Exchange 2010 CAS Server Scalability and Limitations

I have asked recently this question , one of our costumer wanted to know what the CAS server limitations were in Exchange 2010. As the questions sounds broad to be more specific the client wanted to know how many  OWA users could continue to use the services if worst happens , in the scenario we end up losing such number of CAS 2010 Servers. Good question to be honest and here is pretty much all you need to know to get the bottom of this question.

First of all in the scenario of giving design decision, what makes the most sense? how many servers do we need for giving scenario ? Million thanks to Exchange team,  use the Calculator, it will tell you what you need and most of the leg work is being done by the calculator itself.

 

Calculator

http://blogs.technet.com/b/exchange/archive/2009/11/09/3408737.aspx

Now, should you be combining roles or not? It all depends in general , as good practice the simpler design is the “Best one” keep this in mind. So this being said, most engineers have been combining at least HTS/CAS roles anyways.

Should combine MBX role as well? the answer is depends and depends but same rule applies, “ keep it simple” it really is not a brainer, if you are going to architect messaging solution for 100K users , ask yourself how may DAGS you will need ? Take a look at business model (how they are operation in current) and its logical segregation use that to align your design solution and remember the DG boundaries are 16 Servers.

Now you are going to deploy 2 DAGs and each DAG will have 10 Servers in total you end up having 20 MBX Servers. Remember the Exchange calculator will do the hard work for you and how many combined roles servers (HTS & CAS ) you would need ? to accommodate this design ? The answer is going to be , 1 to 1 ratio ( simple isn’t it ) so

E2013 Data Loss Prevention Steve Chew

Option 1 = 60 Servers

HTS Servers	20 HTS Servers
Combined Servers ( CAS & HTS )	20 CAS Servers
DAG1 ( MBX )	10 MBX Servers
DG 2 ( MBX )	10 MBX Servers

Option 2 = 40 Servers

Combined Servers ( HTS & CAS )	20 CAS Servers
DAG1 ( MBX )	10 MBX Servers
DG 2 ( MBX )	10 MBX Servers

Option 3 = 20 Server

Multi Role Configuration	20 ( HTS & CAS & MBX )
DAG1	10 ( HTS & CAS & MBX )
DG 2	10 ( HTS & CAS & MBX )

Now remember the golden rule keep it simple, the “simple deployment” is the best one so use your logic and common sense to decide which one of these makes the most sense? You have to evaluate your own specific requirements as this fluctuates per given scenario and costumer.

Same story goes for choosing the platform, should we take advantage from virtualization technologies or deploy hardware solutions? I personally favor the virtualization deployments fir reasons I have talked about on my previous blog post.

Exchange 2010 Virtualization and Combining Server Roles?

http://smtp25.blogspot.com/2013/01/exchange-2010-virtualization-and.html

Finally Exchange 2010 CAS limits? How many OWA connections a single E 2010 CAS server can support up to? In the scenario the E 2010 CAS Server is running on top of Windows 2008 R2? The issue we will talk about is not the application limitations but remember there is no application called “limitless” there will be a point when things won’t work the way they should be. (Response time, Client end user experience etc.)

Plan your server provisioning steps and document is incase “bad” happens, this way you will know how long it will take to bring another CAS server into existing CAS ARRAY , remember once the infrastructure is set correctly adding CAS servers into existing CAS pool behind the HLB is simple work and it should be seamless to your end-users. Plan your Change control process, procurement etc. (this is why Virtual infrastructure makes a lot of sense, easy and fast server deployment.)

Each Client connection is made up of source IP Address, Source Port & Destination IP and destination port (The TCP/IP basic) as long as CAS Server does have additional IP Address , the CAS server can scale 60K outbound connections per source IP address

A CAS server is not limited to 60000 TCP connections. It is limited to 60000 unique combinations of source IP, source port, destination IP, and destination port for each IP defined on the CAS server.

This means that a CAS server with a single IP address can support more than 60000 TCP connections at the least. Remember there will be CPU and Memory issues when so many connections start hitting the single CAS Server, and this is exactly why you need to plan your infrastructure and make sure in a bad scenario the services are going to be provided to end users.

