Extracted from a thread began on @OP+1jgxdnf0h
207 replies (most recent on top)
Pretty sure DI Studio, data prep solution, etc. clients ultimately invoked CAS Table-related actions _after_ querying various external data sources, invoking stored procedures in the compute server, doing joins, etc.. NR and related Devs outside of the CAS team certainly possessed the ability to write custom CAS actions and may have done so to help optimize their Viya data prep implementation.
The basic idea as I recall was to create wide, relatively flat CAS tables extracted from necessary data sources and suitable for analytics processing. Capabilities provided by the SingleStore partnership began to blur those lines.
Perhaps someone here who also has strong Apache Spark knowledge/skills can opine as to merits and deficiencies in the Viya approach in comparison to Spark. Personally found the Spark UI to be fairly primitive, but that was a few years ago.
My understanding was that NR's team worked on the plumbing of the CAS server per the mandate by OS. If I recall correctly, there was a data prep solution in line with the plumbing, and that all solutions were to run their data through the data wrangler after being placed in memory via CAS server. Or something to that effect, if I understood it correctly.
… data management internals …
Don’t recollect NR, nor any of her team being directly involved in the development of CAS core management internals. Certainly they would’ve provided relevant requirements and feedback, yet were mostly involved in DI studio client development work in Java. FedSQL support was eventually added for CAS, yet because CAS was not designed to be a relational database, the SingleStore partnership became an opportunity to explore deeper CAS integration with a high performance distributed RDBMS — beyond the Viya data connector methodology.
Found the “butthurt” guy posting on the “/pol” board (Politically Incorrect) of 4chan:
So let me get this straight. Because USA and Russia have started actual peace talks about Ukraine, EU is so butthurt …
https://boards.4chan.org/pol/thread/497860153
Suspect he’s one of those wackjob incels.
Wow, he's back!
Look Oliver, I took it easy on you last time because you were pathetic. I won't be so kind again.
Now go home and get your shine box.
That's a nice jump in logic there.
You can have your opinion, and I can have mine. I can respect your opinion, but I ask that you respect mine as well.
Many think Viya was the death knell for SAS. I don't think it was. I think the mindset around the remark "such a person should be dismissed" speaks volumes about SAS culture and why SAS is in trouble. There's a snobby tone to it, a dismissive "othering". I met plenty of those types of people at SAS.
Who would want to do business with company full of these people for any longer than they had to? I sure has he-l wouldn't. Once presented with alternatives, I'd take one of the alternatives and get the he-l away from them as quickly as possible.
Hindsight provides so much clarity, or at least the illusion thereof. In retrospect, a way to get the funding @240+1jk94kx0x speaks of, would’ve been to initiate significant VRBOs followed by layoffs if needed, beginning as early as 2016. Clearly, JG did not want to do this, given his relative generosity and good natured loyalty to long-term employees.
Probably his age had something to do with the decision as well. When one lives into their 70s, they begin to prioritize simpler things in life like family and JG likely did not want to caused stress on others nor their families through layoffs. So SAS kept plodding along, hoping Viya would succeed. Ironically, working on some of the more leading edge and refined aspects of Viya provided the very skills and industry exposure that many of its top practitioners needed to leave SAS during the pandemic stimulus bo-m.
To free up future capital for new projects or investments, profitable public traded companies regularly layoff hundreds, even thousands of Employees from divisions, maintaining mature “cash-cow” products or lackluster/failing, newer products. historically, SAS did not do this. The times they now are a changing.
Internal political dynamics/power shifts and the extreme focus that Viya got when rolled out in 2016 are probable reasons why SAS did not pivot away from Viya to more fully embrace an open source foundation. Who else thinks JG and OS would’ve done this?
I wanted Viya to succeed; but my opinion is from outside it. I’ve been learning from the OP, who makes a strong argument that Viya was a good idea, in 2011-13.
It’s true that there are always customers who will pay for performance, and that V9 was not architected for parallel processing or the Cloud. It’s also true that, in 2011-13, few people could have predicted the explosive growth of open source.
Yet by 2015, Spark had become one of the most active open source projects, and R supported all the most popular analytical techniques of SAS/STAT.
So by that time, it was becoming clear that Viya would compete against open source. But you can’t do that; price always wins. So by 2015, or at latest 2017, SAS should have changed direction.
