Slashdot Mirror
PostgreSQL - Oracle/DB2 Killer?
dagnabit writes "At Yahoo News, there's a story about a company which is investing in/supporting the PostgreSQL crew. Ultimate goals include "a planned expansion to 120 employees and the ultimate possibility of going public." So that enterprise-class open-source RDBMS may not be too far off after all... "
The comment about gaining "mastery of the code" must be taken in context. The current team of developers are not the academic folks who originally developed it at Berkeley, and came together relatively recently (over the past three or so years).
/. article reference you mention was NOT entitled "Why MySQL sucks". It was simply a statement as to why the OpenACS team choose Postgres instead. We understand that MySQL fits a narrow niche quite well. We understand what that niche is, and more importantly isn't, much more so than do the developers of MySQL, judging from their documentation (foreign key constraints are almost always only used for documentation, so enforcement really isn't needed? Feh!)
Obviously, being stone-cold to the code at the beginning it took awhile for them to learn and, as they put it, master it.
So the statement made when 6.5 was released was meant as a milestone accomplishment, if you will.
"We decided to take up and fix Postgres. In order to meet this goal we needed to collectively (not individually) master all the code. With 6.5, this milestone has been accomplished."
The original authors and others from UC Berkeley who worked on it in earlier years have moved on to other things. The current team took the code, learned it on their own, and as of 6.5 had made vast improvements and fixed many bugs. 7.0 - released today - fixes even more bugs and makes referential integrity and various other features available.
Regarding memory leaks that cause periodic crashing of the RDBMS, those were almost entirely fixed by 6.5, which was released a year ago.
The
You might as well make an effort to quote the title correctly, even if you didn't understand what the piece was saying.
That site, BTW, is running on Postgres V7.0 Beta (virtually the same version as today's PG release). It was slashdotted. My, oh my, enough to bring strong servers to their knees 'til they beg for mercy. Not this time - the dual P400 running AOLserver, Postgres and OpenACS never had to take off the sweats, much less work hard. It was running several other sites at the time, as well as being used for development that day by its owner, who wasn't even aware of the slashdot traffic until someone pointed out that his piece had (unexpectedly) been mentioned here. He then poked around and realized that the machine was serving lots 'o pages.
Each page served out of that server involves several hits on a nearly 10,000 line datamodel stuffed into Postgres. There is ZERO caching of queries done on that server at the moment. Nada. Each page built dynamically out of the database.
Those who don't think modern Postgres is fast enough for this kind of task ought to step back and think again. Those who think MySQL provides anything like the data safety of a true RDBMS ought to read a basic book like Gray's "Transaction Programming".
if you fill out the form for a free t-shirt on their web site, it asks you what features you want & they hit pretty well what PostgreSQL is missing - good blob support the biggest (and easiest to fix from the developer's standpoint) it'd be nice to see them dig in with some of the other features like replication & clustering.
the big question i have (paranoia alert!) is that PostgreSQL is under a BSD license, so they don't need to release any changes they make if they don't want to. it didn't sound like that would be the case, but is this the first case of an OSS startup working with BSD-licensed code? it's a whole new ballgame compared to GPL.
The PostgreSQL team has come a long way. They inherited a codebase with known bugs and some fairly serious failures. They have worked hard and have turned the quirky Postgres of yesterday into a powerful and dependable tool. Anyone who has used past versions of Postgres will tell you that the latest versions are a completely different beast.
PostgreSQL 6.5 has been out for nearly a year, and the 7.0 release will be out any day. In fact, I am currently running one of the release candidates in production because of the added performance. PostgreSQL has a full set of reqression tests, a rich set of features, a dedicated development team, and an extremely helpful mailing list.
What PostgreSQL doesn't have, however, is a marketing department that is going to feed you pap. The developers are perfectly willing to tell you which parts of the database are in a state of flux. Historically the list of features that were less than stellar was fairly long, but that is no longer the case. The flip side of the PostgreSQL team's honesty is that, when they tell you that a feature is ready for prime time, you can bet that it is indeed ready to be put into production. How often have you put commercial software to the test only to find that some of the new features are lacking? With PostgreSQL there is no guessing which parts are ready to go.
