< estimate capacity and performance for Managed Metadata Service (SharePoint Server 2013)Estimate capacity and performance for Managed Metadata Service (SharePoint Server 2013)

摘要:了解如何規劃使用測試的容量與效能資料的方式部署 SharePoint Server 2013 的受管理的中繼資料服務應用程式。Summary: Learn how to plan to deploy a Managed Metadata Service application for SharePoint Server 2013 by using tested capacity and performance data.

本文包含的資訊和相關的 Managed Metadata Service in SharePoint Server 2013 的調整大小和效能最佳化的建議。我們也會提供有關如何設定服務,以及結構的最大效能的服務應用程式資料庫的一些最佳作法。使用此資訊來決定是否在計劃的部署適合在測試中提供的容量和效能限制。This article contains information and recommendations that are related to sizing and performance optimization of the Managed Metadata Service in SharePoint Server 2013. We also provide some best practices about how to configure the service and structure the service application databases for maximum performance. Use this information to determine whether your planned deployment fits in the capacity and performance limits that our tests provide.

在 SharePoint Server 2013 中的下列新功能直接影響受管理的中繼資料服務及容量規劃的重要。在我們的測試資料集和測試案例中包含從這些功能的負載。The following new features in SharePoint Server 2013 directly affect the Managed Metadata Service and are important for capacity planning. The load from these features is included in our test dataset and test scenarios.

  • 在摘要 (「 我的網站或小組網站) 中的雜湊標記Hashtags in feeds (in My Sites or Team Sites)

  • 使用受管理的導覽的網站Sites using managed navigation

  • 新的用戶端可以呼叫 CSOM 端點New CSOM endpoints that clients can call

一般 SharePoint Server 2013 部署具有類似特性為我們的測試資料集,我們建議的受管理的中繼資料服務應用程式會假設前端網頁伺服器角色的電腦執行。受管理的中繼資料服務應用程式的資料庫可放置在主控其他 SharePoint service 應用程式資料庫的 SQL Server 執行個體。For a typical SharePoint Server 2013 deployment with similar characteristics as our test dataset, we recommend that the Managed Metadata Service application run on the computers that assume the front-end web server role. The database for the Managed Metadata Service application can be placed on a SQL Server instance that hosts other SharePoint service application databases.

對於包含大幅多個項目比相較於我們在測試中所提供的資料集的部署,我們將會嘗試指導您如何透過我們變更其中一個變數並保留我們的測試資料集的其他元件常數的測試的結果。具有較大型部署的使用者應考慮這些結果和大小其 SharePoint Server 2013 伺服器陣列據此。For deployments that contain significantly more items than compared to the dataset we provide in our tests, we'll attempt to provide guidance through the results of tests where we change one of the variables and keep the other components of our test dataset constant. Users with larger deployments should consider these results and size their SharePoint Server 2013 farms accordingly.

如需容量管理及如何規劃 SharePoint Server 2013 的一般資訊,請參閱容量管理與大小調整為 SharePoint Server 2013For general information about capacity management and how to plan for SharePoint Server 2013, see Capacity management and sizing for SharePoint Server 2013.

簡介Introduction

我們建議在一般 SharePoint Server 2013 部署中具有類似特性為我們測試資料集的受管理的中繼資料服務應用程式能夠只執行為前端網頁伺服器的電腦上。受管理的中繼資料服務應用程式的資料庫可以置於主控其他 SharePoint service 應用程式資料庫的 SQL Server 執行個體。We recommend that in a typical SharePoint Server 2013 deployment with similar characteristics as our test dataset, that the Managed Metadata Service application be enabled only on computers that run as a front-end web server. The database for the Managed Metadata Service application can be put on a SQL Server instance that hosts other SharePoint service application databases.

請注意包含大幅多個項目比中我們在測試中所提供的資料集的部署,我們提供不同的建議。在我們的測試我們嘗試指導您如何使用測試結果我們變更其中一個變數並保留其他元件常數。具有較大型部署的使用者應考慮這些結果和大小其 SharePoint 伺服器陣列據此。Note that for deployments that contain significantly more items than are in the dataset that we provide in our tests, we provide separate recommendations. In our tests we attempt to provide guidance using the test results where we change one of the variables and keep the other components constant. Users with larger deployments should consider these results and size their SharePoint farms accordingly.

