A conservative guess:
10years$100,000annual salary30 developers=$30,000,000 salary investment
That is just a salary investment guess. No idea on the infrastructure investment(offices, hardware required pipeline software, etc)
For comparison. MVA which was a smashing success likely was way less of an investment.
https://www.hayworth-miller.com/obituaries/william-bill-heffner-jr
The depth of Bill’s tremendous contributions extended from pre-MVA SAS all the way through CAS. He was one of few who absolutely mastered Data Step compiler, WG and WX internals. Bill was an extremely supportive co-worker and friend to many at SAS!
Very sad to hear of SASWMH's passing
“ A less polite way to say that is that SAS has a long history of promoting sycophants, incompetents, and downright toxic people to management positions.”
Those who were there sadly witnessed decades of this! Also, one of us, a true “SAS R&D Mason” recently passed. RIP SASWFH!
@12k’s analysis changed my thinking.
I’ve denigrated Viya, and urged more development of MVA SAS. But if one accepts that the old platform’s decline is ongoing and inevitable, it makes more sense to place bets on the new one.
If SAS built compelling applications on Viya, they would sell, and they would sell the platform.
Alas, SAS is “constrained by product vision and the architectural talent.”
A less polite way to say that is that SAS has a long history of promoting sycophants, incompetents, and downright toxic people to management positions.
There remain decent and competent managers at SAS, but they must compromise with the others. Historically, that mix has not created compelling applications.
Well said @12k
Sadly, it's the end of the road for SAS. Both core product platforms have no future.
Any buyer will sweat the asset.
If you're at SAS and aren't financially secure enough to retire, you should be looking for other options because it's only a matter of time before you're forced to.
@12k You wrote the best summary of the current situation I have seen. Thank you.
Enhancement work like 256 -character variable names and certainly new features for key analytics PROCS definitely should have been/should be added to SAS9. This would help keep many current customers happy.
However, what realistic new revenue generating future does SAS9 have? The bottom line is open source continues to eat away at SAS9. The people currently stumping for dropping Viya in favor of SAS9 are not seeing the full picture. The SAS9 platform architecture is between 20 and 35 years old depending on which parts of it we are talking about. It was not designed for present day computing infrastructure/environments and has not proven to be amenable for exploiting parallelism, distributed data paradigms, etc.
At least Viya provides a framework for developing new AI and other advanced distributed computing features that products for handling big data, training models and advanced analytics require. Yes, there are probably better alternatives from other vendors, yet building CAS and related Viya was no small feat an represents SAS’ best attempt at a modern computing architecture.
Bolstering SAS9 could, at best maybe add $100 million in new revenue. The same could occur by building a couple of relevant new products in Viya, with the very real possibility of considerably more revenue from similar ongoing efforts. At this point SAS appears to be constrained by product vision and the architectural talent necessary to maximize this, yet given their apparent current direction it’s likely what they are banking on.
SAS9 is in decline, on its way to EOL, and no amount of propping it up is going to change that. Within 10, 15 years max, it will not even be a conversation.
“ Everyone seems to know that except for the SAS CEO and his myriad of sycophants.”
Oh well if Everyone knows.
“ To my knowledge, it was never shipped”
So you have no idea…. Cool.
What seems more relevant at this point is why is SAS continuing to act like Viya is the future of the company. They took their gamble on a new direction. It failed. Everyone seems to know that except for the SAS CEO and his myriad of sycophants. When I was near the end of my career (the first rapture), I was involved with the new 255-byte name project and extended variable types. The plug was pulled on both, because they weren't Viya. The last thing I did to one of my own procedures was a customer requested (and quite reasonable) enhancement. To my knowledge, it was never shipped, because it was V9. It is Sunday today; I am watching football. I keep seeing the Allstate "mayhem guy" commercials. I always remember the SUGI (or was it global forum?) video that blatantly ripped off that idea with the "big data guy." Cringe!! Right up there with the "power to know" singer! Maybe it is a total non sequitur, but it perpetually makes me wonder, what the ** was the big guy and all his sycophants thinking? Why are they so perpetually committed to nutty ideas and then never willing to admit that they are wrong? I guess I know what the sycophants are thinking: I want to milk this as long as I can, and I can only do that by not disagreeing with the big guy.
Well, the cost is clearly in the $Billions. The implications are unclear.
At this point I’d be moving developers off of Viya, onto SAS9 and new projects -- but what is SAS doing?
A decade is forever in the software industry. The question should not be how much was invested in Viya. That money is gone and does not seem to be bearing any fruit.. The question should be what is the cost and implications of a failed bet.
I like JG. Which means that I hope he has a better going away party than what my eyes see.
@wv This is in doubt?
Viya could prove to be the albatross around SAS' neck. Surely it must be the largest reason for stagnant revenue. Possibly Viya also could be a major factor explaining so little interest from potential buyers.