Clearly the PostgreSQL database isn't right for everyone. But it's no slouch either. And you only As I mentioned before I am currently using it in production and am quite happy with it. I have several tables with millions of rows in them and performance has been more than acceptable. I have been very impressed.
The PostgreSQL mailing lists are full of developers that have migrated from MySQL. MySQL is great for 90% of database tasks, and it has the advantage of being very fast at straight selects, but if you need ACID features then MySQL isn't going to get you there. PostgreSQL will, however, and it also throws in a raft of useful features like triggers stored procedures (in 4 different languages including Perl and Tcl), views, etc. PostgreSQL's support of subselects could very well improve your application's performance.
Not too mention the fact that PostgreSQL is truly Open Source (BSD style license) while the current version of MySQL is clearly not. That is why you get announcements like this for PostgreSQL and not for MySQL.
I am a bit skeptical about this particular announcement myself (since I hadn't heard anything to this effect on the PostgreSQL mailing lists), but whether this particular venture pans out or not it will almost certainly mean that some more PostgreSQL code will be generated.
Heck, it might even make it into 7.1.
Its possible that the even the mighty Oracle's developer team doesn't know what everything does.
I've seen Oracle copyright notices going back as far as 1975 or so, which implies that their codebase is 25 years old. Are any of the original Oracle developers still on the team? How many of their current developers have even been there more than 5-10 years?
Couple a large project with even modest turnover, and you're bound to have sections of code that few (if any) current developers are familiar with. It's probably almost unavoidable. Why do you assume that commercial databases don't have the same problem?
At least the PostgreSQL developers admit they didn't fully understand the code they were maintaining until more recently. They say they are familiar with all of the code now. Chances are that puts them ahead of Oracle in that regard, not behind. Keep in mind that PostgreSQL derives from Postgres95, which derives from Postgres, which was an academic research database. That's a long history to catch up on. Give them some credit for managing it at all, and for being honest about their progress...
Deven
"Simple things should be simple, and complex things should be possible." - Alan Kay
Wow, we're happy to see the response to the initial announcements. I'm Ned Lilly, one of the four employees mentioned in the CNET/Yahoo article. Wanted to say hey, and try and answer some of the questions I've seen so far. Pls feel free to email me directly too- ned@greatbridge.com.
1) PostgreSQL Features. We took about six months to study the open-source world, and have spent a lot of time testing, benchmarking, reviewing code, etc., and in our opinion, PostgreSQL 7.0 is by far the best choice among open-source databases. My personal impression is that some people might have bad memories of earlier versions (I myself actually used it in an earlier business but had to throw it away), but the improvements leading up to 6.5.x and now 7.0 are incredible. And, of course, 7.1 is going be even better :) I'd defer any specific questions about features to the Postgres developers themselves (http://www.postgresql.org/devel-contrib.html) - they're still running the development effort, and will continue to do so.
2) MySQL & Interbase. So that begs the question, what about other open-source databases. Don't want to start a flame-a-thon here, I'll just say that we think MySQL lacks a lot of desirable features (this topic's been exhaustively debated elsewhere), and Interbase, while it might be a good commercial product, isn't yet open-source. To my knowledge, they haven't released the source for 6.0 yet (only binaries), and it's a fair question whether you can create a robust open-source community instantaneously. PostgreSQL has been open-source for many years, and the current core steering group has done a great job at making the code base clean, uniform, and accessible to other hackers.
3) Our Customers. So who are our customers? There's been some good discussion here on that topic already. The short answer is, we've got some ideas, but we're being fairly deliberate in how we verify those ideas. We've done some pretty comprehensive market research on different sectors of the IT world, including 32 focus groups in four U.S. cities, and we think we'll be able to put PostgreSQL (and other open-source tools) in the hands of the people who will best be able to use them. We don't really want to talk in too much detail about our business model at this point - frankly, we're still refining it based on the potential customer feedback we're getting from this research. That process is how Landmark Communications (the parent company) went about building The Weather Channel cable network from scratch, and we think it can set us apart from some of the other open source companies in the market today.