SAS should have built something else, something that did not compete against open source; or else strengthened Viya so that it had a safe market niche. (Compatibility with V9 could have been provided one such niche.)
But we didn’t change direction, and we didn’t invest more in Viya. Why not?
Partly, we lacked the vision. But importantly, we lacked the funding. After 2013, SAS revenues began declining against costs. We had never IPO’d, so our funding depended on one person.
That person, quite understandably, prefers to run his company at a profit. Not only does he prefer to; he must, because he must eventually sell it, and profitability is its strongest selling point.
By 2015-2017, profitability had begun to require cost-cutting. Travel was restricted. Hardware became more difficult to justify. Benefits were being reduced.
So SAS was not considering any big new investments, in Viya or anything else. Cost-cutting was already under way.
The first buyout occurred in 2018, the second in 2021. Between them, hundreds of people left, and SAS was glad for the attrition. And all along, there’s been a steady trickle of layoffs.
JG deserves great credit for making a big bet on Viya -- in his 70s. Most people would not have the courage, at that age.
Some on this forum have suggested a future in which less-profitable areas of SAS are shut down, and the resulting savings used to make another big bet.
That would be ideal. But if I were in my 80s, I would not be making any more big bets. I’d be doing what he's doing: cutting costs, and planning an IPO.
“ That's my opinion at least. The rest of you can continue with the company-supported mantras posted earlier.”
I see. When you have an opinion It is yours and yours alone. When someone else has a different opinion it is a case of following the company mantra(party line).
My opinion is that such people are to be dismissed.
Some say Viya was the last straw. I think the issue of ignoring customer needs finally caught up with JG; customers had alternatives, and JG's solutions were no longer among the alternatives considered.
We fought internally to work with customers. Blockers, left and right. Rather than hear from customers directly, features were dictated to us by Development Managers. Where did they get their feature lists?
When we got to customers, they only complained about the lack of responsiveness to their issues. Viya didn't end hamper. SAS culture, a lack of customer focus, and JG micro-management did.
That's my opinion at least. The rest of you can continue with the company-supported mantras posted earlier.
Wait, wait, wait. I’m confused. Just when did NR and her team develop the data management side of CAS?
Quick summary of this thread and OP is outta here.
Though they might’ve not seen eye-to-eye on many things, AS and OS were in general agreement that SAS needed to be re-factored and greatly modernized, Especially to exploit cloud computing and massively parallel processing paradigms. JG apparently concurred, because he spent billions of dollars trying to make this a reality.
CAS evolved from earlier technologies JG approved and OS was critical in the genesis of. JG and SAS software was/is proprietary work and the advanced analytics and modeling algorithms, etc., were considered “the crown jewels”. The thought of conceding this to open source, even if so developed at SAS, would have been considered blasphemous and JG would’ve never approved of doing so.
As a company, SAS was at its zenith from 2010-2012, when for two years running we ranked the best place in America to work in the Fortune survey and revenue was still growing faster than costs. SAS held 37% of the global advanced analytics market, greater than our next seven competitors combined. Open source analytics, especially the Python libraries were nascent at best, although R, which JG “wanted to ki-l”, had gained considerable traction for single-user work.
2011-2013 was the era in which CAS was conceived and its intense R&D effort began — when SAS was riding high in the saddle. Very few people could’ve predicted the exponential acceleration of open source data management and analytics in the short years that would follow. The die was cast to build what was ultimately branded “Viya”, the next generation of SAS atop CAS, with revamped micro services, new web clients, etc. Any open source initiatives were primarily client-side support for invoking CAS actions with results returned in a tabular format, e.g. Pandas data frames for Python.
Much if not most of the anti-Viya Sentiment present on these threads is some combination of minimizing (or outright ignoring the above facts) or living under the illusion that SAS9 could continue company growth by plodding along with a few new features being added every year, or yet another new product riding atop an incredibly bloated V9 tech stack that was brutal to deploy on premise and even worse in the cloud.
we don't need to know, but a retrospective can be useful.
“ That is where the rubber meets the road and is all the Viya history we need to know.”
True if you are a simpleton.
@1z3+1jk94kx0x Easy CompanyMan
Who just came back from vacation?
Looks like super genius counter ratio’r is back. Wonder how much HR is paying her?
