How Silk Delivers 10X Performance at Over 40% Lower Cloud Costs
Executive Summary
In this webinar, Ori Wiseman, Director of Sales Engineering for Silk (EMEA), and David Berliner, Senior Director of Product at Silk, discuss how Silk’s software-defined cloud storage platform enables organizations to achieve 10x performance at over 40% lower cost in the cloud.
They explain the key architectural innovations — including thin provisioning, data reduction, independent performance and capacity scaling, and copy data management — that drive massive efficiency gains for mission-critical database workloads in Azure.
The session concludes with real-world examples, customer savings results, and insights into how Silk helps enterprises prepare for AI-driven data demands.
Key Takeaways
💾 Thin Provisioning Saves 33%+ Immediately: Silk customers pay only for what’s actually used, not what’s allocated.
⚙️ Data Reduction Shrinks Data by Two-Thirds: Compression and deduplication minimize wasted space and improve efficiency.
🚀 Performance and Capacity Scale Independently: Silk allows real-time scaling without downtime — no need to over-provision.
💡 Smaller VMs, Bigger Performance: Silk delivers 3 GB/s throughput on 2-core VMs — outperforming Azure’s 96-core VMs.
🧠 Copy Data Management & AI Readiness: Instant, thin clones and “Echo” snapshots accelerate dev/test, analytics, and AI modeling safely.
[00:00:00] Introduction
Ori Wiseman (Silk):
Welcome everyone to our continuing webinar series on cost and performance. Today’s session is titled “How Silk Delivers 10X Performance at Over 40% Lower Cloud Costs.”
Joining me is David Berliner, Senior Director of Product here at Silk and our resident FinOps and TCO expert.
David Berliner (Silk):
Thanks, Ori — great to be here. My role focuses heavily on helping customers understand total cost of ownership and where Silk creates measurable savings.
Ori:
Perfect. Today we’ll dig into how Silk achieves those cost and performance advantages, particularly for large relational databases in Microsoft Azure — where native options often fall short for enterprise-scale workloads.
[00:02:00] Understanding Cost Inefficiency in Cloud Storage
Ori:
So, David — what makes Silk so different from native Azure storage when it comes to cost efficiency?
David:
It starts at the data layer. With Silk’s thin provisioning, customers pay only for the storage they actually use — not what they provision.
In Azure, if you deploy a 100 TB managed disk but only use 2 TB, you’re still paying for all 100 TB. Silk eliminates that waste.
Across our customer base, we’ve seen an average 33% immediate savings just from thin provisioning.
Ori:
And that’s based on real customer data — not theoretical numbers?
David:
Exactly. These are live results from deployed environments, and for customers new to Silk, the savings are often even greater.
[00:04:00] Data Reduction and Compression
David:
Beyond thin provisioning, Silk applies data reduction, removing extraneous zeros and compressing data. On average, this shrinks data by two-thirds — a major efficiency gain.
Ori:
Does that include deduplication too?
David:
Yes, for some workloads. While deduplication is less impactful for databases, it’s valuable for other data types. The combined effect of compression and thin provisioning significantly lowers total storage costs.
[00:05:00] Independent Scaling of Performance and Capacity
Ori:
Let’s talk about performance.
David:
In Silk, performance and capacity scale independently. Most native cloud services tie IOPS to storage size — so to get more performance, you must buy more capacity.
With Silk, you can scale performance by adding C-nodes without changing storage size. It’s instantaneous, with no downtime and fully automatable via API.
Ori:
That’s powerful — and you can even script scaling around known demand spikes like Black Friday or end-of-month processing.
David:
Exactly. We’ve built full API support for automation and policy-based scaling.
[00:07:00] The VM Performance Gap
Ori:
I looked at Azure’s own VM documentation and found a huge bottleneck — throughput scales only with core count.
For example, to get 1.5 GB/s throughput, you’d need a 64-core VM.
David:
Right — and that’s where costs explode. Cores drive both compute and licensing costs, especially for SQL and Oracle.
Ori:
Here’s the crazy part: we’ve tested Silk running 3 GB/s throughput on a 2-core VM — performance Azure can only match with a 96-core instance. That’s 48x fewer cores for the same result.
David:
Exactly. Customers can run smaller VM shapes with greater performance — that’s where the 40%+ cost savings really compound.
[00:11:00] Customer Example: Healthcare Industry
David:
We’ve seen this proven in the field. One healthcare customer doubled their environment size while reducing overall cost by 26%.
In their data warehouse, they used Silk to switch from an expensive M128ms VM to a D96 — a smaller series with 33% fewer cores — and achieved 76% total VM savings.
Ori:
That’s remarkable — smaller footprint, better performance.
[00:13:00] Extreme VM Comparison
Ori:
There’s a top-end Azure VM — the M176bds_v3 — offering 10 GB/s throughput at $60,000 per month. Compare that to the E192ids_v6, which costs $13K/month and, with Silk, delivers 35 GB/s throughput — more than triple the speed at one-fifth the price.
David:
That’s the definition of better performance at lower cost.
[00:15:00] Copy Data Management and “Echo”
Ori:
Let’s talk about copy data management, one of Silk’s most powerful capabilities.
David:
Most organizations maintain multiple data copies — sometimes 20 or more — for analytics, dev/test, or AI. With Silk’s thin cloning, those copies consume minimal space. You only pay for incremental changes, not full copies.
Ori:
And Silk’s Echo feature takes it further — near-instant copies that are writable, secure, and refreshable in minutes.
David:
Right. One healthcare customer reduced nightly ETL from 14 hours to 15 minutes, and a payer customer cut dev/test refresh time from a week to minutes.
[00:20:00] Business Value and Operational Impact
David:
Our Business Value Assessments (BVAs) capture both hard and soft cost savings.
For instance:
Hospitals using Silk saw real-time access for clinicians by eliminating ETL lag.
Developers gained faster refresh cycles, improving productivity and agility.
Ori:
Those soft gains often translate into millions in avoided downtime and faster time-to-market.
[00:21:00] Preparing for AI and “Agent Chaos”
Ori:
We can’t skip the elephant in the room — AI.
David:
Right. AI workloads, especially agentic systems, create unpredictable database traffic — we call it “Agent Chaos.”
Large language models are non-deterministic — you won’t get the same output every time. That unpredictability can hammer production databases with random spikes.
Ori:
Exactly. Silk isolates production from those risks. With our cloning and performance architecture, customers can safely run AI training or inference on copies — not live systems — with no performance hit.
[00:25:00] The Broader AI and Data Landscape
Ori:
Relational databases still hold the majority of enterprise data — nearly 30% of all global data, representing tens of zettabytes.
As AI scales, those relational workloads will take the brunt of new demand. Silk helps organizations absorb that surge without re-architecting their environments.
[00:27:00] Next Steps and Closing Remarks
David:
If you haven’t already, check out the earlier session with Microsoft’s Michael Johnson — it dives into how Silk partners with Azure to evolve cloud infrastructure.
Visit silk.us
for our Cloud Cost Calculator, where you can compare your current spend versus Silk savings.
Or reach out for a technical discovery session or cost assessment — we’re happy to model your specific environment.
Ori:
Thank you, David — and thanks to everyone for joining.
Connect with us on LinkedIn to continue the conversation, and we’ll see you next time!
