Creator of ZFS, Jeff Bonwick, leaves Oracle | Hacker News Reader

Creator of ZFS, Jeff Bonwick, leaves Oracle

blogs.sun.com

96 pointschibea67 comments

Did Oracle buy Sun to strip it and liquidate the intellectual assets? What's their plan here? It seems like they have zero interest in building the products.

Oracle is a consistently profitable company. They know how to run a business. They know what they're doing. If the old Sun and their culture wanted to survive, they needed to run a better business.

They certainly know how to sell their database. Whether this translates into an equal success with their Sun IP has to be shown. The good thing for them is that they acquired pretty mature products, so they can probably "coast" along and still make a reasonable profit. It's still open whether they can foster enough innovation to keep things going with Java, Solaris… Their track record isn't very good in that respect.

Of course, it all boils down to whether they need to innovate in the first place. By now, Java is well established enough in the enterprise sector, just as BEA and Oracle DB are.

ZFS, however, is not in that category, and everything I've heard says that it needs more work and a lot more polish.

You should try putting it through its paces before coming to that conclusion; I've heard different things from other people, and I currently use it for my media storage.

The problem is that for this type of thing credible reports of failure carry more weight than reports of success. If you're making a big bet on the reliability of file system you want to mitigate risk, and for all I know either you haven't tickled a bug I would or you've just been lucky so far.

When I looked hard at this 2-3 years ago, it was clear to me that without a tape backup system I couldn't justify purchasing, ZFS just wasn't there yet. I've heard nothing since them to convince me it has been sufficiently polished to get over my threshold or required reliability.

ADDED: I don't think it can be said that ZFS is established to the same degree that the parent's examples are (Java, BEA and Oracle's DB).

On the other hand, I think that file systems are closer to the heart of a database company than quite a lot of other Sun technologies (e.g. JavaFX). Oracle does sponsor btrfs – which is even more in its infancy.

Also, Jeff leaving doesn't necessarily mean that ZFS is dead to Oracle. Maybe he left because they decided to focus more on btrfs, maybe it's just a personal issue.

Indeed, 20 years at one company is a long time, plus his personal financial upside became strictly capped to salary once Sun was assimilated by Oracle. There's all sorts of reasons he might jump at a new opportunity, many of which don't mean the end of ZFS.

Of that I have no doubt, but what I originally thought they would do; use their business resources mixed with Sun's engineering resources, does not appear to be their plan at all. They seem to be losing all of the best Sun engineering talent and do not seem at all concerned about about that fact.

That leads me not to conclude they are failing at what I originally thought they were doing, but succeeding at something else and I am curious as to what that is.

They bought suns business. They will probably replace engineers and continue projects they like, they will lose engineers focused in products that are cut.

I think part of suns problem was engineering that was going in for engineering sake, not business sake, and oracle is going to lose those engineers.

I don't know what businesses other people have collaborated with ... but to me the people behind the products I'm using matter a lot, I'm not dealing with faceless entities.

Personally that's precisely why I'm using Postgresql ... I can talk to the core team of developers and get their input whenever I feel like there's a problem with Postgresql. And for a web service we worked on, we even hired a Mysql developer to come and help us (was also on Sun's payroll at the time).

How could I trust a product based on ZFS or how could I trust them to build my products on top of Java or any of their Sun-derived solutions ... when the top talent that worked on those products left?

Not to mention I don't like hypocrites ... before the Sun acquisition they called on Sun to make the JCP more open. Is this an US thing? Can companies fuck with you because it's just business?

Oracle seems to me like a sweat shop. Nothing wrong with that and I'm glad they are making lots of money ... but in my shop you can get fired for suggesting Oracle products, and it's not just me, there's a whole generation of software developers that won't touch their cash cows.

There's a whole generation of developer that have chosen Mysql or Postgres or implemented their own shit, and did not go for Oracle's DB ... that's revenue and opportunities lost for them, and I'm only seeing this trend getting stronger.

Sun had more copyrights than anyone except IBM. I suspect that's the reason. And it could have even been a defensive thing. Sun was obviously failing badly. If they didn't buy Sun someone else eventually would and then that company could almost certainly start suing oracle over something.

Simply profiting does not mean that you're playing the game optimally.

