June, 2011

The Exchange Interoperability documentation team is pleased to announce the launch of Exchange Server Interoperability Guidance that supplements the information available in the Microsoft Exchange Server protocol documentation and the Microsoft Exchange Server and Outlook Standards documentation. Whether you’re an Exchange interoperability expert, just getting started, or somewhere in between, we’re confident that this supplemental information will be a valuable resource for you.

In the first article, Exploring the Microsoft Exchange Server Open Specifications, Kim Brandl reviews the types of protocols, structures, algorithms, and standards that the Microsoft Exchange Server Open Specifications define, discusses how to get started using the specifications, explores the kinds of content that you’ll find in the various types of specifications, and addresses some key points to help you derive the maximize benefit from the specifications. Check out Exploring the Microsoft Exchange Server Open Specifications now, and stay tuned for additional technical articles coming soon!

Kim Brandl
Exchange Interoperability
Microsoft Corporation

What are the throttling settings for Microsoft Exchange Online as part of Microsoft Office 365? Before I answer this question, let me give you bit of background information. Microsoft Exchange includes bandwidth throttling to help manage server access. This supports service quality and consistent performance. The throttling components of Exchange Online are especially important, given that network resources in the datacenters are optimized for the broad set of customers that use the service. So, to properly engineer solutions that work with Exchange Online, you will need to account for the Exchange Online throttling settings.

We are specifically focused on throttling of Exchange Web Services (EWS) and MAPI with RPC over HTTP (as used by Microsoft Outlook) and with hard limits set by the Exchange Online system. Any mention of specific setting values were obtained for the service version 14.0.100.0. You will want to have your tenant admin obtain and confirm setting values.

Note   An on-premise Microsoft Exchange deployment may have a different throttling policy. Administrators can modify throttling settings for Microsoft Exchange on-premises. Administrators cannot configure throttling policies for Exchange Online.

What is the impact of exceeding throttling budgets?

Exchange throttling is based on the amount of Active Directory, Exchange store, and other resources a client consumes. A budget is allocated to each client to limit the resources a particular user consumes. This budget is recharged every minute.

EWS clients can exceed their budgets by sending too many concurrent requests, or by requesting too many resources on the server. When this occurs, additional requests are queued until one of two things happen:

• The application is under budget. The request is then processed and the client experiences an increase in latency.
• The request has been queued for more than 60 seconds, after which time the server sends an error response to the application.

When MAPI clients exceed their budgets, additional requests are returned with an error, until the application is under budget. There is no queuing of requests.

How do I get throttling settings?

Tenant administrators can view throttling settings by using Remote PowerShell and the Get-ThrottlingPolicy cmdlet. For information about how to connect Remote PowerShell for your tenant, see Install and Configure Windows PowerShell and Connect Windows PowerShell to the Service on Microsoft TechNet.

Use the Get-ThrottlingPolicy cmdlet (without arguments) to get all the throttling policy settings. Throttling policies might change to enhance service performance, so you should confirm throttling policies. We suggest that you make your client applications configurable to adjust for potential changes in throttling policy for Exchange Online. Additionally, because Exchange Online and Exchange on-premises might have different throttling policies, you may want to account for this when you design your client applications.

What are the common throttling settings?

The following throttling settings are common to both EWS and MAPI (RPC over HTTP) clients:

• MessageRateLimit – Identifies the maximum number of messages that can be sent per minute. This is configured to count sent messages on a per IP address basis.
• RecipientRateLimit – Identifies the maximum number of recipients that a sender can send to in a 24-hour period.
• ForwardeeLimit – Identifies the maximum number of recipients that can be configured for Inbox rules when the user account is using the forward or redirect action.

What are the MAPI with RPC over HTTP throttling settings?

The prefix RCA in the MAPI RPC/HTTP settings represents RPC Client Access. The following throttling settings are specific to MAPI RPC over HTTP:

