Here's what the judges said...
"Based on the provided benchmark data, SAS Viya used 87% less computing when running “like” models against linear, logistic, random forest, and gradient boosting algorithms.
When training models, SAS Viya appears ahead of the competition in enhanced efficiency, cost reduction, scalability of models, and closing the time-to-insights gap."
Yawn. Will customers make a buy decision on it? Nope.
The time and energy SAS marketing (or whomever is tasked with this cr-p) puts into rigging up these silly "benchmark" tests is pathetic, especially when it's some small, dying publication like InfoWorld.
If Gartner ratings don't move the needle on decisions, this certainly won't.
"deceptive marketing campaign is the name of the game."
Product richness and usefulness sells better than deception.
SAS Viya - InfoWorld winner and marketplace loser.
That InfoWorld award matters little to those who got laid off.
"I spent a short stint at SAS and from my time there it’s hard for me to tell if stunts like these are the result of pure technology ignorance or an intentionally deceptive marketing campaign."
SAS maybe strange and stubborn but not d-mb so it can't be pure technology ignorance. They just want good positive news in the media so a little bragging, exaggeration, or manipulation albeit deceptive marketing campaign is the name of the game. Other companies probably do the same.
“Worth noting : the benchmarks being referenced are the most nonsensical thing I’ve ever seen.
The “benchmark” compares Viya and Spark - two supposedly big data and distributed compute frameworks. Yet the benchmark compares the two on relatively small data and on a single machine.
This is the equivalent of comparing the speed of two airplanes by racing them on the ground.
To further prove this point - the benchmark actually shows that while Viya outperforms SparkML it is way outperformed by the single machine version of the open source libraries.
This makes sense since again - you would not expect to see a performance gain from a distributed processing framework on a single machine so the single machine libraries are more of an apples to apples comparison.
I spent a short stint at SAS and from my time there it’s hard for me to tell if stunts like these are the result of pure technology ignorance or an intentionally deceptive marketing campaign.
Curious what those of you who spent more time at the company think?”
If you are a nerd like me and want to validate the benchmarks on your own you can download the full benchmark report showing what is mentioned above here: https://futurumgroup.com/sas-viya/
Worth noting : the benchmarks being referenced are the most nonsensical thing I’ve ever seen.
The “benchmark” compares Viya and Spark - two supposedly big data and distributed compute frameworks. Yet the benchmark compares the two on relatively small data and on a single machine.
This is the equivalent of comparing the speed of two airplanes by racing them on the ground.
To further prove this point - the benchmark actually shows that while Viya outperforms SparkML it is way outperformed by the single machine version of the open source libraries.
This makes sense since again - you would not expect to see a performance gain from a distributed processing framework on a single machine so the single machine libraries are more of an apples to apples comparison.
I spent a short stint at SAS and from my time there it’s hard for me to tell if stunts like these are the result of pure technology ignorance or an intentionally deceptive marketing campaign.
Curious what those of you who spent more time at the company think?
"Wait... "InfoWorld"? So, SAS is excited about accolades from some company that most people have never heard of?"
InfoWorld has been around as a publication since the 70's. No idea how credible or not credible it is these days (do people still get magazines?) but if you're in a computing field and are over 35 or so and 'have never heard of' it I have a pretty good idea how credible YOU are.
SAS is #25 in TIOBE's Program Language Index.
https://www.tiobe.com/tiobe-index/
https://www.tiobe.com/tiobe-index/sas/
Wait... "InfoWorld"? So, SAS is excited about accolades from some company that most people have never heard of?
In other news, Bob down the street said in his HOA newsletter that SAS is amazing based on the data that SAS provided him. Whew! Now just sit back and watch those phones ring with new orders!
The Forester was able to pick and choose some of our best engineers to work on Viya. For this reason, I absolutely believe it performs fast.
I believe Viya has two possible markets:
So I believe that Viya performs fast, but is not selling fast, because it faces price barriers in both markets.
In the SAS market, the price barrier might be removed by making Viya 90% compatible with SAS. There is evidently an ongoing effort at this, though other posters have called it a hard problem. And as others have pointed out, some SAS users will never switch, because they run SAS on departmental servers or PCs that can’t benefit from Viya.
In the R and Python market, the price barrier can't be removed. But some users will pay well for fast performance — big banks, for example. The suggestion by other posters to rearchitect SAS/Fraud and SAS/Risk on Viya is evidently being implemented, and I expect these products will sell.
Please correct me if you know better. I believe Viya can target two partial markets: SAS users who can benefit from a modern architecture, and R and Python users willing to pay for performance. Users who run SAS on departmental servers or PCs require a separate product that shares code with Viya.
"Based on the provided benchmark data..."
Full stop right there.
Love this. This reminds me of the diamond council recommending how much to spend on an engagement ring...
Okay, Viya performs fast. So...why is it not selling fast?
Where is the metric on "$ / second of (compute + loaded labor rate) time"?
That's where the value lies. Is it there?