While small organizations can't afford a research division, Oracle certainly can. You'd think a company in the business of storing data might want to keep the guy who created ZFS, especially given how lively the database world is right now with NoSQL implementations, etc.

Based on what James Gosling has said, perhaps Oracle's general HR policies just aren't compatible with retaining people at Jeff Bonwick's level.

There could also be bureaucratic turf wars going on, i.e. the total loss of autonomy/decision making power that Gosling reported suggests that those already in Oracle are happy with how they run things and have no intention of sharing those responsibilities. One can also imagine Oracle having a culture where those acquired from a money losing company don't get much respect for that alone.

This happens everytime a company is taken over by a larger one. Except for a few VPs who are part of the M&A team - everyone at the smaller company always ends up beneath the same person at the larger one.

Everyone in tech knows this - which is why everyone who can leaves before a 'merger'. And why you never hear of the innovative smaller company's work 6months after the takeover.

I'm not sure that's entirely true (well, except for the reporting thing). Some companies have a reputation for acquiring smaller companies with the express purpose of keeping their staff, e.g. Cisco (at least in times past) and Google.

But, yes, most of the time this is an important part of what makes the vast majority of high tech acquisitions fail miserably.

Even G often seems to perform quite poorly. Viz Dodgeball / 4square. Thoroughly alienating Dennis Crowley seems like an enormous mistake. Or Paul Buchheit. Or the Delicious guy who worked at G but didn't find it a welcoming environment. If they were serious about social, they'd do whatever it takes to keep people like Joshua who've built successful social products.

I'm coming around to Zuckerberg's opinion that social is something a company is built around, not something you staple on or buy after the fact.

> Or the Delicious guy who worked at G

This comes across very weird...

I know... Wrong website for jokes like these. Couldn't help.

There's a Joel Spolsky podcast about Qt (pronounced cute) where he talks about the cute guy giving the talk - to the confused embarrassment of the other speakers.

They know how to run one kind of business, that much is clear.

The question was "what is their plan" not "is Oracle a competent company".

I suspect somebody said the same thing about GM or Chrysler

Oracle bought Sun to make more money.

That's clearly the "why". He (and I, for that matter) want to know "how".

Because the alternative was that IBM or HP would buy Sun - or they would simply disappear.

The result (in either case) would be a lot of IBM hardware salesmen crawling over Sun's customers saying things like: If you want to switch to DB2 to go with your new IBM servers we can do you a deal.

Oracle is technically irrelevant because they aren't technically focused. They aren't even close. Their core product hasn't been gaining any steam for years and they're riding out that gravy train. They have to keep purchasing other companies to remain relevant.

So from the thread below it sounds like Oracle actually acquired Sun primarily for their customer base alone, a large pool of cashed up blue chips for Oracle's sales force to feast on. The engineering culture and aptitude at Sun would from this strategy just be considered overhead, thus the mass exodus of talent. So perishes a once mighty titan.

Still, will be interesting to see what happens with the seeds flung off from this tempest.

That's too bad. ZFS (along with crossbow) is one of the best things in Solaris/OpenSolaris.

I was about to say Oracle has a serious human resources problem, but I noticed that, maybe, the problem is in calling humans "resources".

Agreed. To Oracle, these are just pawns in a game. The winner of the game (which is always Oracle) gets to fund Larry's next yacht.

The alternative option was that Sun just went bankrupt and everyone lost their job, so choose between some leaving and all losing.

Aut Larry aut nihil? As James Gosling pointed out just recently, IBM was the second option, not bankruptcy. IBM was thought as worse when it came to laying off people and outsourcing their jobs, but they tend to respect their IP and engineers more -- which is a pretty sad revelation, considering that we're talking about big ol' Inferior But Marketable[1] here.

[1]: http://www.etypewriters.com/hailto.html

Actually, I think that properly managing the company and preventing the bankruptcy would be a better third option.

I know, I know... It's easy to say it now, that hindsight is always 20-20, but come on, who outside Sun thought their market strategy was sane? Who really believed they were on the right track? And, as much as I like McNeally (he was really fun), it's as much his fault as Schwartz's that Sun had a problem with "financial reality" (as Gosling so mildly said).

