Extracted from a thread began on @OP+1jgxdnf0h
207 replies (most recent on top)
“ Looks like we should now close this thread and declare a unilateral cease fire”
No, it doesn’t look that way. I think it is just getting going and getting good.
The 35-hour work week was widely publicized as one of the primary company perks. The advertised reason was that 35 hours of focused, quality work was the expectation because working beyond often was just an inefficient use of time. Quantity does not imply quality. If you chose to work more, and that work was productive, then thank you. Really. Your work ethic is good, and you were likely pursuing problems that were of interest to you. But that choice was yours. Unless you negotiated for additional pay for that work, the idea of being "underpaid" is one that can be debated.
"Viya might be interesting cloud-wise if a customer can actually run more than one user ;)"
Wait.......what?!?
“ Some people were comparatively incompetent, but most were only required to work 35-hour weeks. Even a great leader would find it hard to motivate great work from people who aren’t being paid for it. ”.
Many of the “comparatively incompetent” rarely worked a full 35. It’s highly unlikely most of em could have ever been motivated to do “great work”. Meanwhile a significant number of us doing 45 to 50 hours solid a week consistently were underpaid for our efforts.
… sound’s like someone(s) experienced SAS’ sordid history of customer deployment fiascos.
Looks like we should now close this thread and declare a unilateral cease fire because apparently SAS has been slowing going to he-l in a hand basket for the past 25 years now.
FWIW, Hadoop was a steaming pile of 💩 without any assistance from SAS.
To summarize we have the rouge’s gallery of:
SPSD -> TKTS -> HPA -> LASR -> CAS
All attempts at solving the long standing problem of parallelizing data with SAS. Apparently JG should have put someone else (@hb+1jk94kx0x perhaps?) in charge instead of AG and then eventually OS. In fact few others did take their turn at it and failed to deliver anything of substance 🤦♂️
“His personality and the way he treated people meant he couldn’t rally the rest of the company behind him or the rest of the company was comparatively incompetent and just couldn’t align with him or keep up.”
Both opinions are correct. I had a couple of brief encounters with him, and immediately decided that I didn’t want to work for him. A good leader attracts followers, but he sometimes did the opposite.
Some people were comparatively incompetent, but most were only required to work 35-hour weeks. Even a great leader would find it hard to motivate great work from people who aren’t being paid for it.
Perhaps a less arrogant leader could have succeeded with Viya, if they had been willing to listen to customers who said, this isn’t what we want, it needs some changes.
But a leader still needs followers. Perhaps it had become too late, in SAS’s complacent culture, to organize enough hard workers to produce competitive software.
"The arrogance and rudeness was extraordinary, but it was his lack of willingness to actually listen to customer feedback that was his biggest downfall."
A thousand upvotes to that! Was OS intelligent? Yes. But that was completely overshadowed by his arrogant attitude of being unwilling to listen to customers. Intelligent, yes. Smart, heck no.
Anyone who had a customer facing role at SAS likely agrees that OS was a catalyst to driving customers AWAY FROM SAS. Hard to like that if you loved SAS.
OS was Big Jim's largest mistake.
From where I stood it was nothing more than the third attempt to justify the billions spent on the failed HPA/Hadoop/LASR cr-p.
Hear, hear!
Ann told me how excited she and Jim were about seeing OS groomed to be the future leader of SAS. That’s all anyone needs to know about the confidence that was placed in OS
Is it though? Is it really? Because he was either shown the door or took exit on his own. Were they very excited about that?
His personality and the way he treated people meant he couldn’t rally the rest of the company behind him or the rest of the company was comparatively incompetent and just couldn’t align with him or keep up. Your opinion probably falls into the latter camp while many here express the former opinion.
- CAS/Viya is so much more ... any intelligent person knows so
Perhaps but it's tough to tell when the "intelligent" people can't present their point using remedial logic (see ad hominem, "against the man"). Unsupported claims justified by threatening tone? Aggressive attacks against naysayers? Man, I'm having flashbacks!
I willingly programmed those god-awful actions for years so I'm clearly in the d-mb category. Pray tell: what does CAS offer besides clumsy distributed computing and a new set of horrible apis?
Viya might be interesting cloud wise if a customer can actually run more than one user ;)
- Was CAS/Viya designed perfectly? NO
Hee hee. Was CAS even designed? From where I stood it was nothing more than the third attempt to justify the billions spent on the failed HPA/Hadoop/LASR cr-p.
Viya was necessary for the CI/CD infrastructure.
"As to his bedside manners with customers, likely different story. A lot of R&D geeks are challenged in that respect, yet many venerable long-term SAS customers treated us like rock stars at conferences anyway. "
That's an interesting observation you make. It does not surprise me one bit, the loyal SAS users from our customer base, were VERY loyal as they built their careers on SAS. But they're dying off - quite literally, I went to 4 funerals of loyal SAS users in the last year.
