krow's Journal: Interesting note on IO

VA 4450 chokes on 47 inserts per second while a 22XX was able to due 78 last night in testing without issue.

My cheap ass dual proc tyan board can do 112 on IDE drives (I was trying to stress it or the 22xx, I just wanted to see if it was better).

The 4 proc Intel motherboard inside of the 4450 sucks ass for IO.

Right now Slashdot has four of these (one is on loan). One is just dedicated as a search box. I leave one as a writer and use two as readers (BTW if you notice page glitches with out of date data this is the issue, I doubt we have hunted down all of the bugs in having writers and readers at the moment).

Want to bet a bunch of cheap dual proc/2gig boxes would blow away the 4450's (aka 4 proc/4gig boxes)?

Re:Once again, thanks for the insight.

[krow]

Well I did find two bugs but neither of them point to anything that says why your people object was empty (the data was in the database though so I could rebuild it).

Zoo works like this. There are tables that track all of your relationships, when you add/delete/whatever a user we look at those tables, update your relationships, and then rebuild a hash that is stored as a blob. When you come into the site we thaw out that hash and make use of it.

So your hash was empty but all of your data was sitting around. Normally this will only happen if something really bad happens to the site/DB while saving or your hash gets to big for the table (I recently increased this though so I don't think that will be happening any longer to anyone).

Account rebuilt, sorry that it happened in the first place. Once I check out a few more things I will probably rebuild all of Slashdot. Takes about 20 minutes and I have to lock Zoo during that time.

Re:Once again, thanks for the insight.

[Xaoswolf]

HTTP/1.1 200 OK Date: Fri, 07 Feb 2003 18:53:34 GMT Server: Apache/1.3.26 (Unix) mod_gzip/1.3.19.1a mod_perl/1.27 mod_ssl/2.8.10 OpenSSL/0.9.6g X-Powered-By: Slash 2.003000 Connection: close Content-Type: text/html; charset=iso-8859-1 OK

The server encountered an internal error or misconfiguration and was unable to complete your request.

Please contact the server administrator, [e-mail] and inform them of the time the error occurred, and anything you might have done that may have caused the error.

More information about this error may be available in the server error log.

=========

Any idea as to why I get this any time I try to add a friend? Everything else seems to work fine.

Re:Once again, thanks for the insight.

[krow]

Found a bug that caused your account to not be rebuilt (when we go to store your object there was a bad piece of SQL in case that is specific to the amount of data that was stored in relationships for your account).
Bad SQL equals no page delivery in this case. By the time I hit send there will be new code in place and the servers will have been restarted to fix it.

Sorry about that. I can't test your account completely but I was able to find at least the one bug (so I am not promising that there won't be more...).

Re:Once again, thanks for the insight.

[heliocentric]

I am getting the same error type when I view relationships. In talking to other /.ers I have not yet found anyone who DOES NOT get the error. Those talked to include banky, AntiFreeze, sporty, gmhowell, and possibly one or two others whose names escape me at the moment. Perhaps this is a larger problem than is currently thought?

Re:Once again, thanks for the insight.

[heliocentric]

For example, this is the error message I receive when I view my relationship with krow:

HTTP/1.1 200 OK Date: Fri, 07 Feb 2003 20:58:02 GMT Server: Apache/1.3.26 (Unix) mod_gzip/1.3.19.1a mod_perl/1.27 mod_ssl/2.8.10 OpenSSL/0.9.6g X-Powered-By: Slash 2.003000 Connection: close Transfer-Encoding: chunked Content-Type: text/html; charset=iso-8859-1
OK
The server encountered an internal error or misconfiguration and was unable to complete your request.
Please contact the server administrator, [e-mail] and inform them of the time the error occurred, and anything you might have done that may have caused the error.

More information about this error may be available in the server error log.

slash's sql boxes

[tf23]

So how are you benchmarking the writes/second?

And does the amount of data, per write to the db, affect how many it can do per second, or is it instead a table/row lock issue?

Re:slash's sql boxes

[krow]

Go into the mysql client and type \s, it will show you.

IO affects how many writes we can do a second. Especially on a table that has no locks.

a bet!

[JDizzy]

Sure, I'll wager one Dollar on the 2gb systems, or whatever. Probably a sucker bet, but its only a buck! You would figure that IO on a quad box would have to deal with more complex schedular tasks that would get in the way of fast IO. Not to mention the weird affect where large memory systems have to have more complex and agressive VM systems. Put it all together and I would always wager that a 2Gb system would out perform a 4gb system, especially with less processors.

damn

[unterderbrucke]

'can do 112 on IDE drive'

damn, mine does 15 on IDE :-(