This post points out that a LinkedIn search for Viya yields only three job postings in the United States:
https://www.thelayoff.com/post/@3jfx+1oW1I0bm
What would it take to make Viya succeed?
Better architecture? Better products on top of it? Better compatibility or migration path?
Or is it better to give up, and put more resources into V9?
If SAS makes Viya compatible with SAS V9, it could sell to existing customers who want the performance benefits of a modern architecture. In that sense, Viya can be "fixed".
However, selling to existing customers doesn't create any large new revenue stream. That requires innovation, which SAS culture has always discouraged.
"...I think the real problem here is that the Good Doctor has been searching for a son worthy of his name for 20 years, and none of the candidates have proven up to the task. None of his surrogate sons have been able to run the company half as well as he himself was..."
The Good Doctor thinks he's omnipotent and he's searching for an omnipotent heir to prolong his vanity. The 'Bu---r was the closest match he was going to get.
I don't think he ran the company that well. He was simply in the right place, at the right time, with the right product, and the right grassroots support. An episode that was lucrative, but will not be repeated.
What would it take to fix Viya?
The wording of the question assumes that Viya is worth fixing. Answers to a few key questions might provide better illumination:
The Good Doctor is a brilliant businessman. He took statistical software originally designed by Jim Barr for US agriculture and grew that business to serve all types of customers worldwide.
What’s needed now is a different skill set. The original revenue stream is losing market share to open source. The only solution is to innovate new revenue streams.
Brilliant innovators are rare. Apple went through a series of leaders before they brought Steve Jobs back. Microsoft floundered until they promoted Satya Nadella. At Twitter, we can all wish Linda Yaccarino well.
@btfc+1p0p8YLr:
"The last two CTOs failed...what about the current CTO?
Is he capable of making a difference?"
Of course not. The current CTO was the head of Cyber before being named CTO. Cyber never made a profit despite consuming considerable resources, so he has not demonstrated any practical "business expertise". He failed to create a minimally viable product. After he was named CTO, the Good Doctor presented him with a choice: continue Cyber, or EOL the product and close the division. He chose to close the division and Cyber no longer exists as a division in SAS. He brought all the people who had fled R&D for Cyber back into R&D.
I think the real problem here is that the Good Doctor has been searching for a son worthy of his name for 20 years, and none of the candidates have proven up to the task. None of his surrogate sons have been able to run the company half as well as he himself was.
Too many Lucy's, with too many footballs.
"Arrrrgghhhhh!" Yell the all Charlie Browns' foolish enough to go for the kick.
The last two CTOs failed...what about the current CTO?
Is he capable of making a difference? Or is it more of the same as the last two?
Is the CTO position at SAS cursed so that everyone that tried it failed and never lasted long? Or is it because the Good Doctor did not appoint the right one to begin with?
Is it difficult to look outside of SAS and bring in someone technically capable and competent, but not arrogant, and also have both technical and business expertise in tech (software), a visionary and bring this person in as CTO? Just open the big wallet and bring in a really excellent one and reward the CTO appropriately for incentive. Surely, the Good Doctor won't want to see his legacy go down. SAS is his baby. Please do something to save this baby.
"....the real people running SAS: SAS middle management, from the managers to the senior directors."
Yep, and this group of middle management is one of the reasons why SAS is the way it is today and the owners have no idea that this is a problem (all the lying or saying what their upper bosses want to hear instead of the truth) or if the owners do know about the problem they're not doing anything to change and correct this. This middle management continues to thrive during a time when SAS is facing an existential crisis.
I’ll agree that the Big German should not be blamed for the death of SAS, but he accelerated it. Viya consumed great resources, over eight years of lost opportunity, while SAS competitors were not sitting idle.
The products built on top of Viya were not compelling. If they had been truly innovative, they could have attracted new customers, to buy the product and the Viya platform along with it.
The lack of SAS compatibility meant that Viya could not attract existing customers, either.
Both these mistakes were made by the Big German. Perhaps arrogance was their cause; perhaps he was just never well qualified for his job. Either way, he shouldn’t get a pass because of the entitled SAS culture. True, the principal owner bears responsibility for that culture. But the Big German knew that constraint when he took the job.
Better products, and better compatibility, are still achievable goals, but each requires large investments. If it were my call, at this late date I’d reduce product development to put more resources on compatibility, and hope that would be enough.
I had heard that a "personnel issue" resulted in the departure of the Big German from Single Store. The insinuation was that it was a firing.
I got on the train, and was subsequently thrown under it. "Schabenbuggered", so to speak.