Maybe that's why OS found it so hard to cope with customers who challenged him and asked him difficult questions, as most of the time (particularly in NC) he was, as you say, treated as a rockstar.
Further afield it was a different story. I for one, found his behaviour in front of customers to be utterly deplorable, and deeply embarrassing. The arrogance and rudeness was extraordinary, but it was his lack of willingness to actually listen to customer feedback that was his biggest downfall.
OS was the son that JG wished for.
CAS/Viya is so much more than what you’re describing and any intelligent person who actually has taken the time to learn it knows so. At that point in its history SAS had a long since given up on creating simple consumer grade analytics/visualization software for the masses.
Was CAS/Viya designed and built perfectly? NO. Was it a valiant attempt at creating a new generation of industrial strength, scalable analytics and ML, adaptable in cloud native architectures? YES, yet for a number of circumstances (enumerated in several installments over the last two years), the ongoing effort necessary to make it successful in the market fell short, while in the same timeframe, the market as well as related tech trends rapidly shifted.
At some point 2010-2012 JG anointed OS as “the most important employee in SAS”. I personally am deep in the history of SAS and acquainted with the Goodnight’s on a first name basis. On a walk around campus sometime in 2010-12, Ann told me how excited she and Jim were about seeing OS groomed to be the future leader of SAS. That’s all anyone needs to know about the confidence that was placed in OS.
I also worked directly for OS and aside from a deserved, figurative “boot in my a-s” on a couple of occasions, I found him to be a very effective and engaging leader with an extremely high level of situational awareness when it came to technology and the people who built it. As to his bedside manners with customers, likely different story. A lot of R&D geeks are challenged in that respect, yet many venerable long-term SAS customers treated us like rock stars at conferences anyway.
The idea was to make CAS Action syntax flexible and dynamic enough for virtually any parameterization required for even the most complex advanced analytics, ML, etc. actions/procedures. CAS action parameters are defined using a Lua-based syntax parsed on the CAS server side. On the output side, CAS tkctb results tables are self-describing objects that can be effectively marshaled to any supported client, including CASL, Python, Lua, Java, R, etc.
Can someone explain the genesis of the CAS action syntax? The mind-numbing "complexity" of it continually bewildered me and I often wondered what customers thought of it.
“ Characters in CAS are UTF8 native so maybe that’s what you mean. ”
No, I think he means Workbench.
- CAS was overall a sound strategy
Yup, what the world was waiting for was another 4gl. More convoluted syntax, please! Oh, large server farms ...
- - OS was a genius among geniuses
Perhaps but he was an awful CTO and he destroyed SAS culture without adding substantial benefit.
He su-ked as a boss.
"What got you there will get you no further."
- From 'Necessary Endings' by Dr. Henry Cloud
So much Viya sizzle! Yet, gross revenue is still flat and has been for years. Inconvenient facts. Is there a better metric than that to measure Viya's underwhelming impact?
Fortunately, but only recently, CAS has come together with MVA/TK SAS.
From its beginning, CAS was composed of TK extensions and built upon the TK core and many existing TK services. Characters in CAS are UTF8 native so maybe that’s what you mean. MVA is 40 years old. Talking about time to move on!
CAS internal APIs were certainly incomplete early on. However, to say that it’s design was limited to embarrassingly simple parallel RISK computations is a gross misstatement of fact. That might have been the case when CB and the initial Building S crowd “took their wack at it”, but certainly was not true once OS enlisted his hand-picked team to begin the real McCoy, fall 2013.
By 2015, CAS provided a generalized multi-threaded SMP and MPP compute framework with failover capabilities. Were all possible patterns of usage covered? NO. Was CAS capable of supporting a number of cases? YES.
Oliver and the team were certainly open to and worked hard to build out new requirements as product teams began to adopt CAS as their distributed data and compute engine.
Remember, compatibility between V9 workloads and CAS was NOT initially mandated. CAS and the products built atop it, ultimately branded as SAS/Viya were considered part of next generation SAS. The idea was to get cloud native, showcase new capabilities and then backfill with compatibility as customers, both new and existing provided feedback on Viya.
CAS was overall a sound strategy in the time it was conceived
what would geniuses like you recommend SAS to have built?
CAS was born in a special research group focused on the embarrasingly simple parallelism of RISK evaluation. It was a limited proof of concept that prematurely went production. (That showed both by its unfinished internal APIs and it unfinished exterior usability and suitability for products to be built on top of it.). It’s designers eschewed compatibility with the existing system, ignoring the problem of customer migration.
In what way is that a sound strategy?
How about building something that customers would have wanted?