The best advocate SAS has is loyal customers. If customers are not interested in Viya, how can anyone expect buyers to have serious interest?
“That’s when corruption of the implementation started”
Strong statement. Got anything to back it up?
Sigh … siloed projects building what should have been common infrastructure or services was one of my biggest frustrations in a long SAS R&D career. This was often due to Directors and VPs competing with or disliking each other. Their executives mostly failed to deal with this dynamic which has contributed greatly to the company’s decline.
@rr I don’t know of anyone who tried to stop Viya, but people did suggest it should be made compatible with SAS9.
Even if it were only 80% compatible, that would have made it easier to sell to existing SAS customers.
SAS is full of excellent critical thinkers.
Were there any attempts to put the brakes on Viya? In poker terms, a weak hand was overplayed entirely too long...
@mz "CAS needed a native columnar implementation from its beginning. CAS design began in mid 2013... columnar is often better for analytics workloads, especially on wide flat tables when various column subsets are projected."
This was known at the time. My project at SAS had a columnar data store in 2011. Silos...
@n4 Here’s a fun fact. You know that 25% of revenue that SAS brags about reinvesting in R&D? It used to be a good bit higher. When I first started doing presentations for SAS in 1997, it was 33%, and it was almost surely higher before that. The percentage has declined slowly but steadily.
Goodnight often liked to say that SAS puts 25% of revenues back into R&D. Revenues have been around $3b for the last decade at least. What % of R&D went into Viya over the last decade? Given that SAS9 has more or less stood still for the last decade, I would guess 75% was spent on Viya.
So unless Goodnight has been bullsh-tting for years about the 25% of revenues going to R&D, my rough estimate would be 75% of 25% of $3 billion / per year = $560 million per year.
So total spend on Viya to date...probably not much change from $5 billion....what a colossal waste of money!
Case in point—CAS needed a native columnar implementation from its beginning. CAS design began in mid 2013 and Apache Arrow was not initially released until 2016. However, Open Source Parquet could have been used in place of HDAT for native storage (direct integration with Parquet was added to CAS later) using a purely columnar model.
There are some performance tradeoffs, but columnar is often better for analytics workloads, especially on wide flat tables when various column subsets are projected. For these cases, memory footprint is reduced and analytics requiring multiple passes of the data can still perform well.
There are also applications/solutions cases where the Spark I/O model beats in-memory HDAT. This, along with a lack of tightly integrated database capabilities added to CAS’ limitations. True cloud native and columnar support plus tight integration with for example, PostgreSQL would have taken a greater initial CAS design effort, yet could have produced a Viya platform infrastructure much more suited for the modern age.
Had it not been for the V9 cash cow funding Viya, it would have never happened.
Ironic that Viya was intended to ki-l the entity giving it life.
IIRC, Oliver was not only CTO but COO for much of the post-ASR implementation, and even before that he was extremely influential due to his proximity to Goodnight. He could (and maybe should) have nipped that in the bud if he wanted to. I suspect he wasn't able to, for reasons. It is the case that CAS is not the perfect solution to every problem, and not every developer (particularly the action developers) was as talented as the developers on the CAS core team. It's only gotten worse since basically everyone left.
then 200 Devs from Analytics+Data Management (C) writing CAS Actions, Interactive V* products (Java and eventually Go) plus another 100 testers and miscellaneous DevOps/IT Infrastructure support folks jumped on Viya.
That’s when corruption of the implementation started - away from its pure design to an impure one due to the influx of pet features, non-standards compliant actions, and just plain poor coding.
@ff Completely agree. I feel sad for him and the company that it did not work out.
JHG deserves credit for making such a big bet in his 70s. I wish it had worked out better.
@ba+1k5vgnpd7 Hard to say for sure but you are probably in the ballpark now.
Interestingly that only accounts for say 10% of revenue per year. Give or take a couple % in either direction.
The first 2 years of Viya was predominantly CAS development + ASRD ML Devs. Probably 30 very well paid Devs averaging $200K/year or more considering:
salary + bonus + benefits + company match 401K + employer SSA + office/computing overhead
Once CAS was functional and some of other supporting infrastructure was in place, then 200 Devs from Analytics+Data Management (C) writing CAS Actions, Interactive V* products (Java and eventually Go) plus another 100 testers and miscellaneous DevOps/IT Infrastructure support folks jumped on Viya. Then another 100 from various solutions groups. This went on for another 6 or 7 years at least.
Pretty safe to say SAS has spent well over $2B on Viya development, testing and packaging alone … that doesn’t include go-to-market and sales overhead activities. Would not be surprised if the totals are approaching $5 billion. We’re basically looking at what grew to be a massive company-wide initiative for the past 10 to 12 years.
Way more than 100k salary and way more than 30 developers.
You aren’t even in the ballpark.
30 million is a giant nothingburger.
Conservative guess is right! For one, most of the salaries/benefits of the folks working on it were probably closer to at least double that. And add a lot of person-years in testing as well.