It was painful to watch because we knew how cool technologies they had were, what they could accomplish while, at the same time, knowing it wouldn't last.

I don't think that's it. Every corporation thinks of its employees in a dollars and cents kind of way.

The operative problem here seems to be that Oracle doesn't value senior technical resources as much as some other companies might, or as much as those resources are used to being valued (and really, few people react well to a cut in autonomy, pay or respect.) They aren't willing to invest what is required, probably in the non-monetary sense as much as the monetary sense, to keep the top talent that came with the Sun acquisition. Personally, I agree that this sounds like a mistake, but this sort of attitude is not new for Oracle, and eh, they've done okay so far, so what do I know?

I agree re: your second point, but when I worked at IBM it was explicitly discouraged to use the term 'resources' to refer to people.

Personally I find the people that use this outdated term stop doing so when continually referred to as 'Outlook resources'.

Really? IBM refers to layoffs/redundancies as "Resource Actions", which is surely one of the most mealy-mouthed corporate expressions

Yes, really. PMs would publicly mock new-to-IBM PMs who used mid-90s business speak. When I was there they also referred to layoffs as 'getting rid of people'. It was a pretty good environment - honest and respectful. Perhaps your office was different.

That's extraordinary and not at all what I would have expected from IBM. I wish there were more places where people ditched corporate doublespeak. Especially that vile phrase "human resources". Soylent Green is human resources.

Any place that has any kind of "employee valuation policy" doesn't value their employees.

> Personally I find the people that use this outdated term stop doing so when continually referred to as 'Outlook resources'.

I use to call them "Powerpoint resources", but I think I'll adopt your designation.

I think you're being unfair - some of them are "Visio resources". Tho' they call themselves "architects".

Every corporation thinks of its employees in a dollars and cents kind of way.

Reminds me of my brother’s old company ( http://www.freud.com/ ). Their website says:

We employ humans, not resources. Job opportunities, contact the HR manager.

Was the fun part that they suggested contacting the Human Resources manager?

Well yes, that was the point I was making :)

Based on what James Gosling has said, as a Sun Fellow Jeff Bonwick would have experienced a significant compensation cut and loss of autonomy (near total of the latter for Gosling). Which most people will interpret as a lack of respect.

As many others have wondered in this set of comments and others, just what is Oracle's cunning plan to make their purchase of Sun worth the investment???

Charging more for Solaris support and providing less of it, for starters.

And thus making it compete with a product that's freely available.

I don't think that will work.

It would work if Solaris was something people really desired. Right now, it's not. People are very happy with Linux and need some incentive to move to Solaris. Paying more is not a good one.

And don't forget that no-longer-OpenSolaris is still freely available ... and might even become something now that it's free of the dead hand of Sun (at least for it being a truly "open" project).

I for one am not happy with Linux and would love to run Open/whateverSolaris ... if it has serious driver support. That's been the Achilles heal of x86 Solaris for more than a decade (e.g. something I directly experienced in 1999) and fixing that will likely be a determining factor in its future success.

Unfortunately that's very painful hard work and lots of it, I wouldn't predict success on general principles. But one can hope.

There are existing businesses who bet on Solaris and have it widely deployed. Moving to Linux may not be trivial. That is, it may not be beyond the FYO point. [1]

The IBM AS400 isn't something that people really desire, but there are a lot of businesses that rely on it and IBM still gets paid to support mainframes.

[1] http://dtrace.org/blogs/bmc/2004/08/28/the-economics-of-soft...

> The IBM AS400 isn't something that people really desire

I liked OS/400, you insensitive clod. If I come across a 5250, I'll buy it immediately.

Not sure what I'll connect it to.

But, going back to Solaris and OpenSolaris, it's easy to switch a deployment to "legacy mode". Unix machines are Unix machines and it's not nearly as difficult to move from AIX/HP-UX/Solaris/IRIX to Linux as it would be to move from MVS, unless your binaries insist on running only under Solaris. Legacy mode means no new deployments and no expansion unless justified. It's slow death.

True, in the case of very proprietary platforms (like z and iSeries boxes) it takes very long, but it's death nevertheless.

In order to be kept alive, the Solaris boxes have to be able to perform tricks Linux boxes can't and, to a large extent, this is not the case.

