When do you introduce a search server. The current farm was searching document at 6 hour intervals with incremental search only focused on a small segment of the 2 terabyte content in the farm. The reason for such slow search was because of the impact of running the search more aggressively. It has always been know the impact of search on the application tier of a farm but I never thought anyone would not see the need to scale and isolate given the need to produce search results in a small to medium farm.
So when do you introduce a search server into a SharePoint farm.
The volume of content that you have in your search index affects what resources you need to host the farm. Work out approximately the number of items that you plan on making searchable. Here are some examples of items: documents, web pages, SharePoint list entries, and images. Remember that each entry in a SharePoint list counts as one item.
When you have established a figure, multiply it by what you think the expected growth of that content will be over the next 12 months.
For example, if you’re starting out with 12,000 indexed items, and you expect the volume of that content to triple over the next 12 months. You should plan for 36,000 searchable items.
It’s not always easy to assess how big or small to make your search architecture. The size of your search architecture depends on the volume of your content, the crawl rate, the query throughput, and the level of high availability that you require. There are sample search architectures that we advise using as a basis to plan your own farm. The sample search architecture that you choose depends on how much content has to be searchable:
|Volume of content||Sample search architecture|
|0-20 million items||Small search farm|
|0-80 million items||Medium search farm|
|0-200 million items||Large search farm|
|0-500 million items||Extra large search farm|
Although these sample search architectures use virtual machines, you can use both physical servers and virtual machines according to the strategy of the overall SharePoint Server 2016 solution of your search architecture.
If you have up to 20 million items, the small search farm will probably be the most suitable farm for you. We’ve estimated that this search architecture can crawl 50 documents per second, and serve in the order of 10 queries per second. With a crawl rate of 50 documents per second, it takes search 110 hours to crawl 20 million items in the first full crawl.
If you have between 20 and 80 million items, the medium search farm will probably be the most suitable farm for you. We’ve estimated that this search architecture can crawl 100 documents per second, and serve in the order of 10 queries per second. With a crawl rate of 200 documents per second, it takes search 280 hours to crawl 80 million items in the first full crawl.
In my case the performance of the farm was very poor. The general introduction of an isolated search server would help improve the crawl process but the solution could also be achieved by improving performance by increasing RAM and retuning the search crawl. To test this I went to my O365 tenant and to the existing site Farm and preformed a search test:
- Up loaded a small document with a unique word pattern that could be searched and timed for how long it took to introduce the document to the crawl results.
- MY O365 Ternate was almost immediate. The document showed up within seconds.
- The Farm took hours.
It’s obvious there needs to be some changes.