OS left at the end of 2020. Hundreds of us left around that time. It was easy to find better compensation during the pandemic stimulus years.
This was after seven years of SAS revenues declining against costs. For that reason, cost-cutting at SAS was well underway.
SAS had two buyouts and many quiet layoffs. But the cheapest way to cut costs is attrition. They wanted us to leave.
Okay, that being said, we need to get back to SAS messaging:
"SAS is the greatest place to work!"
"SAS technology is the most awesome analytics technology!"
"JG is a king among men, who built the most awesome company ever!"
The biggest reason Viya failed is simple. Viya simply did not address enough customer pains to move the revenue needle. SAS, as a last resort measure, could not get Viya to stick even when it was given away. That is where the rubber meets the road and is all the Viya history we need to know.
TBC … the OP actually cares about SAS’ and Viya’s success.
What proprietary information? NDAs expire. You don’t know what the OP signed nor when.
What is being posted here can be ascertained from various press releases, public blogs, Patent filings, conference papers, LinkedIn, etc.
The OP is not the person(s) running down Viya, pointing out compatibility issues, etc. They actually care about SAS’ success.
Why do you come here and post proprietary information and intellectual property? It’s a violation of your employment and separation agreement. Do you have no morals?
OP here.
Before returning to the originally scheduled multi-part brief history of CAS/Viya, I would like to opine on recent comments regarding why Viya has not grown to a level of significant success.
Let’s rewind five years, just before Covid swept planet Earth. Viya 3.4/3.5 was in the hands of several large, long-standing traditional SAS customers. Most were only running Viya experimentally alongside V9. Emphasis was on scale, performance and the possibility of interoperating with client side source languages like Python. There didn’t appear to be a lot of negativity toward Viya, yet there was a lack of understanding among customers about how the best use it and what it meant for them.
There was interest in developing partnerships and working with large very wealthy/technologically progressive international customers on major Viya deployments. This was very challenging work, that a small number of R&D and customer facing SAS technologists were getting into. We were getting excellent feedback from some major customers as well as insight into how they could use CAS/Viya to solve some massively scaled problems that at the time, open source and other products were not able to handle well.
Deficiencies in CAS design were certainly being revealed and addressed. There was tremendous momentum among this small group of people in diverse roles, spread across the company in various geographies. In some respects, it was was like being in a growing,startup with a challenging new technology and the corresponding passion, expertise, effort, and long hours necessary to make it successful.
Then the Covid tsunami hit. It cannot be emphasized enough, how much this impacted the fledgling Viya effort over the next 18 months. Not, the fact that WFH became the default and not the fact that some employees contracted the virus and were down for a few days or weeks. No, it was the overall churn and shifts in tech hiring and the impacts this had on SAS’ internal ability to not only sustain but also accelerate the efforts that this relatively small group of Viya experts initiated and who were in large part the primary evangelists of the deep technical skills that many more SAS R&D and customer facing technologists needed to grow.
Between 2020 and 2022, SAS lost Oliver, most, if not all of the ASRD Machine Learning team, all of whom are PhD level and fundamentally world class in the knowledge of ML algorithms and as importantly how they were implemented in CAS and our rapidly growing relationship with Nvidia and interoperability between distributed/threaded CAS and Gpu-based CUDA processing. We also lost many of not most of the key players on the Gopher team, along with other founding CAS design/development team members. Some took the VRBO but many simply left out of frustration with various aspects of SAS as a company and particularly compensation when compared with what their skills were worth to other tech companies for which they could now work remotely, without having to relocate to the West Coast.
That’s before we begin to talk about very important folks who left during the so first and second “rapture” when in total over 100 veteran systems and analytics geeks left SAS for good. Some of these folks no doubt retired for good, but others also “cashed in” on many Tech companies “gold rush” to hire the best engineers they could.
While it is certainly the case that many excellent, brilliant and highly productive developers and customer facing technologists remain at SAS, the above losses are incalculable when one considers the lucidity and rapid ability to address problems, amend compatibility issues and work directly with customers on Viya POCs possessed by those who departed.