 

Resources

Calculator

http://blogs.technet.com/b/exchange/archive/2009/11/09/3408737.aspx

Exchange Scale Limitations Spreadsheet

http://gallery.technet.microsoft.com/exchange/Exchange-Scale-Limitations-34306f77

Respectfully,
Oz Casey, Dedeal ( MVP north America)
MCITP (EMA), MCITP (SA)
MCSE 2003, M+, S+, MCDST
Security+, Project +, Server +
http://smtp25.blogspot.com/ (Blog)
http://telnet25.wordpress.com/ (Blog)

Exchange 2010 Virtualization and Combining Server Roles ?

 

One of the most frequent asked question around design and deployment scenarios , should business go for virtualization when it comes to deploying Exchange 2010 and newest versions. This has almost be the hot topic to be honest and I wanted to highlight some of the bullet point to help for those who need to give the similar decisions on your deployment project.

Before we deep dive into Virtualization we need to be clear about combined roles in Exchange 2010 deployment and talk about it how much it make sense ?

Another significant difference in between Exchange 2010 and 2013 is the simplicity, and doing things in easy and more efficient way. The server roles reduction in Exchange 2013 is one of the example, so let's take a quick look server roles so we do understand and give better decision on combining server roles.

 

Reduced Server Roles

I am personally favoring to go for Virtualization when it is possible and makes sense for the business and here are some of the compelling reasons why I think the Virtualization is the way to go.

The Exchange server application is more and more focusing about the idea of simplicity, getting things done in simple way while providing better services and of course this statement will expand from server perspective to end user experience.

Virtualization Benefits ( Exchange 2010 and 2013)

• Reduced server hardware costs
• Power and space saving
• Improved server utilization
• Fast Server provisioning
• Under-utilized hardware, and can reduce hardware and maintenance costs
• Virtualization provides organizations additional choice and deployment flexibility to meet business requirements and lower IT costs and complexity
• Combining Roles does it make sense?

The answer IMO to this is real short and yes, if you ever design any environment you will quickly see the reduction on the server numbers once they are combined, Remember the most reason ratio for CAS and MBX server, one to one. So if you have 10MBX Server you will need 10 CAS Servers and most probably you will combine CAS & HTS Servers , this is what most people have been doing since Exchange 2007.

Now imagine if you combine HTS, CAS and MBX roles the server count in the same example reduces 50 percent and this is why I am all for combining roles which makes the most sense.

Microsoft Official Stand for Virtualization

Does Microsoft support combining roles and Virtualizing Exchange 2010 servers ? The answer is yes , see below official support statement from MSExchange Team.

http://blogs.technet.com/b/exchange/archive/2011/05/16/announcing-enhanced-hardware-virtualization-support-for-exchange-2010.aspx

• Hypervisor-based clustering, high availability, or migration solutions that will move or automatically failover mailbox servers that are members of a DAG between clustered root servers, is now supported.
• The Unified Messaging server role is supported in a virtualized environment.
• Combining Exchange 2010 high availability solutions (database availability groups (DAGs)) with hypervisor-based clustering, high availability, or migration solutions that will move or automatically failover mailbox servers that are members of a DAG between clustered root servers, is now supported.

When Exchange Server roles are combined there will be additional administrative overhead managing Exchange Servers and also few things to take into considerations

• HUB transport Server and mail submission Service default behavior
• DAG implementation, FSW Consideration in DAG
• Multi Role deployment Server patching overhead and recommendations

Here is more comprehensive Information.

• Robert’s Rules of Exchange: Multi-Role Servers
• Patching Multirole Servers

Conclusion: As you can see from this short post , there are good enough reasons, why to start with multi role deployment and take advantage of Virtualization technologies when deploying Exchange 2010 and 2013 versions.Remember the more simple design is the “best” design in most of the cases. Exchange 2013 has many futures simplifies such deployment and brings on a table huge IOPS reduction.

Respectfully,
Oz Casey, Dedeal ( MVP north America)
MCITP (EMA), MCITP (SA)
MCSE 2003, M+, S+, MCDST
Security+, Project +, Server +
http://smtp25.blogspot.com/ (Blog)
http://telnet25.wordpress.com/ (Blog)

Exchange Administration Center ( EAC )

EAC is web-based management console in Exchange 2013 which is replaced EMC in Exchange 2010.  The MMC base console used to be slow when it loads and new business logic in Exchange 2013 forced to have new EAC in my opinion. Most of us,  most often appreciate new futures and try to adapt them as quick as possible. When certain things are done in certain way over years and people get used to it, it could be bit disappointment to some of you out there to see brad new management interface.

