Slashdot Mirror
Sun Announces Support for PostgreSQL
jadavis writes "Sun announces 24x7 support for PostgreSQL on Solaris 10. From the article: 'Today Sun announced that it will be integrating the Postgres open source data base into the Solaris 10 OS and providing world-wide 24x7 support for customers who wish to develop and deploy open source database solutions into their enterprise environments. Sun is working with the PostgresSQL community to take advantage of the advanced technologies in the Solaris 10 OS, such as Predictive Self-Healing, Solaris Containers and Solaris Dynamic Tracing (DTrace).'"
http://www.sun.com/software/solaris/news/111705.j
More about Postgres specifically:
http://www.sun.com/software/solaris/postgres.jsp
Damn, blew the link - should've (should have, not should of) previewed.
No folly is more costly than the folly of intolerant idealism. - Winston Churchill
There is a blog from a Sun Engineer about databases, etc.. He talks about PostgreSQL, how to improve its performance, etc... You can find it here
Both of my friends, who run CPanel cheap webhosting companies, offer PostgreSQL. For free. It's a little elephant icon right next to the MySQL icon in CPanel. I can assure you, they do exist. A quick check in the CPanel user manual shows a whole section devoted to PostgreSQL. As to how hard it is to set up from scratch - I couldn't say. But here's one way multiple users can use it extremely easily and quickly.
From a commercial point of view Solaris 10 isn't in the roadmap yet - looking at our customers it's evenly divided between solaris 8 and 9 (couple of solaris 6) and zero solaris 10.
The shift will probably start happing in the next year or so... then we'll have to buy another sparc box to support it (any excuse...)
Vs pgAdminIII ?
Nice GUI admin tool. I like that much better than silly web applications.
beeing multiuser is just as easy in postgresql as mysql..
But if you for some unknown reason must have a web tool, there is phppgadmin
I don't think opensource databases are becoming any more of a threat than they were in the past. They really do cater to a different market. This is WHY you see SQL Express and the new Oracle license.
Here's the deal. The company where I'm the SysAdmin has 3 databases we support - DB2 (Linux and AIX), SQL Server (financial product decision made outside of our department without our consultation) and PostgreSQL.
DB2 runs our core database for our enterprise application. All databases were investigated at the onset of this project and DB2 came out on top. SQL Server is in house for a shitty financial package (Navision) and another legacy system. PostgreSQL is our data warehouse.
Because of some issues surrounding our DBA team and the fact that SysAdmins often have to cameo as DBAs in a quick pinch, I've come to learn quite a bit about DB2. It has its warts and bugs but it's 100 times more robust than PostgreSQL and 1000 times more robust than MySQL (which we use for a few self-managed databases here and there - intranet stuff/nagios).
We're currently migrating our data warehouse to a new hardware set and at the same time upgrading from 8.0.3 to 8.1 of PostgreSQL. This requires a restore of the database to migrate. This 80GB datawarehouse took the better part of a day to restore on a box that was 10 times faster than the original. Reading from different volumes on different controllers on our SAN on an x445 with 8 CPUs and 16GB of memory took 8 hours to restore!
This box used to run DB2 on Linux (we just migrated to AIX and a new SAN) and could restore a 100GB production database in 45 minutes.
The box wasn't being used. I/O wait was at 1% the entire time. Each of the 8 CPUs was 90% idle the entire time. Of course memory was maxed out because PostgreSQL uses the OS to cache for it but we weren't using any swap. This was using the native PostgreSQL compressed backup format.
Oddly enough for PostgreSQL, I had less insight into what the database was doing during that time than I would have with DB2.
In DB2 I can make memory changes on the fly - db cfg, dbm cfg and speed this process up. I can use db2mtrk to see what my memory is doing. I have things like bufferpools to allocate memory where it's really needed.
With postgresql, I can change a text file (which I love) but have to restart postgres for a lot of them to take effect. Some db2 changes require an instance restart as well but not many anymore.
Some of the problem lay with me and I'll admit that but some also lay with PostgreSQL.
The whole point is that DB2 and Oracle don't normally go after the same market as MySQL and PostgreSQL. Are there companies using those databases in place of DB2 or Oracle? Sure. And I'm sure they're very happy and have a nice humming system. Our warehouse runs wonderfully on PostgreSQL and there are no complaints but more often than not, the markets simply don't intersect.
"Fighting the underpants gnomes since 1998!" "Bruce Schneier knows the state of schroedinger's cat"
Sun haven't been particularly enthusiastic about open source in the past.
