Measuring SLA Performance Is Still Largely Do-It-Yourself
by February 2, 2018

Filed under: Networking Technology, Performance Monitoring

The Linux Foundation recently announced the LF Networking Fund to combine the multiple open source projects it’s got in the works. These include Open Daylight, Open Platform for NFV and the Open Network Automation Platform, among others. Part of this initiative includes standardizing projects on a single architecture and processes. These various projects in development may bring interesting new technology to market; the Open Network Automation Platform, for example, offers blueprints for carriers and service providers. The platform code, usable in pieces or as a complete platform, manages physical and virtual network functions along with hypervisors and cloud environments. An API manages third-party integrations.

As lots of us have learned the hard way, you get what you pay for—which is very disappointing when you picked the cheapest option and it doesn’t work. This expert’s advice argues that not all ISPs are created equal, and that choosing the best local provider makes all the difference for end-user experience. One key point: there aren’t performance-based SLAs currently available for ISPs, which can really cause problems when you bring SD-WAN into the mix. One tip here is to try to ensure there’s only one Tier-1 backbone to reduce the number of hops and need for peering. Finding the number of connected Autonomous Systems can give you an idea of how many direct connections a network has—the more the better. We know that hop-by-hop network monitoring can get you the details you need. We’ve certainly heard some wild stories on the topic of poorly performing ISPs, like the $1 million credit this customer got from their ISP after they used AppNeta to prove the cause of performance issues.

On the top of the list of apps that need modern monitoring: Here are some basics on managing the performance of enterprise video apps. There are some good big-picture tips here on the essentials of ensuring good performance for video app users. These include using new, or newish, devices that can support these apps. The network, of course, is a big element in good video—one piece of advice is to configure switches and routers properly and reduce packet loss. If cloud is part of the video puzzle, it’ll help to see if the provider’s points of presence are close to the office. We’ve heard from customers that voice apps can be similarly problematic to video apps, and as frustrating to end users when they don’t perform well.

Finally, this week in public cloud: The network speed of EC2 just got a little faster, with AWS removing the 5 Gbps limit on network connections to and from EC2 instances. So connections between EC2 instances and S3 resources, plus EC2 instances in different availability zones, will both now be faster. There are other new features available in AWS now, many announced at re:Invent. As you’re building your cloud with AWS, you’ve now got more details to consider and potentially some on-premises adjustments to make.