5 Lessons From eToro's Azure Migration Every Cloud Infrastructure Leader Should Steal

Milliseconds are money at eToro. So when the trading platform serving millions of customers decided to leave on-prem hardware behind for Azure, the storage underneath it couldn’t afford a single bad day.

We recently sat down with Ori Weizman, Silk’s Director of Solution Architecture for EMEA, and Yitzchak Wahnon, DBA Team Manager at eToro, for a fireside chat on exactly how that migration happened. Fair warning: Yitzchak made Silk earn this one. Here’s what came out of the conversation.

Why did the DBA get kicked out of the bar? Because he kept joining tables without permission. (Ori’s joke, not ours. We’re leaving it in.)

Watch the full conversation on demand here!

1. Azure’s own storage couldn’t keep up

eToro’s database is basically one giant, fast machine handling all trading activity at once — no room for a storage layer that hesitates. When the team tested Azure’s native Elastic SAN, it came up short almost immediately: eToro needed roughly 140,000 IOPS at sub-millisecond response time, and nothing Azure-native could touch that. Without solving storage, moving the rest of the stack to the cloud wasn’t going to move the needle.

2. They tested Silk for six months before trusting it with real money

No leap of faith here. eToro spent over half a year throwing the kinds of failures at Silk that actually worry them — dropped nodes, host issues, zone outages — before letting anything critical near it. In Yitzchak’s words: “We were very careful, and we really made Silk work hard in the POC.” Disk layer first, then server by server, tuning as they went, until performance held up under pressure.

3. Silk found eToro savings they weren’t looking for

Somewhere along the way, this stopped being just a storage deal. Silk pointed eToro toward Azure’s newer E-series VMs, which beat the pricier M-series they’d been running. Then helped them move TempDB off Azure’s ephemeral disk and onto Silk storage, cutting out an entire VM disk cost. Silk even worked directly with Microsoft’s engineering team to help shape a new, cheaper VM option for its own capacity nodes — and passed the savings on.

4. Upgrades used to require a babysitter. Now nobody even logs in.

On a platform where a single I/O hiccup costs real money, “we’ll patch it overnight” wasn’t going to fly. Early on, eToro watched every upgrade closely. Today, Silk’s engineers run them end to end, and eToro just checks the result afterward. The same goes for Azure’s own hiccups — planned maintenance, surprise host failures, the works. Silk’s system catches the warning, moves the affected node, and emails eToro when it’s handled. No one has to be awake for it.

5. Even eToro’s most skeptical team came around

Yitzchak didn’t sugarcoat it: eToro’s IT team wasn’t thrilled about bringing on a storage vendor at first. Two and a half years later, his words were “genuinely a pleasure to work with.” That’s the kind of result Forrester set out to measure independently in its Total Economic Impact (TEI) study of Silk customers — cost savings and performance gains reported in aggregate, not tied to any one company by name.

Curious what those numbers actually look like? Register for Part 2 of our Forrester TEI webinar series to see the full breakdown of what enterprise teams are saving by scaling cloud performance without the chaos.

eToro’s story isn’t a one-off. It’s a pattern we’re seeing across the customers we work with, especially in financial services, where downtime and latency aren’t inconveniences — they’re line items.