[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: ldap load measuring and reproduction tools



J.Smith wrote:


----- Original Message ----- From: "Mike Jackson" <mj@sci.fi>


The scalability problems which you will inevitably encounter will be a direct result of this poor design.
As a result of this poor design which you have to work with, you are now placed in the position in which everything else you do is going to be far from optimal.


Well at the risk of getting seriously off-topic here (and/or turning this into a flame-war) - this is not necessarily so. Some LDAP products let you do replication based on search filters - which (for replication purposes) effectively give you the same flexibility benefits as the hierarchical structure, but without the high-maintenance downside inherent to the hierarchical structure. If the business needs change, you can simply change the filters to meet them and you're all done. Also, some LDAP server products let you do something you might consider calling 'cascaded replication' - where ldap server A replicates to server B, and server B replicates to server C. This then takes the replication induced load of the single master server you have. And lastly, some products let you set up multiple master servers, increasing the availability and scalability of the 'write' part of LDAP as well. It's just that the product currently used can't/doesn't. So the position im currently in can just as easily be viewed as a shortcoming of the product currently used, as opposed to a fundamental flaw in the layout of the directory.

Nowhere in this email thread have you ever mentioned the name or version number of the "product" you're currently using. On the optimistic assumption that you're actually using OpenLDAP, I'll note that the current release (2.2.23) supports all of those features, and most of those features are also supported in 2.1, i.e., they've been there for a very long time.


On a side note, without knowing the size of your database or your query rate, or the hardware involved, it's pretty tough for anybody to credibly say whether updating 10-20 replicas will be practical or not. Anybody offering opinions on this question in the absence of such crucial details is just blowing smoke.

--
 -- Howard Chu
 Chief Architect, Symas Corp.       Director, Highland Sun
 http://www.symas.com               http://highlandsun.com/hyc
 Symas: Premier OpenSource Development and Support