The thing is, Sun has had to do a lot of work behind the scenes to get to where it is today on Open Source.
For example, it's about half a decade now since the project to open-source Solaris was started. There was an incredible amount of legal, engineering and commercial work to be done to get there.
These things don't happen on a whim.
Um, kill -HUP forces postgres to reload config, it works for some (all?) configuration changes, and I didn't notice it being a real restart- clients don't get disconnected.
Fix me if I'm wrong, i didn't use this feature much. But it worked for me when I needed it.
--Coder
- DROP the java front ends for everything. We get gray waiting for loadtimes.
Sometimes I think Sun really didn't think out the Java GUI experience very well before implementing it. The reason you get those blank screens during load times is how swing threads. It uses the same thread for event handling as for screen redrawing. From a programming stand point, I'm sure it makes it much simplier to use their API's for simple GUI's. However, when you've got tools written for system administration that will almost definatly take some time to process an event, it makes for a bad end user experience. Java is a great language. However, their poor implementation of the GUI API's makes the end user experience bad. And ultimately people who use java programs think the whole language sucks because of a bad user experience with the GUI.
If an officer ever threatens to taze you, say you have a pacemaker.
When they made the Postgres engine SQL compliant, they changed the name to PostgreSQL.
Sun's main issue is PR, I suspect. When IBM does something good, it makes sure everyone knows. But that doesn't seem to be McNealy's style...
Indeed. Sun is only just learning about PR (mostly the hard way).
> took 8 hours to restore!
You may want to check out the comments about the "checkpoint_segments" configuration parameter here; tweaking that appears to improve bulk loading performance considerably.
PostgreSQL is doing a fine job for my database, although it's a much smaller installation than yours. Only 4M records, but, hey.
The Army reading list
interesting. Do you remember when Sun was going to release OpenSolaris and they held back. The reason they held back was that internal benchmarking found that Linux 2.6 was killing everything that they had. They had to redesign and recode their networking (according to a friend of mine who did this, they borrowed heavily from the OSS world for ideas; but he swears it will beat them for a while). In addition, the reason to hold of their filesystem was that they found out that it was losing BIG time. It is in a massive rewrite. It turned out that Linux scaled and ran quite well.
Some of us like having beefy Sparc or Opteron SMP machines that perform predictably with Solaris, rather than the erratic behaviour we've seen with Linux on SMP Intel hardware.
Odd. I have not seen any of this erratic behaviour. In fact, I have 5 machines at my house and 5 others in a business that run great(all on 2.6).
The 2.6.x Linux kernel has also been a serious disappointment in terms of reliability, a definite step back from 2.4.x.
Actually it hasn't been. From what I have ran, it has been flawless. In actuallity, I know that there are rough edges, but not in the core. And the rough edges are places that Solaris does not do.
I prefer the "u" in honour as it seems to be missing these days.
Libraries. All 'Linux' tells you is where the system calls live (which number for which call). Red Hat Enterprise Linux tells you what libraries you can expect to find installed - i.e. what you can install without requiring additional dependencies. This is why some companies only support RHEL or SuSE (or whatever), rather than 'Linux'.
I am TheRaven on Soylent News
Yes, you think that's insecure, but the truth of the matter is that giving individual users their own MySQL username and password does not make it any less insecure. I am of the opinion that it's better not to lull people into a false sense of security: if they can see how sharp the blade is, they will be more careful when using a powerful tool.
Fact: it's trivial for any user with an account on a box to read any other user's files, even in their cgi-bin, since they must necessarily all be visible to the Apache daemon user {www-data on Debian systems}. And there's no way to obfuscate the database password: ultimately, the script has to send it to the server in the clear, so all you have to do is make a copy of the relevant file and replace a line that looks something like with That's bad enough in itself; but if the hosting company has decided to use the same password for MySQL and Linux login {and therefore POP3, FTP and maybe even shell access if they're on Gold} -- and there is at least one hosting company out there that are doing this {I had a reseller account with them once; I shan't name them} -- then Sir Hacksalot has the power to compromise more than just your database. One doesn't need to be terrifically "l33t" to find out which hosting company a competitor is using {as hard as it may be for you geeks who all have your own servers in your back bedrooms [and no hosting customers, except your own sisters' girly photo blogs, average 3 unique visitors per month, all bots] to believe this, there a lot of businesses who use hosting companies -- and more than a few who get their hosting done through cheapskate resellers}, get another account on the same box, and cause as much trouble as one can with a MySQL-based site.