Fortunately, but only recently, CAS has come together with MVA/TK SAS. Unfortunately, that melding still leaves much to be desired because of the mismatch in system designs. Still, good people continue to try to fix this mismatch but the job is Herculean and Heracles is nowhere in sight.
the innovator's dilemma doesn't render your technology product obsolete b/c you somehow failed to work hard and recruit smarter people, or because everyone else worked only 35 hours per week. unix vendors, blackberry, kodak, blockbuster employees and so on could've worked 3500 hours per week (if theoretically possible), and it wouldn't have made any difference. you should still try to get more of the smartest, hard working people you can, sure, but most of the world's smartest, hard working people will always be working outside your organization. it's necessary to recruit some of them, but nowhere near sufficient, especially in the face of the dilemma. there were plenty of hints this dilemma had started by 2012, but even with the "best" management and the "smartest" engineers, there is no guarantee some smarter company could've navigated it. well, we've rehashed this point so many times as well. good luck, all.
“ The industry has spoken and has moved on. Time for us to do the same!”
And yet here you are
What was the “right product” then? Without revising the history of what actually WAS within and without SAS from 2010-now, what would geniuses like you recommend SAS to have built?
"SPDS was the brainchild of an extremely difficult individual who few could tolerate working with for any length of time"
Someone put initials to the above "brainchild".
The initials are AG (Amitava, or the more familiar Ami, Ghosh). This is an Indian (Asian) name so it does not mean “friend”.
For those external parties doing research, who have questions like this, you can find information on Google and you don’t need OpenAI’s Deep Research to do it. In this case, just search for SPDS.
… LASR = multi-user data AND analytics server, necessary for interactive clients like VA to achieve real-time performance at scale.
the general lethargy and lack of scaled execution on sound ideas at SAS (especially R&D)
SAS culture was actually working against its own success and executed on a strategy with the brightest and hardest working people he could find to do something about it. OS has my respect to this day for this while manifesting consistent intelligence, tenacity, and work ethic in the process. He fought an uphill battle against mediocrity for well nigh 15 years
Rabidly criticizing and alienating a significant portion of the company was a most excellent way to get everyone onboard for the Viya effort. Fostering an us and them attitude was just a plus. OS was a genius among geniuses. He was a man’s man. He was what was needed at the time, we must all admit.
IIRC — LASR was initially built to provided a multi-user data server in support of V9-era VA.
OS cannot be solely blamed for everything in Viya built atop CAS. Other divisions and their leaders had substantial impact on the client side products.
As repeated ad nauseam, CAS was overall a sound strategy in the time it was conceived, especially given that in 2012 open source infrastructure like Apache Spark, Kafka, etc. were nascent and it’s all but certain JG would NOT have green-lighted building out the next generation of SAS atop open source software that we could not control.
The cloud hyper-scalers were quietly and consistently growing analytics prowess on the massive data platforms they already owned. In 2012, SAS was “riding high” although R and Python analytics were growing, yet still relatively niche, especially Python which soon exploded and ultimately took over.
It was the perfect storm that SAS has not weathered well. Very sad indeed.
“Viya is just the latest in a long line of classic fu-k ups at SAS.”
Yup.
“OS… executed on a strategy with the brightest and hardest working people he could find… Perhaps if many others had done the same… SAS would be thriving today.”
Others were not so incentivized. They were paid to work a 35-hour week, so that’s all they did. I never respected their work ethic, but can’t blame them for taking the deal they were offered.
OS did work hard, but did he work in a sound direction? And if Viya started in a sound direction, was there a failure of execution?
I don’t fully understand, for example, why LASR was built. Also, the applications on top of Viya have not compelled many customers to buy it.
I look forward to more of the OP’s explanation.
To accurately summarize the naysayers: Viya is just the latest in a long line of classic fu-k ups at SAS. We enjoyed briefly light in the mid 90s once version 6 was stable. Anyone who was there remembers what a compatibility cluster fu-k that was. It took 5-10 years to get most customers off Version 5, which also had compatibility issues with its predecessor, SAS 82.4. So it seems grasshopper, that doing enterprise data management and analytics software right is the quagmire that Brooks described in his classic TMMM.
What does NOT need to be rehashed is the general lethargy and lack of scaled execution on sound ideas at SAS (especially R&D) from mid-late 1990s until now. After delivering MVA and milking that technology, for all it was worth, SAS, had relatively small pockets of success, yet was unable to scale revenue at a pace that even began to keep up with the overall market we were in.
OS recognized that and busted his a-s to do something about it. He accurately assessed that SAS culture was actually working against its own success and executed on a strategy with the brightest and hardest working people he could find to do something about it. OS has my respect to this day for this while manifesting consistent intelligence, tenacity, and work ethic in the process. He fought an uphill battle against mediocrity for well nigh 15 years in the later prime of his life. Perhaps if many others had done the same, instead of decamping to the RFC for two hours a day or rarely doing anything beyond the standard 35-hour week, SAS would be thriving today.