@9ifc+1p0p8YLr Great post, that brought back 2 memories.
(1) BG proclaiming that everyone needed to get on the train (CAS .. eventually Viya) or be left at the station (V9). We were all scratching our heads about abandoning a $3B money maker to pursue an unproven new paradigm (on the heels of HPA's underwhelming performance in the marketplace).
(2) My manager warning his direct reports in vague but clear terms that anyone expressing any negative sentiments about Viya could find themselves without a job.
@9hyl+1p0p8YLr - Hey Armistead!
"Schabenbuggered?"
https://www.google.com/search?q=schabenbuggered+site%3Athelayoff.com
"Schabenbuggered"?
OS can't be blamed for the death of SAS. He was given a mission and a mandate. His predecessor AS, who I also liked and wished was still at SAS, was trying to architect a Viya that was modular, updatable, performant, and capable. He failed for the same reasons OS failed. And I suspect that the reason JG gave OS AS's job was because AS couldn't magic all of that out of the time JG gave him. In other words, he kept telling the old man "No, it's not ready yet".
So of course JG gave CTO to OS and the very first thing that OS said when he took over was: "We will ship software." AS then left SAS, OS purged all of AS's loyalists ("Schabenbuggered" at this very site is the reason I'm here), and then got himself named COO as well as CTO! Either of those jobs is a more than full-time commitment. So SAS shipped Viya half-baked, and embedded more technical debt in its first few releases than the company has been able to address in all the years since.
Which did OS favor? It doesn't matter because no matter what he did he was set up to fail but didn't realize it. Maybe a few months after OS took over as CTO, maybe a few more, he started having what they called "skip level" meetings with his direct reports' direct reports. Why? Because his direct reports were essentially lying to him about the status of the work they were doing, or even what they were doing, and they were unable to answer key questions about the work they were doing. OS would call individual contributors at their desks to ask them the status of specific features of CAS or Viya, because it was the only way he could get a straight answer.
Why did he do this?
Well, the long answer is: the culture of SAS had created a "failure is not an option" mindset and that ALL of SAS middle management, from managers to senior directors had learned, from long experience, that the way to survive as a "leader" at SAS is to lie a little, never be too specific, don't paint yourself into a corner, and form management "cliques" that agreed they would all go down together when decisions were made that were larger than any individual group or team.
The short answer is: you get fired if you fail as a manager at SAS.
I blame JG for that, not OS. I blame OS for taking the COO position, failing to understand the vision of Viya because of his irrational hatred for his predecessor, and his inexperience. His inexperience failed to prepare him to deal with the real people running SAS: SAS middle management, from the managers to the senior directors.
That's not to say OS didn't have his issues. He was prickly, but essentially fair unless you hadn't done your research. He was a bit of a prima donna, but also ambitious and willing to put in the hours to make it happen. His team loved him, and when he was replaced they all gradually slipped away, some retiring, some leaving their positions in ASR, and others leaving R&D for other divisions.
But there's plenty of fail to go around. AS was unable to convince JG to give him more time. OS didn't figure out what the problem really was until too late, and then couldn't do anything about it. JG was complicit, setting unreasonable expectations but also probably responding to market demand. It's such a sad story, and when the story is finally written the death of SAS will probably be attributable to a confluence of events that could only have been prevented in hindsight.
I have to take slight issue with “there was NO path for the SAS 9 users to move to Viya - they simply couldn't do it.” That’s broadly but not universally true. There are some (but not enough) exceptions, like my former products, SAS/OR and SAS Optimization. Our team did quite a lot to ease the SAS9 to SAS Viya transition, and for PROC OPTMODEL in particular the changeover is almost trivial. But in the rest of SAS/OR there’s no easy path forward, and in many cases no path at all.
Interestingly, "The Big German" only lasted 2 years outside of SAS as a CIO at another tech company, then 4 months later (the gap would suggest he got fired) is back where he was before working at SAS - as a Professor at Virginia Tech.
You gotta wonder what happened there. Behaviours that were acceptable at SAS, maybe not so much at SingleStore with investors like Goldman Sachs and Google Ventures keeping an eye on things.
It seems the guy that "The Good Doctor" entrusted with making an existential bet on the future of the company, is not cut out for working in the commercial world and has retreated to the naval-gazing world of academia.
History will show that "The Big German" destroyed SAS.
The "Big German" (love that) - his ego is what will eventually ki-l SAS. When they hatched the idea for Viya (before it had that name) - the BG's ego allowed him to think that his new and spiffy CAS server was SO great and SO powerful - and his new CAS actions were so all-encompasing... that the existing SAS 9 world would FLOCK to it - and in short order there would be NO SAS 9 customers.... just Viya.
