Cost-Performance with a Side of AI: Cut RDBMS Costs, Prep Data for AI, and Fund Your Next-Gen Projects

Silk & Microsoft Webinar: Cost Performance with a Side of AI

Topic: Cutting RDBMS Costs, Preparing Data for AI, and Funding Next-Gen Projects

Summary

In this session, Silk and Microsoft experts discuss how organizations can reduce relational database (RDBMS) costs while accelerating performance and readiness for AI workloads.

Through a real-world customer success story with SimCorp, they illustrate how Silk’s cloud platform improves throughput, lowers compute and licensing costs, and supports next-gen AI initiatives.

The conversation also explores Microsoft–Silk collaboration, AI trends like Agent Chaos, the Natural Language Web, and the role of Agentic AI in future architectures.

Key Takeaways

Silk functions as a high-speed virtual SAN in the cloud, optimizing database performance and cost.

The Silk–Microsoft partnership led to new Azure VM SKUs and measurable customer savings.

SimCorp’s migration to Silk achieved a 7GB/s throughput and reduced core usage by half.

Silk’s copy data management enables zero-footprint clones, critical for AI data modeling.

Discussion covers AI workload strain, agent management, and natural language web evolution.

Transcript
[00:00–02:00] Introduction and Overview

Ori Wiseman (Silk):
Welcome, everyone, to our webinar “Cost Performance with a Side of AI.”
I’m Ori Wiseman, Director of Sales Engineering for Silk in EMEA. Before Silk, I worked at Microsoft and AWS.

Joining me today is Michael, Cloud Solution Architect at Microsoft in Denmark.

Michael (Microsoft):
Thanks, Ori. I’ve been in IT for about 25 years, most of them at Microsoft. I’m thrilled to discuss how Silk and Microsoft collaborate to drive performance and efficiency for enterprise workloads.

[02:00–03:00] The Silk + Microsoft Partnership

Ori:
Silk is a proud Microsoft IP Co-Sell Partner, with deployments directly in customer tenants.
This partnership removes commercial friction and helps enterprises reduce Azure costs through optimized infrastructure and licensing.

Michael:
Yes, Silk’s Marketplace eligibility and MAC reductions make it easy for Microsoft customers to deploy and save immediately.

[03:00–06:00] What Silk Does

Ori:
In essence, Silk is a high-speed virtual SAN in the cloud. It supports any relational database workload—Oracle, SQL, PostgreSQL, MariaDB, and more.
We help customers cut licensing, compute, and storage costs, while preparing their data for AI and LLM solutions.

[06:00–09:00] Customer Success Story: SimCorp

Ori:
Michael, let’s talk about SimCorp, where our teams first collaborated.

Michael:
Yes, SimCorp was modernizing their on-premise SimCorp Dimension to a SaaS model. As they approached go-live, a core component was discontinued, creating a major performance gap.

Ori:
We jumped in quickly. Within a week, Silk deployed a POC that outperformed Exadata benchmarks, reducing run time from 2.5 hours to 20 minutes.
They were able to move from a 128-core VM to a 64-core VM—maintaining performance and cutting costs dramatically.

Michael:
The collaboration was intense but seamless. Silk’s solution stabilized performance and allowed SimCorp to go live successfully.

[09:00–12:00] Post-Deployment Optimization and New Azure SKU

Ori:
After deployment, Microsoft identified further optimization opportunities.

Michael:
Yes. We realized Silk could benefit from a better CPU-to-storage ratio. That led to the creation of a new Azure Lsv4 SKU, co-developed with Silk’s engineering team.
The result? 30% more capacity and lower costs—now generally available to all customers.

Ori:
That’s a huge win for the ecosystem and proof of how closely Silk and Microsoft collaborate.

[12:00–15:00] Reducing Costs with Copy Data Management

Ori:
Another cost-saving tool is Silk’s zero-footprint cloning.
It lets you instantly create virtual database copies without doubling storage. A 10TB database cloned through Silk only consumes additional space for changes—so 11TB total instead of 20TB.

Michael:
That’s a game-changer for dev/test and AI data preparation environments. Many customers see 10:1 data reduction through this approach.

[15:00–18:00] Azure VM Innovation and Performance Boost

Ori:
Microsoft’s new Mv3-series VMs can deliver up to 10 GB/s throughput but cost about $60,000 per month.
With Silk and the E192sv6 VM, customers achieve 35 GB/s throughput at sub-millisecond latency—at just $13,000 per month.
That’s over 3.5× performance at one-fifth the cost.

Michael:
It’s an impressive example of joint innovation between Silk and Microsoft engineering.

[18:00–22:00] The AI Landscape: Managing “Agent Chaos”

Ori:
Let’s talk AI. You’ve described something called Agent Chaos. What is it?

Michael:
Agents are task-driven layers built on top of LLMs. As more are created, systems experience massive data strain and latency challenges.
Silk helps manage this by maintaining consistent, low-latency performance for these workloads.

Ori:
Exactly. Many of our customers are using AI for real-time insights on proprietary data—and soon, they’ll use it to write data too, creating new demands on databases.

[22:00–25:00] The Natural Language Web

Ori:
At Microsoft Build, there was a keynote on the “Natural Language Web.” Can you elaborate?

Michael:
Sure. It’s a vision where websites become conversational. Instead of clicking filters, you’ll simply ask:

“Find me a direct flight under $1,000 next Tuesday.”

The site responds instantly through AI agents using the Model Context Protocol.
This evolution, though exciting, will add tremendous backend demand—exactly where Silk’s high-performance data layer proves essential.

[25:00–27:00] Data Modeling and Development Flexibility

Ori:
Silk’s cloning and snapshotting make it easy to spin up multiple test environments instantly.
In the old days, backup-restore cycles took hours or days. Now, you can refresh or replicate databases in seconds—perfect for AI model training or parallel development.

Michael:
Yes, it allows teams to test LLMs safely with live data while keeping production systems isolated.

[27:00–29:00] Agentic AI: Real-World Examples

Ori:
Let’s shift to Agentic AI—AI that acts independently in the background.

Michael:
Right. For instance, a dentist I know uses an AI assistant that scans X-rays and flags potential issues before review.
Similar systems in hospitals have already caught critical conditions early, preventing major health crises.

Ori:
Exactly. These background AI agents are also transforming cybersecurity—scanning for anomalies or ransomware continuously. Silk’s data resilience helps enable that level of monitoring.

[29:00–30:00] Closing Thoughts

Ori:
This has been a fantastic discussion—thank you, Michael, for joining and sharing your insights.
Our next session will feature David Berliner, Senior Director of Product at Silk, for a technical deep dive into Silk’s architecture and AI capabilities.

Michael:
Thanks, Ori. Always a pleasure collaborating with you and the Silk team.