| Pages: [1] :: one page |
| Author |
Thread Statistics | Show CCP posts - 0 post(s) |
Mortor Calieph
   |
Posted - 2006.03.09 20:09:00 -
[1]
Hey, something to check out and see if it might improve. My years of using RAM disks for compiling faster felt justified when I heard you guys moved to the RAM-SAN. The game sure feels a bit snappier now. Below is a link I took from one of my industry mailings:
'Father of DB2' to Mentor Super-Fast Database Startup
|
Noctoc
|
Posted - 2006.03.11 20:10:00 -
[2]
Looks interesting, but I couldn't find any real world comparisons between ants and mssql or oracle. Most of the info on the web site is marketing bs, I never trust performance comparissons on manufactures web sites. Also they say ants is a lot cheaper than mssql, but I could only see a +/-$500 diffrence in licensing fees. I known that sounds a big diffrence but when your talking $25k for a license (per cpu) its not that big a diffrence.
But if it gives us a performance boost in EVE I'll take it 
Due to lack of funds the light at the end of the tunnel has been turned off. |
Mortor Calieph
|
Posted - 2006.03.16 06:49:00 -
[3]
Yeah, need to ignore the adverts. I wasn't pushing anything from their adverts. I was just interested in the methods they were describing. Will have to see what becomes of MySQL now with the Oracle thing anyway.
|
Solntsev
|
Posted - 2006.04.10 09:08:00 -
[4]
first of all, being a 'father of DB2' may or may not be a compliment in the eyes of technical folks ;] secondly, lock-free concurrency algorithms aren't something new and exciting - they've been around and in heavy use for quite some time. in fact, some of the algos are at the core of how distributed transactions are handled in modern rdbmses. however, there are quite a few issues with using them the way the article seems to imply, the biggest of which is the difficulty of dealing with the worst case performance downside risk and reconciling their inherent semantics with the need for transaction success or failure to be deterministic. if anything, the most interesting technology for ccp to look at would be the resurrection of multi-value dbms tech in cache, since a lot of stuff in eve is much more suited to that model and cache gives you mv without the scourge that is mumps. unfortunately, it is unlikely that this is practical at this point - i'd expect ccp to have a lot of effort invested in sql server stored procs that wouldn't port over at all. this is simply too much of a risk at this point.
-s
|
| |
|
| Pages: [1] :: one page |
| First page | Previous page | Next page | Last page |