Cloud computing is the undisputed champ, when it comes to convenience and agility in enterprise computing. According to the 2016 “RightScale State of the Cloud Report,” 95 percent of businesses surveyed have a cloud strategy in place.[1] Initially, many of those implementations are likely software-as-a-service (SaaS) offerings for email, customer-relationship-management (CRM), and other applications that can easily replace on-site app deployments. But companies are increasingly tempted by cloud-based options to replace or supplement the infrastructure in their on-premises data centers.

Cloud-based infrastructure can be a great option for companies looking for ways to quickly provision resources without investing heavily in physical infrastructure, support, and management. Many cloud-seekers use virtual deployments from services like Amazon Elastic Compute Cloud (Amazon EC2). Virtual infrastructure is quick and economical to provision, but it might not offer the control, security, and—in particular—performance that some businesses need.

Bare-metal services, like Rackspace OnMetal I/O v2, offer the promise of real-time or near-real-time performance for a variety of needs, such as online transaction processing (OLTP) for payments processing, inventory control, enterprise resource planning (ERP), or customer-relationship management (CRM). Bare-metal speed is also critical for businesses using online analytical processing (OLAP) databases for data mining and analytics.

Testing Bare-Metal and Virtual-Server Performance

Is there actually a performance difference between virtual and bare-metal servers under load? We wanted to know the answer, so our engineers put together a series of tests to compare high input/output (I/O) database performance for OnMetal I/O v2—a bare-metal cloud-server deployment—and Amazon EC2—a popular virtual hosting service.

Our testing pre-populated a large database, generated TPC-C benchmark workloads, and then scaled up virtual users over time to measure transactions per minute (TPM). The goal of our testing was to come as close as possible to duplicating real-world I/O performance in the lab.

Stay Tuned for the Results

We’re putting the final touches on our test report now, but initial results show some surprising performance differences. Check back soon for the full report!


[1] RightScale. “RightScale State of the Cloud Report.” January 2016.

Share this: