SLAs Need Teeth

Posted: 05/2002

SLAs Need Teeth

By Peter H. Salus 

WE’VE ALL EXPERIENCED POOR NET CONNECTIVITY — ranging from no connection, to slow connection to static downloading. To improve the quality of our connection, and our ISP relations some providers have been offering service level agreements

But, in fact there is no way of really proving out those SLAs. So, the ISPs have been providing lip service to the concept of real-time, verifiable service level within and outside their networks. Moreover, most ISPs will guarantee only their part of the network. After all, that’s all they can control.

So how can you put some teeth into SLA enforcement? The Internet is a vast network of networks. Even a simple e-mail to a nearby friend will pass through a dozen or more systems. It does me no good to know that Texas.Net or Cable & Wireless or Digital Island will guarantee their fragment of the network, if my bits must pass through networks of other ISPs where performance is not guaranteed.

The real issue facing ISPs and carriers with regard to off-net SLAs is that customers cannot tell the difference among providers when surfing the Internet, downloading streaming video or sending e-mail. If your customers are experiencing poor performance, they are not pacified by assurances that the problems are generated by other carriers carrying traffic for you. They recognize bad performance and connect that performance to the company they are paying for Internet services.

SLAs must reflect that there is a tremendous difference between the circuit switching that is inherent in telephony systems and the packet switching that transpires on the Internet.

Historically, SLAs deal with only one aspect of performance — a guarantee of availability for 99.5 percent of the time for a segment of the network, for example. (This would provide an average of seven minutes per day or 3.5 hours per month of downtime for maintenance, etc.) This metric for performance comes from the telco Five-Nines standard. Today, this won’t suffice; 99.999 percent reliability is nice, but end users need SLAs that also reflect performance in terms of packet loss, latency and jitter.

Why? Large companies employing tens of thousands in a worldwide enterprise are dependent upon the Internet for internal communications as well as external access. They may have contracts with two or three ISPs in each location, but it is unlikely that they are the same ISPs in every location. Without real and accurate information about latency, reachability and packet loss, end users just don’t know enough about access, transmission and availability.

The use of the Internet — for web services, for e-mail, for bulk transfer, etc. — continues to grow. As is the case with landlines, customers now expect to be able to reach their ISP on demand; access delays are virtually eliminated from our connected world.

Performance issues are now the problem. Enterprises will not transfer key applications to the Internet until they have faith in the performance of the Internet. This situation is a two-edged challenge for ISPs and carriers: the guarantees must cover performance from point to point regardless of whose network is carrying the packets and SLAs must measure performance in a meaningful way.

Customers are not stupid. They know that persuading any manufacturer or service provider to take on these challenges is a very difficult task. If we truly are going to see SLAs with teeth, we need to answer another question.

What incentive would encourage a large carrier to invest the resources necessary to provide meaningful SLAs? Essentially, there are two: revenue opportunities and peer pressure.

Telcos and ISPs are in business to make money. If one carrier or one very large ISP adopts a guarantee with externally verified SLAs and meaningful performance criteria, customers will flock to them — sustainable guarantees are good for business. Rather than lose customers, other carriers and ISPs will follow suit.

How long until this becomes a reality? The truth is some carriers already are exploring ways to implement meaningful SLAs. It is only a matter of time before they are standard throughout the industry.

Meanwhile, a few ISPs already have started offering off-net SLA measurements. In 2001, one very large ISP started offering an off-net SLA with metrics provided by a web-monitoring company. While this is a step forward, only a single metric is offered: latency. The basic technique is web page retrieval, which provides neither packet loss, nor reachability information.

While such a peering SLA measurement technique works, it has a few drawbacks. But these aren’t important where the big picture for ISPs is concerned. The important part of this story is that meaningful SLAs are on their way.

Once ISPs and carriers start offering these guarantees, one additional question remains. Who would be the judge or adjudicator? If carriers are serious about offering meaningful SLAs, there will need to be some form of oversight to ensure that the SLAs are being met. A guarantee is meaningless unless it is enforced.

For this reason, it is not merely enterprises, but ISPs and carriers themselves that need third-party publicly available performance measurements. Moving from a model where ISPs and carriers oversee their own guarantees to a model where an outside independent party provides oversight will engender faith in the guarantee. Such third-party measurements will keep SLAs in line with reality; and SLAs with teeth will prompt telcos and ISPs to make sure their networks — and those they peer with — are up to snuff.

This is the same sort of problem confronted by manufacturers of household appliances and automobiles. Just like those other industries, ISPs and the telcos need an independent third party to test their performance and provide public results — the Internet equivalent of “Consumer Reports,” personalized for every customer.

Finally, SLA tools and performance measurements must be close to real-time: historical pictures of past performance glitches are worthless here. It’s nice to know that there was a glitch last week that resulted in severe packet loss or extreme latency; however, it’s better to know about it in time to do something. Your customers don’t want to use SLAs to collect reparations for poor service; they want SLAs to guarantee reliable service.

Near real-time performance event detection, analysis and interpretation has been called “the Holy Grail” of Internet performance. SLAs must guarantee end-to-end performance in meaningful term. Third parties must be able to adjudicate it. Only with valid, rational measurements and SLAs with teeth will carriers and ISPs achieve real success.

At Matrix NetSystems, real-time performance measurement and optimization inside and outside the network, is not a “thing to come,” but a reality. Today and going forward, ISPs have no more excuses, nowhere to redirect the blame when an SLA is not being met, other than at themselves.

Dr. Peter H. Salus is chief knowledge officer at Matrix NetSystems Inc., a provider of Internet performance measurement solutions. 

Leave a comment

Your email address will not be published. Required fields are marked *

The ID is: 69281