Strategies for Large Deployments

[kmccrery]

Hi, I am new to Thunderstone so please bare with me if my initial questions sound stupid. I recently purchased two search appliances for Virginia Tech to replace our free search from Google. I am in the process of setting things up and I have a few questions about strategies for large-scale deployments. Here is my scenario: I have between 300-500 sub domains under vt.edu I plan to index. Each of these domains typically represent a college or department level web site. I would like to provide each of these domains with their own site search. I would also like to have separate walk frequencies for some sites since some are more dynamic than others. My initial though is to set each domain up as its own profile and then do a meta search profile for the primary search site to query. However, it's quickly becoming obvious that this will require some considerable maintenance. Can anyone suggest a better method. Is there any way to better manage a large number of profiles? Something like a single XML file that can be uploaded to the appliance would be ideal.

One other issue I have come across is the using the Best Bet feature with the meta search. In the scenario above, I can assign Best Bet words at the profile level but not at the meta search level (because it has no URLs). However, the Best Bet results are not showing up in the meta search profile. Seems like Best Bets should appear on meta searches or you should be able to define keywords at the meta search level. Any ideas?

I appreciate any advice.

Ken McCrery

[kmccrery]

Looks like I solved the issue with Best Bet results from individual profiles not appearing on the meta search. I just needed to use the same Best Bet Group name in the meta search. Still looking for a way to more easily manage a large number of profiles though.

[John]

You may find the Backup/restore settings a useful way of managing a lot of profiles.

Depending on how the individual sites are organized you may want to group some of them together, and use the categories to provide the finer level search.

You may also want to create a separate profile for the site-wide search and have all the other profiles replicate into it, as a meta search of several hundred profiles may have a negative performance impact, and that is likely to be a frequently used search.

[kmccrery]

I'm looking at the backup/restore option as a way to manage a large number of profiles. It looks as though each profile gets a unique value for SS_dataspace and SS_db. Example:

<applianceoption>
<Name>SS_dataspace</Name>
<Profile>profile-name</Profile>
<Value>/usr/local/morph3/texis/profile-name.45e756d57</Value>
</applianceoption>

"45e756d57" is a unique part of the string that changes for each profile. Are there any details on this value? What if I created entries for new profiles in my XML file and just used random characters for this value? Would my appliance recognize it? I'm basically trying to get away from having to use the Web admin interface to add hundreds of domain profiles.

[mark]

Currently you have to create the blank profiles before you can restore settings. Creating them will assign the dataspaces automatically.