Steve McGhee

People often ask me about the differences in architecture between Amazon Dynamo (the 2007 SOSP paper), DynamoDB (the AWS serverless NoSQL database), and Aurora DSQL (the AWS serverless SQL databases). I captured the key points on durability, high availability, programming model, and trade-offs in my blog: https://lnkd.in/gkMG_tVD TL;DR: Despite the shared name, Dynamo and DynamoDB have little in common architecturally, and make very different trade-off decisions.

631

8 kommentarer

I keep being asked how Ably performed during the AWS us-east-1 outage this week. The bottom line is that there was no customer impact: no downtime, no errors and imperceptible impact on latency. Ably is hosted on AWS with services operating in multiple regions globally. Each region scales independently based on traffic, and us-east-1 is normally the busiest region. When AWS services in us-east-1 started failing, the Ably data plane kept operating as normal. The AWS disruption meant that we couldn’t add capacity in that region, so new connections were routed to us-east-2 instead; this is a routine intervention that we make in response to disruption in a region. Existing connections in us-east-1 stayed live, serving traffic with the same latency and zero errors. This was our globally-distributed system doing exactly what it was built to do. Read the full breakdown: https://lnkd.in/eeVsie4a

126

4 kommentarer

Researchers from Wiz recently disclosed CVE-2025-49844 (RediShell) in Redis, and it’s a good reminder for anyone building or operating managed or multi-tenant systems. I’ve always had a soft spot for anything related to Redis and Valkey since my time at AWS ElastiCache. This finding is a great design exercise in thinking about how we isolate shared components like Redis in managed environments. Limiting resource sharing is important, but it’s not enough. To really contain risk, we also need to narrow the attack surface by profiling applications and defining what they should and shouldn’t be allowed to do. There are many approaches to achieving that, from specialized runtimes to unikernels. But for cloud providers who need to secure large fleets and apply protections without downtime, eBPF offers a uniquely practical path. It can enforce fine-grained controls or apply security policies across thousands of running instances without restarting processes or draining nodes. That combination of precise control, immediate rollout, and zero disruption makes eBPF one of the most effective tools available for tightening sandbox boundaries in real-world systems.

46

6 kommentarer

Super excited to share the our agent-gateway oss project is getting an overhaul! Agent gateway is a rust-built, dedicated gateway for MCP, A2A, LLM, and inference use cases. Updates include: 👉 Update to latest A2A and MCP specs 👉 Kubernetes Gateway API support 👉 More sophisticated traffic policies and resilience 👉 LLM Gateway and unified API 👉 Inference Gateway / Extensions Go try it out now, blog links in the comments 👇

124

12 kommentarer

We recently announced the GA of Oracle Database@AWS. Customers can now run Oracle Exadata Database Service and Oracle Autonomous Database on dedicated infrastructure on OCI within the AWS U.S. East (N. Virginia) and U.S. West (Oregon) Regions. In addition, to meet growing customer demand, we plan to make the service available in an additional 20 AWS Regions around the world. InfoWorld has the details: https://lnkd.in/gswnvM9z

108

8 kommentarer

Today is an exciting day for the open source movement! The OpenInfra Foundation Board has unanimously agreed to take the next steps in joining forces with The Linux Foundation. As someone who helped create the OpenInfra Foundation (originally the OpenStack Foundation) 15 years ago, I'm thrilled about this natural evolution in open source collaboration. This move doesn't change our core operations - our board, governance model, and community-first approach remain intact. Instead, it aligns us more closely with an ecosystem our users and developers are already deeply engaged with. This decision reflects the changing landscape of open source infrastructure. Today, organizations are combining multiple open source projects like OpenStack, Kubernetes, Linux, PyTorch, and Ceph to build modern, scalable, AI-ready infrastructure. By joining the Linux Foundation, we're positioning ourselves to collaborate more effectively across these technologies and support the evolving needs of the industry. I'm excited about what lies ahead. This isn't just a win for OpenInfra projects - it's a fantastic move for the entire open source ecosystem, ensuring its continued growth and success well into the future. Here's to the next chapter in open source collaboration! #OpenSource #OpenInfra #LinuxFoundation #FutureOfTech Read more: https://lnkd.in/dJbi8yZU

240

32 kommentarer

Exciting news for the Go community: the Go team at Google is working with Anthropic to build and support an official Go MCP SDK! I'm humbled by how the Go community came together in my proposal to share their support for a unified MCP solution. As most gophers know, Go is an amazing language for building micro-services, CLIs, cloud-native, and much more. And now, I'm very bullish on the future of building AI enabled applications with Go and MCP!! https://lnkd.in/gxMd5hKd

1 322

13 kommentarer