So my recommendation is, forget about GUI if you are going to stay in Exchange Server business (-: , learn and use PowerShell and who knows next version we might not even get EAC (-: , pure PS to manage next generation of Exchange Server.

To get to EAC this is what you have to insert into your browser ( from your CAS Server or replace the local host with CAS Server FQDN or IP Address)

https://localhost/ecp

Now this will give you EAC

If you are in Co-existence with Exchange 2010 , same URL wont work for you , so need to use this one

ECP= Exchange 2013

https://localhost/ecp/?ExchClientVer=15

ECP= Exchange 2010

https://localhost/ecp/?ExchClientVer=14

Now if you miss old console and keep cursing things as you get used to EMC here is a nice trick might be able to lessen your pain (-: , yehh I hear you….

Create custom MMC , see the video , I have also added ADUC and ADSIEDIT into my console. I wish at the least MMC console was bit more flexible to do more with it.

http://technet.microsoft.com/en-us/library/jj150562.aspx

Respectfully,
Oz Casey, Dedeal ( MVP north America)
MCITP (EMA), MCITP (SA)
MCSE 2003, M+, S+, MCDST
Security+, Project +, Server +
http://smtp25.blogspot.com/ (Blog)
http://telnet25.wordpress.com/ (Blog)

Exchange 2013 Surviving Server Roles and Business Logic

In exchange 2010 we do have 5 Server roles and the business logic build in middle tier, which means MAPI request goes to HLB ( if exists ), from there is goes to CAS Servers in Exchange 2010 Architecture, then the Exchange 2010 CAS start talking to Exchange 2010 MBX Server. The process in Exchange 2013 has changed, the business logic got moved into MBX Server role, while Exchange 2013 supports RPC/HTTP only dropping support for RPC/TCP. This means MAPI client will use RPC packets wrapped inside the HTTP to talk to their respected MBX Server.

Some of the compelling futures build into Exchange 2013……

• Exchange 2013 is much simpler to deploy , less servers roles, simple name space and better site resilience support.
• Exchange 2013 introduces huge cost saving by offering faster, and smarter messaging application.
• Exchange 2013 does have built in "application awareness" , new Exchange 2013 Application can detect and fix most of the problems automatically while the service continuity is not interrupted, ( much better seamless end user experience)

 

IOPS Reduction !!!!

This translate better performance obviously and cost saving, Faster the application is, better disk utilization and usage will be, allowing low end disk to be used and utilized by Exchange 2013.

Comparing exchange 2013 IOPS with some of the previous versions of Exchange Server

Exchange 2003---------------------------->95.5% IOPS reduction ( compare to Exchange 2013 )

Exchange 2010----------------------------->50-70% reduction in IOPS ( compare to Exchange 2013 )

 

CAS	Client Access  ( Available Role in Exchange 2013 )
HTS	Hub Transport ( Combined into CAS Role )
MBX	Mailbox    ( Available Role in Exchange 2013 )
UM	Unified messaging ( Combined into CAS & MBX Roles)
EDGE	Edge Transport Server ( depreciated !!!! )

 

In Exchange 2013 the Server Roles do have major reduction

CAS	Client Access Server  Role
MBX	Mailbox Server Role

 

CAS Server Role

Domain joined Server ( not placed in DMZ ) existed within the ADDS name space. Provides Layer4  (transport) Load balancing

• Session affinity functionality got moved into CAS Server role.
• CAS Server Light weight protocol proxy Server
• Exchange Aware proxy Server, supports proxy and redirection

 

 

Here where you can find more information in regards to exchange 2013…

http://technet.microsoft.com/en-us/exchange/fp179701.aspx

Respectfully,
Oz Casey, Dedeal ( MVP north America)
MCITP (EMA), MCITP (SA)
MCSE 2003, M+, S+, MCDST
Security+, Project +, Server +
http://smtp25.blogspot.com/ (Blog)
http://telnet25.wordpress.com/ (Blog)