Mark Cavage

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 Comments

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 Comments

Nice insights here for customers adopting or thinking about #kubernetes. Our Pure Storage Portworx team worked with Dimensional Research to interview 500 experts to understand how they are evolving their infrastructure strategies. Learn more at: https://lnkd.in/gNc57Q6X Venkat Ramakrishnan Andrew Custard Kalyan Ramanathan Stephanie (Rancourt)Richardson

93

2 Comments

FinOps as Code gets a functionality bump today: we're expanding the Vantage docs MCP to not only provide information about Vantage features...but it now also has context about the Vantage Terraform provider and help you write and troubleshoot Terraform configurations...and it's awesome. Below is a clip of this in action with Anthropic - when writing Terraform and hitting an issue, you can just ask Claude to resolve it. Claude leverages the Vantage Docs MCP as well as additional context of everything provided by the Vantage Terraform provider and will resolve and give you the templates accordingly. I'm always authentically stunned when I hear from customers moving over from other vendors that they're required to do everything manually via ClickOps on those platforms. This set of functionality will set you free...and honestly seems like table-stakes heading into 2026. Full blog post with all the details is in the comments.

44

2 Comments

Had a great interview with Alan Shimel at JFrog swampUP last week. We chatted a bit about my presentation and demo to optimize DevOps pipelines using GraphRAG. 🎉 Here is a link to the video: https://lnkd.in/gtk2zdCB

48

3 Comments

Back on SiliconANGLE & theCUBE this week with the inimitable Savannah Peterson and the ever-well-spoken Rob Strechay, unpacking Chainguard's introductions of the CHPs framework this week. CHPs are Container Hardening Priorities, and exactly what you need to go along with your SLSA, as you assess your security posture in Production. CHPs levels help you determine "how secure" your containers are in execution along 4 axes (minimalism, provenance, configuration, and vulnerabilities), while SLSA helps you understand how securely they were compiled and built. You're gonna love some CHPs and SLSA! https://lnkd.in/gn3AAD_j

43

2 Comments

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 Comments

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

124

4 Comments