at one-fifth of the cost of their previously most capable model, Claude 3 Opus. Claude 3.5 Sonnet can transcribe text from images, interpret charts and graphs, and write and edit code. This model raises the industry bar for intelligence- improving its writing capabilities to feel more authentic and allowing it to provide more accurate and comprehensive insights than any other foundation model. It is also a powerful model for complex enterprise tasks like context-sensitive customer support, multi-step workflows, and creating revenue-generating apps. Visit the Amazon Bedrock console to get started with Claude 3.5 Sonnet. https://lnkd.in/gudBiVwP https://lnkd.in/gV-YzN3K #AWS #Anthropic #AmazonBedrock #Claude #GenAI #GenerativeAI #MachineLearning

316

5 Comments

Well done #AWS for promptly fixing an S3 billing edge-case! Interesting S3 cost issue in this blog: "How an empty S3 bucket can make your AWS bill explode". In brief: external unauthorised (denied) requests to a new empty S3 bucket were being charged - $1,300 within two days. The author had inadvertently chosen the same bucket name that was the placeholder default configuration in a popular open-source tool!    https://lnkd.in/eNEqedR8 That blog was published on 29th April 2024. On 13th May, AWS fixed this problem: "Amazon S3 will no longer charge for several HTTP error codes". https://lnkd.in/eMXuhhu8 To quote that announcement: "bucket owners will never incur request or bandwidth charges for requests that return an HTTP 403 (Access Denied) error response if initiated from outside their individual AWS account or AWS Organization."

6

Primitives or abstractions? I had the chance to pose this question to Matt Garman, CEO of AWS, at re:Invent last week. His response was clear: AWS is doubling down on both. They’re investing in innovating and owning key primitives such as foundation models, infrastructure, and chips, while also building powerful abstractions—services that enable developers to run diverse workloads and create innovative applications with industry-leading price-performance. Against the backdrop of thousands of attendees buzzing about all things AI, AWS once again delivered one of the best industry conferences. Here are my top 5 insights from the event.

18

4 Comments

When adding all of their clients to a new Confluent Cloud-based architecture, interactive ad-tech platform Skai realized their initial design would require 90,000 topics in order to accommodate everyone. Using many-to-one relationships between producers and topics, and between consumers and topics, Skai devised an ingenious architecture that reduced that number to 1,800 topics. Learn how they did it in a new blog: https://cnfl.io/3WIEQcf

41

2 Comments

We are pleased to announce that you can now achieve up to an aggregate 1.3 million I/O operations per second (IOPS) and 12 GB/s throughput per OCI compute instance with the OCI Block Volume Service. In addition, you can also attach up to 32 Ultra High Performance (UHP) volumes to a single OCI compute instance. This service update is particularly useful for high I/O workloads, such as OLTP, AI/ML, data warehousing and analytics, databases with large data sets and extensive complex processing. https://lnkd.in/eHErX5mC

2

The speed at which these models improve is staggering, so is the cost to build them. Only way to keep up is to leverage principles such as AWS Bedrock so you have the flexibility you need to quickly select the model that works best for your task. https://lnkd.in/gNAR63YC https://lnkd.in/g_FtP6e9

36

Something that's been in the works for a long time: details of the formal correctness proof (in Dafny) of AWS's policy authorization engine, which makes 2+ billion decisions per second about who can access what, on practically every call to AWS. Talk is by Lucas Wagner and Sean McLaughlin https://lnkd.in/eiJYQeqw

214

6 Comments

Amazon FSx for NetApp ONTAP now supports #NVMe-over-#TCP for simpler, lower-latency shared block storage. Read more here...

