A well known saying that comes from motorsport. If you wanna win the race, make sure your car survives the race first.
The fast and the furious
A long time ago on Dutch television, a new show started called “Blik op de weg”, which doesn’t really translate well into English because of it’s double meaning, but roughly it meant “A view on the roads”. The content of the long running show was showing you on-board video’s of surveillance done by undercover cop cars of traffic violators on the Dutch roads, a format widely copied across the world, but pretty original at that time.
More or less the same videos where shown every week of cops catching speeding cars and people that pulled stupid stunts and got their drivers licenses shredded on prime time television, but one episode from almost 20 seasons ago I still remember.
They did an experiment. And it was rather eye opening for me as a much younger driver back then.
Two cars would drive (race if you will) the same stretch of highway (I believe it was from Apeldoorn to Arnhem, but it doesn’t really matter), and of course the first one to arrive would win.
But here’s the catch; one car would drive really fast (200+ Km/h wherever possible), and the other one would stick to the speed limit (120 Km/h on that stretch of highway).
Of course they had fully trained drivers in the cars and made sure everything was done safely, and nothing happened that put anyone in danger, but the outcome of the race wasn’t what you would expect.
As the race started, the “fast” car made great headway for the first few kilometers of the stretch, but as it got more closely to Arnhem, traffic increased and it had to constantly brake for slower traffic, and even though they tried to reach vMax, they never really were able to constantly hold it. The driver had to be constantly on edge to make sure nothing bad would happen. The car that stuck to the limit, the “slow” car, never really lost it’s average speed however, and had a much more constant pace and relaxed drive.
After about 30 minutes of driving the last few km’s brought the cars into the city. You would think the “fast” car would have easily put a sizeable gap between the two by now but no. At one of the traffic lights just before the finish the “slow” car actually passed the “fast” car because it had a more favorable lane and approached the lights as they just turned green. The slow car had won the race.
This race netted a couple of interesting results.
First of all the “slow” car won. By being able to maintain a more consistent speed, the average speed attained over the entire stretch was actually slightly higher than the average speed of the “fast” car because it had to brake for slower traffic, or wait longer for traffic lights as the road conditions changed.
Secondly the driver of the slower car had a much more relaxed drive and arrived in that same state at his destination. The driver of the faster car was much more tired (and maybe even irritated) because of having to constantly be at a higher state of alertness and concentration, due to his higher peaks in speed.
Thirdly, the “slow” car used a lot less fuel compared to the “fast” car that was running on very uneconomic fuel usage ratios due to the high speeds. Speeds that the car was not optimally tuned to. It also burned a lot more fuel because it wasted built up speed by excessive braking, after which it had to regain that speed again by accelerating after overtaking other cars, or having to stop more frequent for traffic lights.
This episode surely made me more aware of the downsides of speeding in my car. I won’t be holier than the Pope though, sometimes I like to go a little faster than allowed and have a car that is built for it. But you won’t be at your destination sooner, you get very tired if you do it for a long stretch on the unrestricted stretches of the German Autobahn for instance, and waste a whole lot of extra fuel, just to have some fun.
But this piece of text isn’t written about the morale of speeding.
It doesn’t matter if you win by a mile or an inch….or does it?
I like to tell stories, and this one is a lead-in towards the topic of benchmarking, and specifically benchmarking the technology of the sector I’m in, or more specifically, Nutanix is in, which is the software defined datacenter space, or as we call it the Enterprise Cloud, because we believe there is more to it than just virtualising some storage in a box.
The thing is, we see a lot of benchmark reports around of our competitors, that focus just on one thing: constantly trying to drive 200 mph. And telling the customer that if they can drive 201 mph, and maybe faster than Nutanix, they have a better solution.
What’s always missing in these benchmarks is a real life scenario. Straight-line speeds are nice, but are they realistic? Are they achievable for longer periods of time? What are you actually testing? Driving on an empty 4 lane stretch of Autobahn between Nurnberg and Munich in the middle of the night (got to admit, I love that stretch), or are you trying to maneuver in the rush hour traffic from my home town of Utrecht and trying to get to Amsterdam?
Almost everyone likes to show you how fast they can get to Munich and drive 200mph in the middle of the night in their Audi R8 for 2 hours straight and hope the tires don’t blow.
But if your workload is to transport 2 pallets full of Nutanix kit to a new customers’ datacenter during rush hour, you need something a little different than a R8. You need something that can handle the load, drives smoothly and is as economic as possible, without breaking down.
Does this mean Nutanix performance can’t be that R8? Oh hell yes, we can be that R8. We can drag race everybody and win reliably. But we are not only that R8, we like to cater to all the workloads, and also be that Scania. Or that MPV that drives your family around safely, and gets you to that nice vacation location in the southern part of France without too much stress, and prevent a half way visit to that dodgy car repair shop in the middle of nowhere that rips you off for refilling your engine coolant after you overheated the engine because you were racing all the way, and getting you to your destination with a huge delay instead of arriving on time and rested.
Keep an eye out for some upcoming Nutanix street performance based benchmarks that focus on the performance in real-life situations of your applications. And no, we are not going to compare ourselves vs vendor A, or vendor B. You can make the comparison yourselves quite easily by just thoroughly reading the different reports and looking at how the tests were performed, because that’s where the differences are.
Different workloads require different tests. Your workload is not IOMeter or HCI bench. Your workload does only not consist of random read 4k blocks.
Your workload is a database, a VDI deployment or an Exchange server, or a mix of all those. How about making snapshots and backups during production time? How does that affect things? Your workload is a user entering data into an app, reading an email, or closing an account and making you money, or helping a patient, and that workload needs to be run reliably all the time, with good fuel efficiency and without stress for the driver.
The reliability should be built into your architecture, the fuel usage is your budget, and the driver is you.
So stay tuned, buckle up and get ready for some true Nutanix street performance!