Balancing EBSCO’s Hybrid Cloud OpenStack Load
EBSCO Information Services, a content discovery service for researchers and educators, has an IT development staff of 525 working in an OpenStack, DevOps, hybrid cloud environment. They build and frequently update “discovery services” applications that support EBSCO’s global customer base that submits, during peak periods, an average of 4.5 million information transactions per day. The company also provides library subscription management services for more than 360,000 serials, including more than 57,000 e-journals, as well as online access to more than 800,000 e-books. It’s a high-performance environment that demands speed and nimbleness.
Three years ago, EBSCO moved toward OpenStack to support self-service infrastructure provisioning for developers while the company began to build out a hybrid cloud environment, with Amazon Web Services on the public side of the equation.
“We went through many iterations, many different levels of service to the development teams, and kept on refining the resiliency of the environment until we eventually settled on a highly available installation on the Juno release of OpenStack,” EBSCO IT Architect Nate Baechtold told EnterpriseTech, “and by that time we had a good bit of adoption from some of our development teams. We had people spinning up infrastructure in OpenStack on demand and then servers would be spun up to put the code on it to run tests on those servers. So the entire infrastructure was being provisioned on-demand and torn down when it wasn’t needed.”
EBSCO’s development team liked the new, faster provisioning model. IT management liked it too because it supported the company’s goal of a faster, more agile DevOps software release model. But the empowerment of self-service provisioning raised the need for improved management capabilities integrated within the environment to ensure that the new OpenStack applications would actually, well, work.
The team looked at a variety of load balancing software products and settled on Avi Networks’ Vantage Platform because it did more than just load balancing.
“It solves two very big problems for us,” said Baechtold. “First, self-serviceability and automation – being able to automate every aspect and to do it safely; and second is the insight and analytics engine, to get much deeper and higher quality insights into how our applications are performing that are behind the ADC load balancer, an end-to-end picture to be able to self-service every portion of load balancing.”
The analytics engine provides insight into how individual services were performing and also how better to load data for machine analytics. “Avi became this APM tool for us where application teams now can run performance tests to see the outcome of average results and outliers that weren’t detected,” Baechtold said.
Chandra Sekar, Avi’s vice president of marketing, said the Avi Vantage Platform creates a programmatic model that mirrors the capabilities of private cloud platforms, providing APIs that can program the load balancer on a commodity x86 server, rather than a specialized ADC appliance.
“I could be a developer writing a multi-tier application,” Sekar said, “and before I hand it off to the IT operations team to deploy the application in my development and QA environment I may want to spin up load balancers to characterize its performance under load. I have the ability to do that with a product like Avi because I don’t have to deal with the procurement of specialized hardware or go to my IT team to do my development and test validation. The process for IT is so much simpler because they have centralized administration for all of these distributed load balancers in the environment.”
He said Avi separates the control plane from the data plane of application delivery, enabling a centralized management console that gives visibility into all the load balancers that administrators have deployed throughout the environment.
“EBSCO has been growing fast and we have clear plans for working with Avi,” said Baechtold. “Our goal is to do live installations in our OpenStack environment and to make it generally available to all workloads in OpenStack. Once that is done, we will make it available for everything – inside and outside of OpenStack – and we will displace our traditional load balancing solutions with Avi. We will look to Avi to give us a continuum of capabilities between our private and public cloud in AWS. Additionally, Avi’s support and vision for containers and PaaS solutions like Kubernetes aligns very well with our own.”