39

5 Comments

It was somewhat scary to take on the role of WW Head of EPIC at AWS. I had to say goodbye to my stable role as WW Business Development Manager of RDS/Aurora (the fastest growing service in AWS history) to move to this radically different, untested role. It was a totally new role I'd created, and I had no idea if and how it would work out! I had to embrace uncertainty. Do you enjoy uncertainty or avoid it like the plague? It turns out that embracing uncertainty can be really good for you. We all have different relationships with it, but in the business world, we tend to index towards trying to make things as predictable as possible. We do things like strategic planning, forecasting, modeling, predictive analysis, etc. to make things more “certain”—and there are obvious advantages to that. Yet the reality is that life (and work) are filled with uncertainties…and probably more than we want! If we can learn to embrace the inevitable uncertainties that come at us, we will be healthier, more successful, and happier along the way. That was actually one of the main points in my talk on "Emotionally Intelligent Leadership for an AI-powered Future" at AWS ExecLeaders London this week. If there is one thing we know about the AI future—it’s that no one really knows what it's going to look like. As best-selling author and business guru Jim Collins said, "Uncertainty is chronic, instability is permanent, disruption is common, and we can neither predict nor govern events. There will be no new normal, there will only be a continuous series of not normal episodes, defying prediction, and unforeseen by most of us until they happen." Welcome to the world of continuous uncertainty, instability, and disruption! Learn more in this article, How Embracing Uncertainty Can Improve Your Life: https://lnkd.in/gPNv4-_R How do you feel about uncertainty and ambiguity? What has helped you have a healthier relationship with these realities? I'd love see in the comments below. #uncertainty #ambiguity, #complexity #success #emotionalintelligence #epicleadership #ai

124

4 Comments

So much potential here for security and risk mgmt as we refine and improve the use of ML for code migrations or transformations (think memory safety as an example) https://lnkd.in/gzjdTRBW

25

Amazon Web Services (AWS) is at war and Andy Jassy has a new consigliere Matt Garman. Amazon.com Inc. has strategically positioned Matt Garman as the new CEO of Amazon Web Services (AWS), much like a wartime consigliere in the vein of the iconic film "The Godfather." Garman assumes command in a critical era for AWS, marked by intense competition for cloud and AI supremacy. This period follows significant shifts brought about by the pandemic, a new wave of developers, and rapid advancements in generative AI technology. AWS finds itself in the midst of a fierce battleground, not only maintaining its cloud infrastructure leadership but also vying for dominance in the burgeoning AI sector. The one other wildcard in favor of AWS with Garman (besides being a product czar) is that he's very partner friendly which will be a key part of keeping AWS ahead. Lots of commentary and our post has the news with reaction from the top analysts. SiliconANGLE & theCUBE has the news. Story Link: https://lnkd.in/gDJFC496

49

12 Comments

Can you trust AI? 🤔 For most companies the answer is no - at least using current foundational models out of the box. ❌ Problems like sensitive data leakage and toxic content are major barriers to implementing production genAI solutions. The route towards trust lies in system architecture and the good news is the field is advancing rapidly, including increasingly easily accessible toolkits of solutions to increase trust in AI systems. 🔥 This is the hot area for new releases from major AI platform players (along with AI security), including AWS (https://lnkd.in/e45S-ehS), DataRobot (https://lnkd.in/eCWvppWR), and Databricks (https://lnkd.in/e6A7cTcH). 👨‍⚖️ Guardrail or judge models - leveraging a second AI model to police the content coming out of an original - are probably the most promising scalable solution. For many the best route is likely to be a new class of foundational models specifically designed for this task such as Llama Guard and Databrick’s new judge LLM. There is also the ability to fine tune a model (think a discriminator model from a GAN architecture) or use a general purpose foundational model. 👮‍♂️ Classic filtering techniques for sensitive data can also be applied to the result of an LLM, whether for PII, company IP, or inappropriate content. This can include regex, classic PII models (such as Amazon Comprehend), etc. 🚻 Human in the loop techniques can also be used, but are far more scalable when used to calibrate automated solutions such as those described above rather than manually verifying outputs in realtime. What’s still missing? 🔎 🎯 While techniques to improve accuracy and reduce hallucination such as RAG and AI agents have made great strides, there remains a big gap in conveying a model’s confidence in its output to end users. 💡One idea is to leverage the probability distributions used for top-p (nucleus sampling) at the token level instead at greater aggregations to convey confidence levels for entire model outputs or significant segments (e.g. each step in a chain of thought structure or AI agents framework). #ai, #artificialintelligence, #GenAI, #GenerativeAI, #llm, #datascience, #machinelearning, #mlops, #bigdata

5

1 Comment

Real-time data, real impact 💪 Join us on November 19th at 1 PM ET for an exclusive fireside chat with LiveRamp’s Chief Architect, Kannan Dorairaj, and Redpanda’s CMO, Mike Boyarski. Discover how LiveRamp uses Redpanda to drive #realtime personalization, reduce costs, and create faster #data pipelines. Don't miss out!👇 https://lnkd.in/grknhDMV

3