The only way around this is for every user to run their own instance of the Apache server as themself, on a different non-privileged port; and to have a transparent proxy on port 80 that redirects requests to the appropriate port based on the host name. This way, users' files don't need to be readable to anyone save that user. Although it still would not be wise to use the same password for the two services, because a database password can still be exposed by careless use of chmod. And I wouldn't like to think how that is going to affect performance.
Je fume. Tu fumes. Nous fûmes!
I've found that Solaris 10 has so many problems that I wouldn't dream of using it even in a box at my house that I test on. It's proven to be trash and not worth the ascii text it's written in.
Now, we are in the same boat. I made several claims that one OS sucks yet I didn't list any references to support my claim.
That force me to discount your entire statment and all creditability.
Solaris 10 is a very nice OS, but my Oracle 10g RAC runs quite nicely on RHEL3 x86-64. (SMP Opterons (DBs) and SMP Xeon (app servers)) We migrated away from Sun for several reasons. Solaris 10 fixed many of them, but today it's a little to late and we arn't willing to fix whats not broke. I mean why add cost and work if everything is working as it should?
I would love to hear where your "serious disappointment in terms of reliability" stems from. I can tell you where my disappointment in Solaris 7-8-9 stem from but I see no reason to bash Solaris. I found Linux superior in almost every way for what we use it for in a weighted comparison. (YMMV) It's why we use it now and will continue to until we find enough reason to switch again. (Obviously once it's time to upgrade again)
Anyway, if you make a claim support your claim with references otherwise you shoot your creditability in the foot.
If it wasn't for DTrace, Zones, and ZFS I would stick with Solaris 8 for even longer.
I'm sorry, I don't often reply on Slashdot, but I have to reply this because it's so pointedly wrong. Sun didn't open source solaris to win support contracts and sell hardware. They open sourced it to generate a community around it and to increase developer and academic interest. The engineers working on Solaris have been wanting to open source Solaris for ages to try and increase adoption. I'm on the OpenSolaris engineering team, so I'm pretty certain I know what's going on around me here.
I am not saying that I am going to run Solaris 8 forever. I am just saying that I am speeding up the transition to Solaris 10 only because of the features it offers. They make my life as sysadmin much easier.
I've used both firebird and postgresql for the last 7 years or so. My only complaint with firebird was that the Java JDBC driver was so much slower than the postgresql driver (this was a couple of years ago). I've been using PostgreSQL exclusively for the past 3 years or so. It isn't as easy to admin as Firebird, but also offers more functionality. For example, PostgreSQL lets you define procedures in nearly any language (python, perl, java, c, c++, etc). PostgreSQL's PL/SQL stored procedure language is (I hear) close to Oracles. Not that Oracles stored proc language is the end-all be-all, but consistency is nice.
The PL's (like PL/PGSQL) are not for optimizations. They are for building triggers, etc.
Ok, there are a few areas where PostgreSQL performance sucks. Try doing an outer join between a moderately sized table (few million rows) and an empty table (zero rows). The result will be a nested loop join which will eat up your processor time for an insanely large time. The answer here is either to add a row to the empty table and vacuum analyze or rewrite your query to avoid the join against the empty table since this is meaningless anyway (this bad plan is a corner case in an optimization to prevent bad plans when tables grow rapidly between vacuum analyze routines).
A second area is that it is possible to create overly complex systems that are impossible to properly optimize. I have run into these cases before. Views of aggregates of views of aggregates are a great example. One should generally try to keep everything as simple as possible.
However.... Your suggestion that web server processing power is always cheaper than DB processing power is interesting but there are times when processing on the db server is cheaper than processing on the web server. For example:
1) If you can efficiently do joins and save the round trips, do them on the db side.
2) Data integrity is best enforced on the DB side if you have more than one app inserting or updating data in a DB. This is in addition to front-side checking and ensures that a bug in one app doesn't screw up data for another.
3) Look into PgPool and Slony-I for load balancing solutions that should enable you to add processing power for select statements at least as easily as you add processing power for web servers. This is best done when you have an app with lots of reads and a few writes (most web apps are this way). If you need something more complex, these technologies can be used to create more complex replication and load balancing solutions including multimaster replication systems in any architecture you can imagine with whatever conflict resolution algorythms you wish to use.
In short, it is not always true that adding processing power on the web server end is the least costly way to approach this.
LedgerSMB: Open source Accounting/ERP