How AQR migrated over 1,100 VMs months ahead of schedule using Amazon EVS

Technology is central to how AQR develops investment models, manages portfolios, and serves clients. As part of a multiyear strategic cloud transformation, the company steadily migrated workloads from two on-premises data centers to AWS. Although most applications had already moved, the VMware-based virtual desktop infrastructure (VDI) remained the final workload that prevented AQR from fully retiring its data centers.

Unlike many other workloads, AQR’s VDI underpinned everyday operations. Additionally, the infrastructure team had built extensive VMware knowledge, customized tooling, and established processes that employees across the company depended on daily. Replacing those tools—or asking teams to learn a new operational model—would require significant effort while adding little business value. “We weren’t interested in reworking what we had in place,” says Michael Raposa, managing director and head of infrastructure at AQR. “Spending all that time converting to a new system wouldn’t improve the end user experience, bring in clients, or generate revenue.”

Therefore, to complete its cloud transformation, AQR needed to migrate its remaining VMware workloads to AWS without rebuilding years of institutional knowledge or disrupting the employee experience. After evaluating several options, the company chose Amazon Elastic VMware Service (Amazon EVS), which provides a fast path to migrate and operate VMware workloads on AWS. That way, AQR would be able to run its VMware workloads directly within its Amazon VPC.

A longtime customer of AWS, AQR became an early adopter of Amazon EVS before its general availability. To confirm that the service met the operational requirements for its VDI, the company worked alongside AWS throughout validation and deployment, providing feedback that helped inform the feature road map of Amazon EVS.

To reduce disruption, the company migrated desktops to Amazon EVS in phases, accelerating the pace as the process exceeded expectations. Employees logged in to the same desktop—with the same applications and settings—before and after the migration.

AQR created a distributed architecture by spreading 36 Amazon EVS hosts evenly across three Availability Zones so that if one Availability Zone became unavailable, the other two could absorb the load. “We wanted to maintain continuity for our end users,” says Raposa. “From their perspective, all they saw was a higher-performance, lower-latency, lower-jitter machine on Amazon EVS, and nothing else changed.”

By overseeing its virtual machines (VMs) and off-loading infrastructure management to AWS, AQR preserved its operating model while removing the need to maintain on-premises hardware. The Amazon EVS team stayed closely engaged with AQR engineers throughout the migration, working transparently to overcome technical challenges and assisting with code updates as needed. This collaborative approach reinforced AQR’s confidence in Amazon EVS, helping the company accelerate the pace of deployments.

AQR migrated more than 1,100 VMs, marking the final milestone in its multiyear cloud transformation. “The project was going so smoothly that we finished 2 months ahead of schedule,” says Steve McKeon, managing director at AQR. Minimizing disruption while achieving workload resilience, the new architecture expanded disaster recovery coverage from a subset of critical employees to the entire workforce.

With the migration of its VMware workloads to AWS complete, AQR retired its remaining on-premises data centers, reducing the operational burden of managing physical infrastructure. As a result, the infrastructure team can now focus more of its time on initiatives that provide business value while continuing to modernize its AWS environment. “The AWS team handles hardware maintenance, which makes our lives so much easier,” says Raposa.