Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
FAST for SharePoint Deep Dive
1. Enterprise Search Customer Case StudyFAST for SharePoint Deep Dive Andy Hutchins – Collaboration Service Line Lead Neil Richards – Solutions Architect David Cohen – Infrastructure Architect 11 April 2011 1
26. Session Overview Client Solution Overview Demo: Configuring a User Context Demo: Configuring FAST using PowerShell Demo: Debugging FAST Key Tools 6
27. Client Solution 7 International company with operations in over 50 countries and 70,000 employees Consolidating a number of back office functions into regional service centres based in Europe, Asia, Australia, Africa and the United States Support their drive for efficiency in the service centers through a single, unified Knowledge Management solution
28. The Client Challenge 8 SharePoint Chaos Content Types File shares Managed Metadata WSS2 Team Sites Site structure Documentum LOB Data Search
29. Client Solution 9 SC SC SC Uniformity of content types through content-type publishing and replication of document library settings Consistent classification driven through a global taxonomy Similar but tailored site structures SC SC
41. Lessons Learned 12 Get your hands dirty. Embrace PowerShell, the QR Server and the FFD Dumper for configuration and debugging Prototype functionality early Undocumented but useful tools Don’t assume FAST understands how SharePoint works
43. Creating a User Context Configuration via UI is pretty simple Web Application account needs specific permissions on the FAST server Add new properties with 3 simple lines of PowerShell $contextProps = Get-SPEnterpriseSearchExtendedQueryProperty-SearchApplication "FASTQuery" -Identity "FASTSearchContextProperties" $newPropertySet = $contextProps.Value + ",<newProperty>" Set-SPEnterpriseSearchExtendedQueryProperty-SearchApplication "FASTQuery" -Identity "FASTSearchContextProperties" -Value $newPropertySet 14
47. Debugging FAST QR Server accessible via http://localhost:13280 on the FAST server (must be logged in) Debugging process Search via UI Copy querystring from %FASTSearch%/var/logs/querylogs/<mostrecentfile> Paste into QR Server query By default you only have 15 seconds before security token times out. 18
48. Key Tools QR Server debugging querieshttp://www.neilrichards.net/blog/?p=100 FFD Dumper for inspecting crawled propertieshttp://www.neilrichards.net/blog/?p=153 MossMan FAST Query tool http://fastforsharepoint.codeplex.com/ FASTSearch2010.chm for PowerShell http://www.microsoft.com/downloads/en/details.aspx?FamilyID=ead44c66-7d02-4edf-9e56-2f56c6f59f22&displaylang=en Information Worker Demo VM http://www.microsoft.com/downloads/en/details.aspx?FamilyID=751fa0d1-356c-4002-9c60-d539896c66ce&displaylang=en PowerGui http://www.powergui.org/index.jspa?sid=0 19
49. FAST deep dive David Cohen infrastructure architect - collaboration
50. FAST for SharePoint Overview FAST for SharePoint (FS4SP) is a separate install from SharePoint FS4SP and SharePoint can co-exist on the same server, but this isn’t supported in production Two Service Applications created to connect the two One “Content” SA, which is where crawling/feeding is configured One “Query” SA, which is where querying is configured Licensing Requirements Requires a SharePoint Server license, per server Requires Standard CALs Requires Enterprise CALs Requires FS4SP license, per server Check with your Large Account Reseller
51. FAST for SharePoint Architecture Rows and Columns Columns add scale for indexing Around 15m documents per column Up to 40m documents when in high capacity mode (reduces query throughput) Rows add resilience and query capacity Depending on hardware, up to 50 queries per second (QPS) per server, is achievable Row 0 Column 2 Row 0 Column 1 Row 0 Column 0 Row 1 Column 1 Row 1 Column 0 Row 1 Column 2
52. FAST for SharePoint Architecture Contd. Roles are defined in the deployment.xml for multi-server installs. The configurable roles are: Admincontains roles like the name and security services Content Distributorshares content amongst the document processors Document Processorprocesses content before it’s indexed e.g. language detection, stemming etc. Indexing Dispatchersends content to the right column to be indexed Crawleryes FS4SP has separate crawlers as well as SharePoint Web Analyzeranalyzes hyperlink structures Querycontrols if the server will respond to search queries Search Enginecontrols the layout of the rows and columns Search Clustersets which are the primary and backup indexing nodes
53. FAST for SharePoint Limitations If you require high availability, the Admin role is key Admin role is a single point of failure Can only be installed on one server in the farm Impact of loss of Admin Server Crawling will stop You won’t be able to restart the Windows service handling queries Mitigation Separate out the admin role to its own dedicated server. This virtualises well. Still requires a license Have a “warm standby” server on hand to restore your admin server onto Scale and Boundaries SharePoint Crawler – 25m documents per crawler FAST Server – 15 to 20m documents per server Queries – up to 50 QPS per server, per row FAST Farm – limitless? Microsoft have tested >400m documents
54. Summary Different set of servers, architecture and performance characteristics Key single point of failure with the admin node. Not yet fully documented by Microsoft. Get the FAST topology right first time Hard to change once it’s deployed Huge scale – hundreds of millions of documents Great performance – thousands of queries per minute
55. thankyou. questions? come and visit us at booth 6 in the exhibition hall use twitterto ask us a question #askavanade visit our web site www.avanade.com find more from Neil Richards http://neilrichards.net/blog @neilrichards
Notas del editor
PrepCleanup user context & site promotionsBrowser 1 Tab 1: Site Collection Settings httTab 2: Search CenterTab 3: Marketing SiteBrowser 2Tab 1: Search Center
PrepCleanup user context & site promotionsBrowser 1 Tab 1: Site Collection Settings httTab 2: Search CenterTab 3: Marketing SiteBrowser 2Tab 1: Search Center
PrepCleanup user context & site promotionsBrowser 1 Tab 1: Site Collection Settings httTab 2: Search CenterTab 3: Marketing SiteBrowser 2Tab 1: Search Center
Does have higher memory and cpu requirements and I’d recommend not to virtualise.Icons of servers instead of boxes