As a result, the BG decided that there would be NO legacy SAS ported over to Viya. Even more than that - Viya would NOT support ANY legacy SAS code/functionality. the CAS actions in Viya were SO much better than the old SAS9 Procs - who would want them anyway.
This caused the bifurcation of SAS into two products: SAS 9 and Viya - and there was NO path for the SAS 9 users to move to Viya - they simply couldn't do it. The technical debt required to recreate what they had in SAS 9 into the Viya environment was so daunting and costly that it was not reaonsable to consider, even IF (and that is a big if) Viya functionally supported all the stuff in SAS 9, which it didn't even come close to.
By the time that the Big Guy (not Joe Biden - Dr. G) realized what the Big German had done, we were probably 2 years down the road into the split. Then Dr G started saying in every big meeting of R&D - "I want ONE SAS"... and the Big German would agree with him - but just kept pressing on with his new baby and ignoring Dr G. I cannot tell you how many meetings I've sat in and heard Dr G say that and the Big German agree - but nothing happened.
After the Big German departed - people finally started to try to figure out how to fix the problem, but by then Viya was so far down the road that it was (and still is) a VERY difficult job to unify it with SAS 9. I think that - the SAS9/Viya unification effort only started in earnest in late 2022 or early 2023.
One problem that nobody was willing to face was the fact that a significant part of SAS's revenue came from small departmental level enclaves of SAS inside large companies. Groups of individuals running SAS on a departmental server or even on PC's were a lot of the customer base... Well Viya architected to run in the cloud on Kubernetes - that isn't an architecture that adapts well to a departmental server or even less to the PC. Even if it works - those departments and individuals cannot support the cloud infrastructure Viya runs on - and even though they are trying to make it work - how in the heck can you deploy Viya on a PC for a desktop user???
I've been involved with SAS for 30 years - first as a SAS customer, then trying to sell SAS to our customers and then in R&D trying to make SAS for our customers.... now I'm sitting on the sidelines - "retired" from SAS and rooting for it to survive long enough for my younger friends (those over 50) to get close enough to retirement age to survive the decline.
Things are playing out pretty much like I thought they would when the Big German rose to power... it would be tough for SAS today even if this colossal mistake was not made and not compounded year after year... SAS 9 HAD TO GO AWAY - it was an architecture that couldn't come into this century, but the transition to new technology COULD have been made if egos had not gotten in the way... it's going to cost a lot of people their jobs and Dr G a legacy he wanted...
Copying from the Elon thread
@4wcg+1oW1I0bm. FTW
A few of us were working to reduce CAS’ RAM requirements by making it stream directly from high performance storage formats and data providers. This is sufficient for a variety of use cases because the entire data set often does not to be RAM resident, especially when the column projection is minimal.
Amen!
What would it take to fix adios and Viya con dios? A blowtorch.
Fascinating post @1cac+1p0p8YLr
love the sarcasm with... "introduce the change in direction that will surely save the company from itself. This time, it will work for sure. " bahaha!
As @ugb+1p0p8YLr said, Viya can't be fixed. To start with, there's nothing wrong with Viya's tech stack, with the possible exception of the Cloud Analytic Server that isn't. The problem with Viya is and always has been that it is not a replacement for SAS 9. In the earliest discussions about "Project X" it was intended that there would be two parallel... product families, we'll say... of SAS software: SAS 9 and Viya, with stated intention that they would meet at the vanishing point in the indeterminate future, but soon.
However, it quickly became apparent that the same people working on both product families had absolutely no intention of doing that. Feature parity was never a requirement for Viya. This was and is a huge mistake.
Those same people working on both SAS 9 and Viya took the existence of another product family as carte blanche to "move fast and break things", including backward compatibility. No customer wants that.
They also smoothed things out, eliminating edge and corner cases that were created to solve real customer problems that no one still at SAS even remembers, basically reverting to the older, less fully-featured version of SAS their customers escaped from by upgrading to later versions. No customer wants that.
And, to top it all off, they created a third product family when they shoehorned CAS and Viya into Kubernetes: Viya 4. So now the build pipeline produces build artifacts for SAS 9, Viya 3, and Viya 4. Imagine that complexity. Imagine the technical debt being buried daily by the constant churn of a build pipeline that almost works most of the time. FFS, those same people were so unhappy that Kubernetes does exactly what it is designed to do and moves load from underresourced nodes to overresourced nodes that their "solution" was to override the default scheduler and effectively break Kubernetes.
