Higth availability indexes

Obed Silva

Hello everybody

 There will be some way to duplicate the Indexes on another server to act as a "server contingency", ie a "Hot Stand-by" of the indexing service.

A company   needs   high   availability of the data and I not know what solution to give.

I need your   help   or   any   idea.

Thnaks.

Appu_Nair

This would have gotten more responses had it been posted at search & indexing.Assuming the company does not have the wherewithal to actually procure a cluster solution many companies implement what I detail below.The basic premise of this solution revolves around understanding livelink and in particular search.

Livelink's search subsystem can be broadly broken up into indexing and searching system.Indexing is acheived by an extractor process that reads volatile changes in the livelink database and makes it available to the indexer.Once livelink indexes it basically writes it out to index files that could sit in shared folders and files.

For e.g if the company can afford downtime on indexing (new objects/updated objects will queue up until the indexer is built again,usually about 2 hrs is sufficient to repair an indexing system) then all you need is another admin server who can read the index files.

So a small e.g may help the math used just simple.

Assume I have a 8 core box then I can thoretically allocate 7 cores for search and indexing and will probably require about 24 GB of RAM.Now if I stand up another identical box and if the index locations can be shared then users will never see the dreaded "no search engine avilable for this data source".But as your livelink grows there could be a situation where my primary index server has occupied enough objects so that the primary partitions are full at this point I will have to overflow it into my secondary at which point if my primary server fails I will not have enough RAM in both boxes to search all the index partitions for a search the user is asking it.I hope you understand what we are alluding to.

Adding a secondary admin server to a livelink install is very easy.When I build this architechture I manually build the index processes on a local disk and then change the index locations to a share so it can be read by other admin servers.Data flow local allows livelink to do its indexing stuff very fast.If local attached tsorage can be sourced then latency can always be brought down.

Look for postings by a few stalwarts like David Templeton,David Kinch,Ben Shapiro,Raymond Tsui to name a few hwo contributes very well to the search and indexing area discussions.