This content is part of the Conference Coverage: Your passport to AWS re:Invent 2016

New AWS application-level load balancing gives IT pros granular control

Enterprise IT users eager for more granular control in AWS get their wish with an upgraded Elastic Load Balancing service that runs at the application level.

NEW YORK -- Elastic Load Balancing, one of Amazon Web Service's older functions and a "major headache" for some customers, is getting a welcome upgrade.

Application Load Balancer, now available in all regions, was rolled out at the AWS Summit in New York on Thursday. The new service offers more granular control by running at the application layer rather than the network layer and can route traffic based on the content of a request.

The new service, targeted at next-generation workloads such as container-based applications, can perform and report more detailed health checks than the original load balancer and can collect metrics for microservices running in separate target groups. It also supports WebSocket and HTTP/2.

"I've been waiting for the Application Load Balancer for a long time," said Jonathan Miller, application security manager at a brand-protection company based in the Midwest.

Native tooling is preferable to a third-party tool from a company such as F5 or Cisco, and it will work for multiple use cases, whether ingesting data migration from on-premises or building newer applications, he added.

With the original load balancer customers must write the functionality on each node in the target group, and demand for application-level load balancing has been "very high on the request level for our customers," CTO Werner Vogels said during his keynote speech.

[Application-level load balancing] has been very high on the request level for our customers.
Werner VogelsCTO,

"This has caused major headaches for our users because you want much finer-grain control over the way the traffic is being sent to different target groups," Vogels said.

Amazon has delved into network services and application delivery controllers out of necessity as it grows its business and move to newer types of workloads, such as mobile and container-based applications, said Brad Casemore, research director at IDC, in Framingham, Mass.

"This is part of what AWS is doing on the networking front to support the workloads customers are running in the cloud, and we'll probably see this continue to be enhanced and extended," he said.

Application Load Balancer may be overkill for more traditional workloads, Casemore said. While Amazon may not see a tsunami of adoption for IT shops that aren't as cutting-edge, it is getting ahead of the market and the need for more sophisticated load balancing so customers don't start looking to other providers.

SciQuest, a software as a service provider based in Morrisville, N.C., manages its own infrastructure on-premises but is transitioning to AWS. It uses F5 currently, but of all the new services being discussed at the conference the new load balancer was of particular interest, said Mark Pugh, senior product development manager.

"It's definitely of interest to us," Pugh said. "The health of any one of the nodes in any one of our clusters is not as simple as a ping. You need to do some app analytics to determine the health of a node."

The original load balancing service, now renamed Classic Load Balancer, is still available to users. The pricing for both services is the same, though the actual cost varies by region.

Serverless step, IPv6 support

Another notable feature rolled out at the AWS Summit this week was Kinesis Analytics, which continues Amazon's push into "serverless" computing with a fully managed SQL query service for streaming data. Applications can be set up in minutes to provide real-time analytics on the massive amounts of data customers cull from websites, mobile apps and other connected devices.

The service is currently limited to the U.S. East (North Virginia), U.S. West (Oregon) and EU (Ireland) regions. A Kinesis Processing Unit, which includes capacity for 4 GB of memory, 1vCPU of compute and corresponding networking resources, costs $0.11 per hour in the U.S. regions and $0.12 in the EU.

Other updates and services added to coincide with the show included:

  • The price of snapshot storage for Elastic Block Store was reduced 47%, and the number of IOPS customers can provision was increased by 66%.
  • Customers can now bring their own keys to Key Management Service. The addition is targeted at users in more regulated industries.
  • Support for IPv6 was added for S3.
  • A Snowball Job Management API to build applications for Snowball jobs and an adapter to treat the appliance like an S3 endpoint.
  • And new plans were added to API Gateway to categorize different levels of access control and usage.

 Trevor Jones is a news writer with TechTarget's Data Center and Virtualization media group. Contact him at [email protected].

Next Steps

These ELB workarounds help improve performance

Cloud users clamor for ELB upgrades

NGINX offers ELB alternative

Dig Deeper on AWS tools for development