I Run Applications, You’re Measuring Packets: It’s Time to Revisit Those Network SLAs
January 12, 2017 Leave a comment
- Enterprises are increasingly focused on, and making decisions around, applications performance. WAN providers’ stock SLAs are still mired in the language of packets and frames.
- Network partners already aren’t top of mind for enterprises’ applications performance discussions despite their key role in applications delivery.
Through our regularly conducted global ICT surveys here at Current Analysis, we caught one of the most recent purchasing trend shifts early on: enterprise lines of business are increasing their purchasing influence, even influencing networking-related decisions. More accurately, our surveys have been validating a power shift that we’d already been experiencing in our enterprise inquiries.
That is, while our enterprise conversations in 2016 were still coming through the IT department, the nature of conversations has been changing. In 2015, for example, discussions were still more around the network: for example, reach and performance of MPLS and WAN services in support of enterprise applications. This year, enterprise networking conversations more frequently started with an application – for example, a problem child that needed specific performance, such as transaction latency – and worked its way backwards into the supporting networks and other services.
Network providers should be key in that conversation. Assuming the application is already optimized and the compute environment properly resourced, what’s left is the performance of the network to ensure the application is fully supported. And network operators do address applications; they can design a service to have deterministic latency, and can assemble SLAs for clients that are based on guarantees for completed transactions.
But, two issues challenge network operators here. One challenge is that operators’ boilerplate SLAs, in terms of openly published metrics, have not aged gracefully. The industry is still measuring packet loss, latency and jitter all based on various degrees of monthly averaging. Some operators do break their boilerplate WAN SLA latency averages further down to regions and even between specific city pairs, and that granularity helps. There have been some forays into these areas: for example, Tata Communications touting deterministic routing for predictable performance and Orange Business Services making a transaction SLA available with its managed WAN optimization services. But, efforts to mainstream applications performance language into stock published SLA guarantees? Not so much.
So, operators are still talking about delivering packets and frames within averaged windows while the enterprise is looking at applications performing up to par. It’s related, but not the same metric. It’s not easy for operators; every application can be different, so any delivery guarantee depends on a range of unknowns. But meanwhile, that means enterprises are stuck doing the investigative work, engaging with various carrier reps, specifying needs in detail, seeing how each operator responds, then having to decide what’s real.
That custom reach-out and feedback process creates the knock-on challenge: enterprises already don’t tend to think of network operators when thinking about applications transactions and performance. Many enterprises tell us in surveys that their networking provider is “IBM” or “Accenture” or even “Cisco,” and to them, it’s true. Carriers should be involved in IT conversations more frequently. But, it’s a tall order if the WAN continues to give enterprises aging stock SLA guarantees that align less and less with the way their business is working.