I'm sure that moving off iSeries, etc. is a much more significant proposition than moving off of Solaris. But the barrier isn't zero. For an installation of any significant size it's almost certainly not going to be as simple as switching a deployment to legacy mode. You've got shell scripts to port, different configurations for core features like kickstart, firewalls, different package management. If you made an "environment standardization" bet on Solaris 10, you've got barriers to overcome.

But I'm not saying that Solaris is going to live. Just that, as someone who uses Solaris 10, Oracle seems to be tightening up. I think that they don't really care that much about Solaris and are just going to squeeze it dry. I'm far more worried what they're going to do with Java.

> In order to be kept alive, the Solaris boxes have to be able to perform tricks Linux boxes can't and, to a large extent, this is not the case.

True, but keep in mind that their competitors are often limited to "enterprise" distributions like RHEL and Ubuntu Server.

You are overgeneralizing. Sun's (and Oracle's) main customer base are "people" who use Solaris/Oracle to manage their millions of customers and billions of dollars. They don't want to mess with it, they want support, and will pay for it. The question is, "how much?" -- and Oracle is very good at getting a profitable answer to that question.

"Your ideas will go further if you don't insist on going with them."

Good luck Jeff Bonwick.

It's good to see something positive coming out of the Sun acquisition: a whole crop of very interesting startups are probably being founded right now.

I'm really uncertain now of the future of Illumos and ZFS. I hope there's more push behind these projects as well as BTRFS. Will be interesting to see where this goes.

Since btrfs is an Oracle funded project, I'm real interested to see what happens to both of these filesystems going forward.

As someone who has actually used OCFS2 on a real production system, I will never voluntarily touch another Oracle-written filesystem.

I think the future of Illumos is safe. They've already got a lot of heavy-weighters behind it, ex-Sun/Oracle and others.

Someone still has to fund them...

His Nov.2009 blog entry on ZFS data deduplication is a pretty good primer:

http://blogs.sun.com/bonwick/en_US/entry/zfs_dedup

P.S. also, the 2008 pics with Linus. I suppose there is no further story to go with those ...

IMHO, with my little work in compression, all that dedup stuff was a bit hyped. And BTW it is older than ZFS, at least Plan 9's network storage had the same feature. But alignment kills looking for matching chunks of data. And "generally best left to the application" sounds like a cop-out.

It's the ease of usage, speed, and very large compression window that makes dedup better than traditional compression.

Wow, major talent leakage. Whoever running a startup right now, if there's a need for low-level programming talents, better snatch these guys.

Hmm I was hoping he would go work for Apple on a new filesystem to replace HFS+... oh well.

Will be interesting to see what his startup is working on when they come out of stealth mode.

Oracle's strategy seems to be:

(1) Don't bother investing too much in R&D or figuring out the next big technology, just buy companies that are alread profitable (to name a few: Sun, BEA, Hyperion, Sieble, Innobase, PeopleSoft...).

(2) Immediately execute massive layoffs to increase profits even more. And suck every dollar possible out of the business for a couple years.

(3) After a while, the profits will start decreasing, but slowly enough (due to market inertia) that they have time to rinse and repeat. Go back to step 1.

Sometimes, Oracle will get lucky and be able to keep running a business good enough that the profits will somewhat be stable enough for years.

"just buy companies that are alread profitable (to name a few: Sun, "

don't know about the rest in your list, can you specify what do you mean saying "profitable" in application to Sun?

during recent years Sun was like an aircraft carrier steered by (actually with all the systems controlled by) a bunch of monkeys - a lot of noise and jerking right an left. When finally it was grounded and started to sink, its crew was saved by another ship.

Why it so surprising that ones who grounded the carrier didn't get to steer the rescuing ship?

I don't know how much of the oracle's academic initiative was a success. From the likes of what is being happening to Sun's Academic Initiative to promote university grad students to take the certification exams such as SCJP now called Oracle certified java programmer. Sun has offered its student members the certification exam for 40 $ which has been recently dropped by oracle making it 40% of the voucher fee which implies 40% of 150 around 90$. The 40% discount voucher is also for a limited period which will be done for good this December. I seriously doubt oracle's contribution towards opensource too.

Reply on news.ycombinator.com