This, IMO, is a significant reason why Viya lost momentum and has likely never regained growth that could have been. Not saying, such growth would’ve resulted in SAS currently being at $4B+ annual revenue. However, things would very likely be better than they are, and more importantly, the company in a better position to continue innovating and growing.
it wasn't just open source that disrupted the market. various companies and products (some built on open source, some not) emerged in the 2011-2013 period. for example, databricks, snowflake, powerbi. the first two are above $3 billion annual revenue (msft probably doesn't break out rev for powerbi). so much talk here about $3 billion flat (declining in real terms) revenue. to go 0 to 3 billion in <12 years is incredible. there are many other examples that arose then or later and are still emerging. what will OpenAI or DeepSeek make? who knows. what did they all see, try to do, do, that we didn't do? was something they did even attempted by us? i believe the answer is no, simply due to the nature of the innovator's dilemma. you listen to your own customers (if even that ... sounds like we did not). something comes from "below" and disrupts everything. nothing you can really do, no matter your mgmt or leadership flaws or strengths or quantity/quality of people. but the attempts (and what were the now-obvious blind spots) would be nice to hear about.
Hi! I'm Nobody. I think OS is posting here.
"Who wants the OP to return to teaching the multi-part history of CAS?"
I'd be glad; I have learned from the OP.
I still don't know what exactly went wrong. I guessed below that Viya might have worked, if we had not shipped too slowly and let the market pass us by.
But maybe the price was too high, or the apps on top of it were not compelling, or there may have been other causes.
Nobody thinks OS is posting here. Ignore the trolls.
Who wants the OP to return to teaching the multi-part history of CAS from the perspective of a long-term R&D Dev who was there for much of the journey including developing lots of CAS related code?
I’m not OS.
'The one who wants to "learn more about Viya" is the same one who will be doing the teaching about Viya's history. Spotlighting his self perceived greatness.'
No.
Viya and related products used some of my work. But I am in no position to teach it; I was distant from it.
It seems to me that the market moved faster than Viya. But I don’t know much about it, am only trying to understand what happened.
"That AS guy kept up with changes in technological direction by reading 1000 articles a day on Pulse. Let’s see you do better."
If that were true, how did he come up with articles he hadn't read already, and when did he get his work done?
"I‘d be glad to learn more about Viya and its history"
Troll alert! The one who wants to "learn more about Viya" is the same one who will be doing the teaching about Viya's history. Spotlighting his self perceived greatness. Once again.
Software projects fail in four main ways:
- We build the wrong thing.
- We build the right thing, but badly.
- We build the right thing, but slowly, so the market passes us by.
- We build the right thing, but then we change direction and blow it.
I saw all four cases at SAS Institute.
It seems to me that Viya is in case 3). But I‘d be glad to learn more about Viya and its history.
Some buyers only invest in growth companies, so would only buy the company “growing at a steady 10% clip”. Microsoft almost always buys growth.
Others invest only in declining revenue streams, so would only buy the company with “flat” growth. Broadcom has bought dozens of such companies.
So this choice is a good analogy for the sale of SAS. It can only attract the latter type of buyer.
A recurring 5-10% margin on $3b can buy a lot of a-s, and pay loads into babies trust funds. Now I'm much more inclined to believe those old stories about 'spare families' scattered hither and yon.
By definition, “flat growth” means revenues are declining against inflation, therefore declining against costs.
The choice did not specify that profit margin “remains steady”. With flat revenues, that’s only possible by cutting costs.
Costs can’t be cut forever — because you can’t cut to zero.
That’s why I’d want to know the rate of decline.
“ Depends entirely on how much I paid for said company.”
You aren’t paying a dime. You get to choose and whichever one you pick is yours.
“ “Flat growth” means revenues are declining against costs, therefore profit margin is declining. I’d want to know the rate of decline.”
Flat growth does not mean revenues are declining against cost thus shrinking profit margin. It explicitly called out a 5-10% margin . How that is achieved(cutting expenses) wasn’t specified but is unimportant if the margin remains steady alongside the revenue.
It’s either a card to play or it’s not.
If SAS IPOs, then both Viya and AI are potential growth stories. Viya is probably growing now, although from a low base; and anything related to AI may grow in future. Growth stories are attractive in an IPO.
If SAS sells privately, then neither Viya nor AI will matter. A buyer must pay at least $5-10B for the $3B legacy revenue stream. Buyers of such large declining revenue streams normally don’t care about small areas of potential growth.
An IPO generally brings a higher price than a private sale. That’s probably why SAS is taking steps to do one.