• RCAMaxConcurrency – Identifies the maximum number of concurrent connections that can be maintained by MAPI RPC/HTTP clients (such as Microsoft Outlook) at any one time. Attempts to open more than the maximum number of connections will fail. Existing connections will remain unaffected. Active Directory calls count toward the number of concurrent connections. 
• RCAPercentTimeInAD – Identifies the maximum amount of time that can be spent by a Client Access server when accessing Active Directory resources on behalf of a client, per minute. This budget is calculated by dividing the throttling value for this setting by 100 and then multiplying the result by 60 seconds. For example, an RCAPercentTimeInAD throttling value of 5 results in a throttling threshold of 3 seconds (5/100 * 60 seconds).
• RCAPercentTimeInCAS – Identifies how much time all concurrent MAPI RPC/HTTP connections can consume on the Client Access server per minute. This is calculated on a per user basis. This budget is calculated by dividing the throttling value for this setting by 100 and then multiplying the results by 60 seconds. For example, an RCAPercentTimeInCAS throttling value of 205 results in a throttling threshold of 123 seconds (205/100 * 60 seconds). If the setting is 123 seconds, two concurrent MAPI RPC/HTTP connections can be open on a Client Access server for a full 60 seconds, thus using up 120 seconds per minute. A third connection can only use the remaining 3 seconds per minute, assuming all three connections occur concurrently and are initiated at the same time. 
• RCAPercentTImeInMailboxRPC – Identifies how much time all concurrent MAPI RPC/HTTP connections can consume for accessing the mailbox per minute. This budget is calculated by dividing the throttling value for this setting by 100 and then multiplying the results by 60 seconds. For example, an RCAPercentTimeInMailboxRPC throttling value of 200 results in a throttling threshold of 120 seconds (200/100 * 60 seconds). This means that two connections can consume 60 seconds per minute (120 seconds total per minute) as long as there are no other connections.

What are the Exchange Web Services throttling settings?

The following throttling settings are specific to Exchange Web Services:

• EWSFastSearchTimeoutInSeconds – Identifies how long an Exchange Search (indexed search) request will continue until it times out.
• EWSFindCountLimit – Identifies the maximum number of search results in concurrent requests. For more information, see Throttling Policies and the EWS FindCountLimit. We suggest that searches all be paged and that the page size not exceed 1000 results.
• EWSMaxConcurrency – Identifies the maximum number of concurrent connections that can be maintained by EWS clients at any one time. Attempts to open more than the maximum number of connections will fail. Existing connections will remain unaffected. Active Directory calls count toward the number of concurrent connections. 
• EWSMaxSubscriptions – Identifies the maximum number of push, pull, or streaming notification subscriptions. If a server has too many notification subscriptions, it will start returning server busy errors. The server starts to return errors when the number of open connections reaches approximately 20 for a user. The charges for EWSMaxSubscription are calculated as follows:
  • Client – charged against the caller, also called the mailbox owner.
  • Intra-forest ExchangeImpersonation – charged against the caller.
  • Server-to-Server – charged against the underlying effective caller.
• EWSPercentTimeInAD – Identifies the maximum amount of time that can be spent by a Client Access server when accessing Active Directory resources on behalf of a client account, per minute. This budget is calculated by dividing the throttling value for this setting by 100 and then multiplying the result by 60 seconds. For example, a EWSPercentTimeInAD throttling value of 50 results in a throttling threshold of 30 seconds (50/100 * 60 seconds).
• EWSPercentTimeInCAS – Identifies how much time all concurrent EWS connections can consume on the Client Access server per minute. This is calculated on a per user basis. This budget is calculated by dividing the throttling value for this setting by 100 and then multiplying the results by 60 seconds. For example, a EWSPercentTimeInCAS throttling value of 90 results in a throttling threshold of 54 seconds (90/100 * 60 seconds).
• EWSPercentTimeInMailboxRPC – Identifies how much time all concurrent EWS connections can consume while accessing a mailbox via RPC requests, per minute. This is calculated on a per-user basis. This budget is calculated by dividing the throttling value for this setting by 100 and then multiplying the results by 60 seconds. For example, a EWSPercentTimeInMailboxRPC throttling value of 60 results in a throttling threshold of 36 seconds (60/100 * 60 seconds). This means that a single connection can consume 36 seconds per minute as long as there are no other connections. Two connections that started at the same time can only consume 18 seconds each per minute. This is based on a threshold setting of 60.

For more information about EWS throttling policy, see EWS Best Practices: Understanding Throttling Policies.

What are the specific Exchange Web Services limits?

The following are the specific EWS throttling limits:

• maxAllowedContentLength (IIS7) – Identifies the maximum accepted content length of HTTP requests. This maximum length is 35,000,000 bytes, which translates to 25 MB of base64-encoded data.
• maxReceivedMessageSize (WCF) – Identifies the maximum message size that is accepted by EWS. This maximum message size is 35,000,000 bytes, which translates to 25 MB of base64-encoded data.

Note   The values for this limit differ for an on-premise Exchange 2010 deployment. In the initial release version of Exchange 2010, the message size limit is 10 MB. In Exchange 2010 Service Pack 1 (SP1), the message size limit is 35 MB.

• maxRequestLength (ASP.Net) – Not used.