4) BSD vs GPL. More potential flame-bait here, so read carefully... :) Another area we spent a lot of time researching was the area of open source licensing. We like the Berkeley/BSD-style license under which PostgreSQL is now offered; we think it's the simplest to understand (it's free for any use, with no restrictions on what you can do with the code), and offers the fewest barriers to widespread commercial adoption. We've heard from a lot of business types that a GPL-style license scares them when it's touching their proprietary business applications. You might wonder if we're just going to freeload on all the community development work, then take the code, call it GreatBridgeSQL, and fork it off into some proprietary product. That would be the stupidest thing we could possibly do - we are 100% bought into the open source development model, and have no intention of ruining what is working so well. Indeed, we're going to do quite a bit of internal development ourselves - on interfaces, APIs, monitoring tools, and business applications - and everything we write, we're going to throw back over the fence into open source for the community to bang on and improve. We'll do the same to other peoples' code - just like any other member of the community.
5) Business Plans. Some of the press reports have talked about a potential IPO, and people have rightly sniffed and said, yeah right. Let me be perfectly clear on this one: Landmark Communications, which is a 100-year old privately held company, doesn't build companies to do a quick public offering spin-out. The Weather Channel, and weather.com (which has 14 million unique users a month), are both privately held. Great Bridge's CEO, Al Ritter, is the former CFO of Landmark corporate, and is deeply steeped in the old-fashioned notion that companies should probably try and be profitable rather than just ask for more and more public money. We have no intention of going down the IPO road until we're satisfied that we've built a business that can stand on its own two feet, generating real live profits by delivering professional support for open source software solutions. We think, frankly, that a lot of tech support (particularly in the database arena) stinks - and we can do better. We think we've got a good product to get behind in Postgres. And if at some point down the road we think there's an opportunity to super-charge the business through an IPO, we might well do it. But unlike a lot of startups, we don't need the money. Landmark's committed $25 million to this phase of the rollout, and if we can make the case to our corporate bosses for another round of investment, there's more where that came from.
6) Hiring. Yup, we're hiring. We want developers, support engineers, even, er... marketing people. We're talking with a lot of people in the PostgreSQL world already, and will be making some more announcements on that front in the weeks to come. If you're interested, please email me at ned@greatbridge.com.
Thanks...
Ned
From the article:
As chief financial officer of media company Landmark Communications, Ritter watched his company miss out on a golden opportunity to invest in Linux software seller Red Hat way before its successful initial public offering. Now he hopes to catch the second wave of the open-source software trend.
...
Great Bridge, though consisting of only four employees today, has grand ambitions, including a planned expansion to 120 employees and the ultimate possibility of going public, Ritter said. But the biggest challenge will be taking on giants such as Oracle, IBM, Microsoft, Informix and Sybase, each of which have their own proprietary database programs.
First of all, a proclamation like this seems to me more like jumping on the open source bandwagon and hoping for a successful IPO (with no long term prospects of giving investors return on their value) than as a project that is destined to make Oracle and Sybase quake in their boots. It is particularly interesting that the a company that is supposedly Open Source oriented is stating all sorts of plans about IPOs and such and little or no talk about technology.
Hopefully I am wrong and this will be a company that will give back to the community in spades as opposed to a bunch of opportunists. Oracle databases currently cost several thousand dollars ($20,000 minimum price) while MSFT's cheap SQL server is about five thousand dollars. If a company can be created that will produce software as robust and functional as Oracle software (believe me that is a daunting task) and yet charge only support costs they may well sweep the DB market especially for small businesses. Of course, it will take a phenomenal amount of work (mayhap an impossible amount) for a four man company to compete with a company that has $6 billion in revenues and one of the most robust products in it's area. I wish them luck.
I always thought databases would be a good candidate for Open Source development. Since many (most?) of the users are developer-minded already, getting a community should be easier than getting people to rally around it.
:)
What I *hope* this new company will do, if it truly emmulates RHAT, is hire full time developers for the project. Hobbiest coders have contributed tons to linux (heck, they invented it), but I'm sure the OS has benefited immensely from having full-time people working on it as a day job. I'd like to see the same for postgress.
I also agree with the other commenters who point out that postgress doesn't need to be an Oracle killer. There are plenty of medium to large projects that don't require a full scale Oracle system. Postgress can become an important alternative for people who need more than Access (or, arguably MySQL), but don't need Oracle and don't want to use SQL Server.
My $0.02
Dana