測試資料集Test dataset

若要提供容量規劃指引,我們會建立測試資料集包括在 SharePoint Server 2010 和 SharePoint Server 2013 中的新功能的功能及 SharePoint 部署上執行測試。下表顯示我們使用的測試資料集:To provide capacity planning guidance, we created a test dataset that includes the features in SharePoint Server 2010 and new features in SharePoint Server 2013 and ran tests on a SharePoint deployment. The following table shows the test dataset that we used:

VariableVariable 項目數Number of items
字詞集群組Term set groups
500 個500
字詞組Term sets
1000 個 (每個群組 2)1,000 (2 per group)
受管理的字詞 (不包含企業關鍵字)Managed terms (does not include enterprise keywords)
20000 (20 個字詞組)20,000 (20 per term set)
企業關鍵字Enterprise keywords
8000080,000
雜湊標記Hashtags
200,000200,000
總字詞 (包含受管理的字詞、 企業關鍵字和雜湊標記)Total terms (includes managed terms, enterprise keywords, and hashtags)
300000300,000
標籤Labels
300000 (每個項目 1)300,000 (1 per item)
字詞庫標籤長度Term label length
每個標籤 30 個字元30 characters per label

測試案例Test scenarios

我們使用下表中的測試此資料集:We used the tests in the following table for this dataset:

測試Test 描述Description 在測試中的百分比Percentage in test
GetSuggestionsGetSuggestions
單一呼叫 web 服務] 和 [單一字元前置字元字串。字串會選擇要比對字詞庫中,類似於中繼資料欄建議 SharePoint 使用者介面中的運作方式的 20%的字詞。A single call to the web service and a single character prefix string. The string will be chosen to match 20% of the terms in the term store, similar to how metadata column suggestions work in the SharePoint user interface.
10%10%
GetMatchesGetMatches
儲存在 web 服務中的條款字詞的 1%會比對的字串。Web service for a string that matches 1% of the terms in the term store.
5%5%
ValidateTermsValidateTerms
若要驗證單一字詞的 web 服務呼叫。Web service call to validate a single term.
5%5%
CreateTaxonomyItemCreateTaxonomyItem
Web 服務呼叫呼叫建立關鍵字的名稱。Web service call to a call to create a keyword with a random name.
5%5%
GetChildTermsInTermSetWithPagingGetChildTermsInTermSetWithPaging
Web 服務呼叫至多個字詞組。類似於所撥出通話 SharePoint Server 2013 來擷取受管理的導覽功能所使用的字詞組,其中結果然後快取中的前端網頁伺服器。Web service call to multiple term sets. Similar to the call made by SharePoint Server 2013 to retrieve terms sets used by the managed navigation feature, the results of which are then cached in the front-end web server.
5%5%
GetTermSetsGetTermSets
若要取得的字詞組呼叫 web 服務。Web service call to get term sets.
5%5%
GetTermsByLabelGetTermsByLabel
Web 服務呼叫的是字詞 Guid 的清單。類似 「 我的網站首頁上載入時進行 SharePoint Server 2013 的通話。Web service call with a list of term GUIDs. Similar to the call made by SharePoint Server 2013 when a My Site home page is loaded.
10%10%
HT GetSuggestionsHT GetSuggestions
Web 服務呼叫以取得單一字元的雜湊標記的建議。類似於所撥出通話 SharePoint 使用者介面時使用雜湊標記的摘要。Web service call to get suggestions for hashtags with a single character. Similar to the call made by SharePoint UI when you work with hashtags in the feeds.
10%10%
HT NewHashTagHT NewHashTag
2 web 服務的通話。若要取得條款,然後另一個則是建立字詞之一。2 web service calls. One to get terms, then another to create the term.
5%5%
HT GetTermsByLabelHT GetTermsByLabel
若要取得現有的字詞的 web 服務呼叫。當湊標記用摘要張貼中使用。Web service call to get existing terms. Used when a hashtag is used in a feed post.
15%15%
HT AddAssociationHT AddAssociation
若要將關聯新增至現有的湊標記字詞呼叫 web 服務。Web service call to add an association to an existing hashtag term.
5%5%
CSOM GetTermsCSOM GetTerms
測試以 GetTerms 詞彙 Guid 的清單。有 2 個別的 CSOM 呼叫。Test to GetTerms a list of terms GUIDs. Has 2 individual CSOM calls.
10%10%
CSOM SetPropertyCSOM SetProperty
測試以 SetProperty 單一字詞。具有 4 個個別的 CSOM 呼叫。Test to SetProperty for a single term. Has 4 individual CSOM calls.
5%5%
CSOM CreateTermCSOM CreateTerm
將字詞新增至字詞儲存區的測試。具有 4 個個別的 CSOM 呼叫。Test to add a term to the term store. Has 4 individual CSOM calls.
5%5%

在我們的測試我們置於詳細權數預期更常使用的作業。In our tests we put more weight on operations that are expected to be used more frequently.