“…a lot of fundamental design/development and innovation went into version 7 and carried forward. ODS being the crown jewel.“
ODS tables shipped over several years. ODS graphics took eight years to ship production. Our competitors did not need eight years to print graphics.
Shipping a product does not make it a success. It it can’t be shipped in a competitive timeframe, it’s a failure.
ODS, like Version 7 generally, was a failure — because while we built it, we fell behind our market. It is an early example of SAS moving too slowly to compete.
What went wrong with Viya?
Wrong product at the wrong time. Too late. It was a horse drawn carriage in a world that had traded carriages for automobiles. Revolution left evolution behind.
The Viya creator was the wrong guy to take it on a dog and pony show with customers. Yeah, a road show is needed, but the person doing the road show needs to care about listening to customers. OS could not have cared less about listening to customers
All this is old news and rehashing it one more time won't make it taste better, nor will that magically make Viya relevant. The industry has spoken and has moved on. Time for us to do the same!
"Those who cannot remember the past are condemned to repeat it.” — Santayana
Those of you with no interest in history are not required to read this thread. Some of us want to learn what happened, for our own interest and education.
I always had the impression that Viya was an abject failure. But the history thus far suggests that, at least in its early stages, there were logical reasons for the decisions.
@OP, as Holmes said to Watson, “Pray continue your most interesting narrative.”
Many were at SAS long before OS was hired. And many of those same folk were there long enough to see the rotten fruit called Viya that has lead to where SAS is now. A company adrift.
Rehashing history is good if one admits mistakes were made. Unless that admission happens, then rehashing is mostly a waste of time.
Honestly, it is too late to rehash history. Why? Because the founders have run out of steam. None of their kids seem interested in running SAS. Whoever acquires SAS likely has no interest in the history which you speak of ad naseum. Their going forward plan will be very different than any of the history.
What was/is your role at SAS? How far do you go? Your comments sound like you’re a sales guy who got hired 3-4 years ago.
Anyone can throw stones at where Viya is today, but to ignore history is to fail to understand how things got where they are and how deeply this trajectory is ingrained into SAS culture and how the founders have operated the company, especially over the last 25 years.
No history lesson needed on Viya. Especially if written by biased OS himself.
Cliff notes version: flat revenue since Viya released. The Viya brain child returned to academia due to his baby flopping. If that is success, I'd hate to see failure.
Definitely agree that a lot of fundamental design/development and innovation went into version 7 and carried forward. ODS being the crown jewel. Some solid low-level services that have survived in SAS since the late 90s were developed as well. FWIW, prototypical early work that eventually became the metadata server started back in the version 7 timeframe
As I recollect, the biggest problem with version 7 was our attempt to mimic design patterns in the MS-Windows object model and make those ubiquitous throughout SAS internals and interfaces. This created a lot of frustration amongst developers across host, core and analytics, especially those deep in the lessons learned from MVA.
I keep reading in the layoff page that Version 7 was a failure along with Viya. No one can deny that Version 7 was not a success in the market, but the similarity stops there. A lot of work went into converting from V6 to V7. Notably, that is where the conversion from 8 character names and 40 character labels were converted to 32 character names and 256 character labels. Much more involved was the conversion to ODS. Both of these conversions were a lot of work. This work went straight into V8--it was not ripped out and redone. V7 was not a success in the market, but it was a critical steppingstone to V8, a successful release. Viya is a steppingstone to the demise of a once great company.
“
TKTS was poorly scoped and lacked technical leadership with true, world class expertise in database technology.
I think that BH and JM would disagree. EVERYONE was told to read up on ODBC..
“
Please articulate how TKTS fit within the overall vision for the SAS architecture starting back around year 2000. Who was it original “lead architect” and what were its primary objectives? Did it have enthusiastic executive sponsorship? What does everyone being told to read up on ODBC have to do with ultimately building a highly complex transactional database management system for SAS atop an existing technology? One can be highly competent in ODBC without knowing much about database internals and what it takes to integrate them into the kind of analytical data access patterns that SAS requires.
The original TKTS architect along with the two people you mentioned are certainly solid citizens in the history of R&D, and proficient database professionals at some level, yet it’s uncertain they possessed the depth of expertise necessary to build an industrial strength DBMS into the core of SAS and scale it to perform on the largest customer workloads. Sadly, such efforts were typical of the time, coming off the heels of the version 7 disaster.
“
Why are you doing this? What’s the point? This has nothing to do with layoffs. I suppose that with the world going to he11 in a basket, it doesn’t matter
“
Because a considerable amount of shade continues to be thrown on Viya by some here who may not understand its origins and why/how it organically grew out of the SAS culture and technical evolution. Interest was expressed in continuing these historical installments by some on the other thread. FWIW, The success or failure of Viya has and will likely continue to have something to do with the layoffs as time goes on