The only beneficiary of Viya has been SAS itself, as Viya presented an opportunity to "modernize" the build pipeline, and that created an empire from which the current CTO could emerge, fully-formed, and begin to rise to the level of his incompetence. I like BH, but honestly most of the SAS old-timers are only waiting for his tenure to come to a bitter end and for someone else (probably JP) to take his place, and with it, introduce the change in direction that will surely save the company from itself. This time, it will work for sure. And it will require no less than three or four years of re-orgs, during which at least some of the senior DevOps R&D leadership who created the problem in the first place will be thanked for their service to the company, but informed that it will no longer be required, after which they'll also take the offer and that will be the last any of us hear of them.
Stupid customers!
Would love to have been a fly-on-the-wall when The Big German (BG) told The Good Doctor (GD) he was done...
BG: "I'm leaving"
GD: "But why, BG? You're my boy, you'll have the reins one day"
BG: "But you've been saying that for years, and there's no sign you want to let go. For goodness sake, GD, just last week I had to clean up the mess you created after you tinkered with the threaded kernel code again!"
GD: "Well BG, we need you. You're Mr Viya...we ain't done yet"
BG: "Yes, I am Mr Viya, and yes it's best darn analytics platform on the market, but our customers are too stupid to realise it..."
How to “fix” Viya:
Elon has presided over multiple successful companies, including Tesla, which regularly flirts with $1 trillion in market cap. And all of his companies, Elon mandates that product design being embedded with engineering and that engineering lead.
SAS formerly was organized this way and was at the same time a highly successful and double digit growing company YoY. Return to this model, half the number of employees worldwide (including the removal of all useless suits and incompetent fools), define objectives along the lines of the business plan mentioned in a recent thread here, retain only existing stellar performers and hirer same.
Keep a skeleton team of competent veteran talent to wind V9 down, surge and bust a-s on Viya for a couple of years — including fabulous new versions of Risk and Fraud. Plan on SAS being a much smaller niche company. That’s the way forward for Viya!
The reason why Viya never took off is two fold...
In the final analysis...Viya is going nowhere and never will, it's a product that doesn't have a market...and SAS9 is slowly dying, it's a product with a market that is diminishing every day as it's user base retires or customers switch it off and force their users to up skill into more modern tech.
History will show that Viya was the last roll of the dice for SAS. The Big German bet the company on it, and knew he had lost that bet long before many others at SAS did...so he fled.
The yearly revenue stream for V9 has always been more than Viya. Assuming the Viya revenue stream is increasing(not a good assumption based on OP's research), and assuming the V9 revenue stream is decreasing(likely a good assumption), what exactly is there to entice a buyer?
To put it another way, if the V9 revenue stream ceased immediately, it doesn't take much imagination to imagine that Viya can not stand on its own.
loadable, not wearable in reference to CAS Actions
The foundation of V9 is 30+ year old technology that does not multi thread nor scale well. It's fine for running traditional SAS jobs on a single machine or remote submission. This paradigm is not being significantly adopted by the younger generation of programmers and data scientists.
Most of the solutions built around V9 are dependent upon a bloated Java midtier and the TK based metadata Server. These components are not a natural fit for public cloud computing. The solutions built to top them are aging, and in need of infrastructure, interfaces, and extensibility that is difficult to impossible to provide based on the V9 stack.
The Viya infrastructure (as I last understood it) is fundamentally the highly paralyzed CAS Analytics Server with its Actions (wearable, data management and analytics specific programs similar to traditional SAS procedures) whose syntax is supported by popular open-source client languages; the TK Compute Server which allows traditional SAS code streams to run; and the Viya Microservices written predominantly (if not completely by now) in GO with a PostgreSQL Persistence Layer replacing the lower level pieces of the V9 Metadata Server.
Viya engineering is supported by git + modern DevOps infrastructure for CI/CD. Not sure how far they've gotten with V9 for this. When I left SAS, the V9 stack was still being source, managed in CVS and significantly built/tested with SDS tools.
These Viya components are generally well thought out and engineered. With continued extension and work, they could support a variety of innovative analytics+data managementinterfaces and products. Unfortunately, management has often stood in the way of getting this done with the necessary concerted effort and additional talent, that is expensive and historically hard to come by, would be necessary to take this infrastructure next-level.
SAS has lacked the will to put the correct people in place to make these things happen.
SAS 9 is still bread and butter (and more) and without it SAS will collapse for sure.
Not many (or enough) customers went to Viya but Viya hogs more resources and attention at SAS.