Watch the Numbers: WAN & Performance Metrics
Haseeb Budhani is VP of Products at Infineta Systems, a provider of Hyper-scale WAN optimization systems, where he is responsible for overseeing Infineta’s product marketing and management and overall product roadmap.HASEEB BUDHANI
Channel surfing in the middle of the night, I recently came across one of my favorite mockumentaries, the wry, “This is Spinal Tap.” It’s one of those movies that is way better than it should have been, and it has become a source of a lot of classic quotes. Here’s Nigel talking about his amp and why it’s so much better than the others.
Quoting from the movie:
Nigel Tufnel: The numbers all go to eleven. Look, right across the board, eleven, eleven, eleven and…
Marty DiBergi: Oh, I see. And most amps go up to ten?
Nigel Tufnel: Exactly.
Marty DiBergi: Does that mean it’s louder? Is it any louder?
Nigel Tufnel: Well, it’s one louder, isn’t it? It’s not ten.
Marty DiBergi: Why don’t you just make ten louder and make ten be the top number and make that a little louder?
Nigel Tufnel: [pause] These go to eleven.
Blind Faith in Numbers
This scene has become a sort of meme for the clueless faith you sometimes see people putting into the meaning of arbitrary numbers. And yes, in the world of WAN optimization, vendors do seem to come up with all sorts of numbers to measure their product in the one specifically correct way that will allow them to tout some ridiculously large performance claim. I’ve seen all sorts of metrics bandied about, such as 1000x for transfer gains, a 300x “performance” boost, 50x reduction, and 100x “improvement.” Let’s accept all those numbers at face value. What then, could they possibly refer to?
Well, the facts about WAN bandwidth reduction are shown in the graph below, which depicts the relationship between percentage reduction and the multiplicative (x) improvement in reduction.
Image courtsey of Infineta Systems.
Analyze the Real Gains
As the graph above shows, any gains above 7x or 8x are marginal. In fact, after about 5x, or 80% reduction, the absolute gain in terms of reduction begins to drop precipitously. Note too that a number like 10x isn’t double the reduction of 5x – it’s only 10% more, from 80% to 90% reduction.
And 50x reduction? First, it’s not going to happen with real data, at least not over a human-measurable period of time. Second, the difference between a 5x and 50x reduction ratio is not 1000%, as it may seem, but only 18%. This is marginally better, to be sure, but with one important qualification: With traditional WAN optimization solutions, such fantastic results only happen in specific branch use cases, and the performance does not remain the same at higher speeds.
So if there are 50 or more employees in a remote location and they all receive the same email with a large attachment, and if all the employees download that one email at the same time, you could achieve 50x or higher reduction for a moment, but certainly not all the time. If the organization’s WAN speeds are in the Gigabits, the reduction ratios with traditional WAN optimization tend to fall off a cliff.
Lately, there have been a lot of claims about WAN optimization systems running at multi-Gigabit speeds. In most cases, what the vendors really mean is that their systems expose 10GigE ports, and that they can process packets at speeds of 1-2 Gbps. But what they tend to not mention is that at such speeds, reduction performance is very low. And the need for high-efficacy reduction is particularly dire when customers are pushing traffic over the WAN at multi-Gigabit speeds.
Don’t Be Misled
Organizations spend large portions of their budgets on WAN bandwidth expenditure and any solution that can reduce their WAN usage significantly is going to make a significant difference to their bottom line. Don’t be misled by large reduction ratios. Remember:
1. Reduction rates very much depend on the data being transferred over the WAN. Log data will be a lot more reducible than CAD files, for example. Use your organization’s own data when testing WAN optimization solutions.
2. Be wary of vendors promising 40x or some other crazy reduction ratios for traffic such as high-speed replication. First, establish a reduction ratio range that will address your organization’s needs and evaluate which solutions get you into that range. Going from 6x (83.33%) to 7x (85.71%) is nice, but be sure to evaluate how much impact 7x will have on your RPO compared to 6x.
3. 10 Gbps packet processing is not the same as reduction at 10 Gbps. Ask the vendor how their reduction performance fares as traffic rates go up.
Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library.