測試拓撲Test topology

我們執行測試實驗室中的環境具有如下圖所示的拓撲:We ran the tests in our lab environment that has the topology shown in the following diagram:

圖 1: 測試實驗室伺服器拓撲Figure 1: Test lab server topology

這個 Visio 圖表顯示測試伺服器拓撲,其中包含裝載 SQL Server 的單一電腦和裝載 SharePoint 伺服器的單一電腦,此電腦可當成應用程式伺服器或前端網頁伺服器來執行。

我們測試入門一部電腦負責執行的 Managed Metadata Service 應用程式及服務的前端網頁伺服器。稍後,我們新增其他電腦具有相同的設定。We started our tests with one computer that is running the Managed Metadata Service application and serving as the front-end web server. Later, we added another computer that has the same configuration.

測試結果Test results

我們使用資料集和所述的案例在舊版的各節來測試執行的特定設定的受管理的中繼資料服務應用程式作業總數。We used the dataset and the scenarios that were described in earlier sections to test the total number of Managed Metadata Service application operations that executed for a given configuration.

我們已測試執行與下列清單中的不同負載設定檔:We ran our tests with the different load profiles in the following list:

  • 綠色區域Green Zone

    伺服器下為 60%使用率。這應該是目標的大部分的伺服器正在執行時的時間。Servers are under 60% utilization. This should be the target for most of the time when the servers are running.

  • 紅色區域Red Zone

    伺服器很接近完整的使用情況。這可視為較平日的更多負載下的 SharePoint 網站所在的狀態。紅色區域中,伺服器回應時間值開始逐漸伺服器會嘗試以符合傳入要求的需求。Servers are close to full utilization. This can be considered a state where the SharePoint site is under more load than usual. In the Red Zone, server response time values start increasing as the server tries to meet the demand of incoming requests.

下表顯示我們使用單一電腦的度量值的結果:The following table shows the results of our measurements with a single computer:

綠色區域Green Zone 紅色區域Red Zone
伺服器回應時間第 50 個百分位數的讀取作業:Server Response Time 50th percentile for read operations:
32 毫秒。32 ms.
44 毫秒。44 ms.
伺服器回應時間的讀取作業的 95th 百分位數:Server Response Time 95th percentile for read operations:
1090 毫秒。1090 ms.
1335 毫秒。1335 ms.
伺服器回應時間第 50 個百分位數對於寫入作業:Server Response Time 50th percentile for write operations:
1837 毫秒。1837 ms.
2038 毫秒。2038 ms.
伺服器回應時間對於寫入作業的 95th 百分位數:Server Response Time 95th percentile for write operations:
2283 毫秒。2283 ms.
3515 毫秒。3515 ms.
測試每秒完成:Tests completed per second:
99
1515
平均 CPU (應用程式伺服器或前端網頁伺服器)Average CPU (application server or front-end web server)
56%56%
92%92%
平均 CPU (SQL Server)Average CPU (SQL Server)
7%7%
12%12%
尖峰記憶體使用量 (應用程式伺服器或前端網頁伺服器)Peak memory usage (application Server or front-end web server)
6 GB6 GB
6.2 GB6.2 GB

我們部署具有相同的設定則新增第二個的應用程式伺服器或前端網頁伺服器虛擬機器 (Vm)。下表顯示具有兩個 VM 伺服器陣列的結果約兩次 (2x) 為舊版測試負載:We then added a second application server or front-end web server virtual machines (VMs) to the deployment with the same configuration. The following table shows the results for a two VM farm with approximately two times (2x) the load as our earlier tests:

綠色區域Green Zone 紅色區域Red Zone
伺服器回應時間第 50 個百分位數的讀取作業:Server Response Time 50th percentile for read operations:
44 毫秒。44 ms.
110 毫秒。110 ms.
伺服器回應時間的讀取作業的 95th 百分位數:Server Response Time 95th percentile for read operations:
1161 毫秒。1161 ms.
1679 毫秒。1679 ms.
伺服器回應時間第 50 個百分位數對於寫入作業:Server Response Time 50th percentile for write operations:
1828 毫秒。1828 ms.
2253 毫秒。2253 ms.
伺服器回應時間對於寫入作業的 95th 百分位數:Server Response Time 95th percentile for write operations:
3321 毫秒。3321 ms.
4648 毫秒。4648 ms.
測試每秒完成:Tests completed per second:
1515
2828
平均 CPU (應用程式伺服器或前端網頁伺服器)Average CPU (application server or front-end web server)
49%49%
88%88%
平均 CPU (SQL Server)Average CPU (SQL Server)
14%14%
28%28%
尖峰記憶體使用量 (應用程式伺服器或前端網頁伺服器)Peak memory usage (application server or front-end web server)
6.1 GB6.1 GB
6.3 GB6.3 GB

下圖顯示先前的兩個資料表中的資料:The following graph shows the data in the previous two tables:

圖 2: 受管理的中繼資料服務應用程式的效能Figure 2: Managed Metadata Service application performance

這個 Excel 長條圖顯示來自先前表格中的 Managed Metadata Service 應用程式效能資料。第一筆效能資料顯示單一應用程式伺服器或前端網頁伺服器,接著針對綠色與紅色區域顯示兩部雙倍負載的電腦。

影響搜尋的易記 Url 編目的易記的 UrlEffect of friendly URLs on Search crawls for friendly URLs

SharePoint Server 2013 包含受管理導覽功能及觸發以確保搜尋編目收到使用易記 Url 頁面的最新版本的兩個計時器工作。測試中發現執行這些計時器工作與受管理的中繼資料服務應用程式沒有顯著影響傳達給讀取與寫入作業的目標相同的應用程式。SharePoint Server 2013 includes the Managed Navigation feature and two timer jobs that are triggered to make sure that the search crawl receives the latest versions of pages that use friendly URLs. Our tests found that running these timer jobs that communicated with the Managed Metadata Service application did not have a significant effect on read and write operations that target the same application.

CSOM 呼叫的影響Effect of CSOM calls

一些測試增加相較於基準案例我們測試程序中所做的 CSOM 呼叫。我們增加 20%的測試 66%以下。測試結果顯示執行測試的號碼拒絕來自 28 以每秒 22。伺服器回應時間會相較下的基準。在執行測試數目減少是因為增加額外的每個動作的相較於較少呼叫 web service 通話方的 CSOM 呼叫前端網頁伺服器的多個通話。Some tests increased the CSOM calls made in our test procedure compared to the baseline case. We increased these from 20 percent of tests to 66 percent. The test results show that the number of executed tests declined from 28 to 22 per second. The server response times are comparable to the baseline. The reduction in the number of executed tests is because of the increased overhead of multiple calls to the front-end web server for each action for the CSOM calls compared to fewer calls involved in the web service calls.

具有多個湊標記作業的影響Effect of having more hashtag operations

在另一個測試案例中,我們增加百分比 35%湊標記作業中 71%我們比較基準測試。結果會顯示執行測試的號碼拒絕來自 28 至 19。伺服器回應時間也包含大於比較基準約 30%。在執行測試數目減少會對應到相當高的雜湊標記字詞組中的字詞數以及更加增加的百分比變更的程序中的寫入作業。In another test case, we increased the percentage of the hashtag operations from 35 percent in our baseline to 71 percent in our tests. The results show that the number of executed tests declined from 28 to 19. The server response times are also about 30 percent greater than the baseline. The reduction in the number of executed tests is attributed to the fairly high number of terms in the hashtags term set and the much increased percentage of write operations in the changed procedure.

另請參閱See also

概念Concepts

在 SharePoint Server 中規劃受管理的中繼資料Plan for managed metadata in SharePoint Server

其他資源Other Resources

在 SharePoint Server 2013 中受管理的中繼資料服務應用程式概觀Overview of managed metadata service applications in SharePoint Server 2013

Plan terms and term sets in SharePoint Server 2013Plan terms and term sets in SharePoint Server 2013