Exchange solution with redundancy

Oh man! Ad-blocking software has been detected! :'(

This website is run by the community, for the community... and it needs advertisements in order to keep running. Blocking our ads means your killing our stats!
Please disable your ad-block, or become a premium member to hide all advertisements and this notice.

Hi all,
We are going down the ISO route and as such I need to improve the Exchange set up we are using.

Currently I have 1 Exchange 2003 server that serves about 70 users across 2 sites, but I need a solution that will provide failover/redundancy in the event that one site goes down from a power cut or fire etc. As part of the process I would be upgrading to EX2010.

The 2 sites are connected via a 100 meg link, and there is a DC at either site. I was thinking along the lines of 2 Exchange servers ( one at each site), but I'm not sure about how to handle the mail stores, so that if one site is down, the other takes up the slack with little or no impact to the users.

I've looked at hosted Exchange services - which ticks a lot of the boxes regarding ISO - but would like another solution(s) which allows me to keep the Exchange servers internal.

Any suggestions?

Cheers
NB

 

Oh man! Ad-blocking software has been detected! :'(

This website is run by the community, for the community... and it needs advertisements in order to keep running. Blocking our ads means your killing our stats!
Please disable your ad-block, or become a premium member to hide all advertisements and this notice.

Exchange 2010 DAG mate, make sure you have MX records set up so that if Site A looses internet the incoming mail will go to Site B.

If the Exchange 2010 dies in Site A then AD Sites and Services will automatically log them onto the nearest subnet, being Site B and vice versa.

 

Cheers mate.

So am I right in thinking that Exchange server 1 would hold a copy of the mailstore and so would server 2, with the changes being replicated between them in near real-time?

NB

 

Yep, pretty much.

 

Yeah I've just been flicking through that site. It seems high availability is a lot better in 2010.

NB

 

Depending on the ISO you are going for you will want a lagged copy of the DB too
so that will provide protection for roll backs if say, a trojan spams thousands of mails across your stores etc

DAG can be very simple in 2010, but plan your CAS deployment properly or the failover will break anything external, generally a load balance that can handle global traffic redirection could assist with this if you want to go that far

 

If you want to stick with Exchange Server 2003 you can use a third party software I played around with is CA ARCserve High Availability and Replication I download the trial version which is very easy to use for a failover but not cheap.

I also like note to Note_Blokes replicated to exchange server does not mean you can do a failover as users mailbox are either created in Exch server or Exch server 2 so if the user mailbox is created in Exch server 1 then outlook will have to configured to connect to exch 1 or even using pop or imap accounts as well. So if it failed there will no failover to exchange server 2 unless you used etc ARCServe HA & Replication software or the cluster administrator.

 

Yes I understand Sparky, it just as I stated that if want to stick with Exchange server 2003 but other than that I understand Exchange server 2007/2010 has the high availability and replication feature included which he better off upgrading to am I right as I have not used Exchange server 2007/2010?

 

Yes I've got the budget to upgrade to Exchange 2010, so the solution would use 2010 instead of the current 2003.

Cheers
NB

 

As was mentioned, a DAG spanning both sites, either in an active/active configuration, or an active/passive.
Advantages and disadvantages are included with both solutions.

A 100mb link should suffice to replicate several thousand users worth of transaction logs between the two locations.

Building two multi-role servers with the CAS, Hub Transport and Mailbox roles all-in-one will keep cost down if you're going with physical servers, but you'll need a Hardware Load Balancer (can get a cheaper one in the $2000 range) due to the CAS roles. If you're going with a virtual setup, you could split the CAS and Hub Transport roles from the mailbox role to make a second set of servers that use Windows NLB if the HLB is not a viable option. I wouldn't recommend using DNS round-robin due a number of reasons (especially if you want to provide OWA, EAS or Outlook Anywhere for your users).

With a two node DAG, you'll want to ditch JBOD and go with RAID, so make sure you get servers with enough disk slots (if going physical).

How many people are we talking about here and do you know their messaging habits (i.e. how many messages they send per day, and what is the average size of these messages)?

 

Yes the servers will have RAID.

There are 70 users and the max attachment size will be capped at 3 meg. The heaviest users send about 20 - 25 messages a day, but obviously not all contain attachments.

NB

 

You could run 70 users off a Netbook if you shoved in enough ram

Having done a very quick Exchange calculator check, this is what I saw:

CPU wise, you can run a multi-role server with a single core. I believe any 64bit capable server these days will come with ~4 cores anyway, so you can go for a cheaper CPU (you'll still have ten times the number of available megacycles there for you).

The design would warrant 8 Gigs of ram due to it being a multi-role server, but ram is relatively cheap these days. Even 4 Gigs would run it, but memory is used up by your antivirus product, monitoring software, etc... so can't be too skimpy with that.

The number of IOPS will definitely be very low, you can pretty much get the cheapest disks you can find and put them in RAID1 (mirrored) to save cash (no advantage in going with RAID5). You could run everything off the disks used for the OS But there's some advantage in keeping the OS, Exchange binaries and Hub Transport queues on the OS RAID1 setup, and the database/logs on a second RAID1 setup dedicated to just them. Don't forget the Recovery LUN... You can get big, slow disks for the database if you want to give your users massive mailboxes (+10GB in size) and enable recovery of deleted items for a period of multiple months.

If your users send 50 messages a day (the smallest number available in the calculator) and they were always maxed out at 3MB in size, you'd still be using less than 10'% of your WAN link between the two locations. You said that they send between 20 and 25, and as message size is often closer to a few hundred k in size, you'll be using only in the range of 1% of your bandwidth between the two sites for DAG replication.

Basically, you don't have to get very powerful hardware to run this setup, just pay attention to having enough ram (the calculator goes in typical increments which is why it starts off at 8 Gigs for a multi-role setup).

And finally, two NICs is good to have so that you can separate replication traffic from everything else. But not that important in the smallest setups.

 

Thanks for this mate, its excellent info

NB

 

FYI, whilst I did say that your design could conceivably run off a single CPU, stick to Microsoft's Recommended Practices as per: http://technet.microsoft.com/en-us/library/dd298121.aspx#PR

i.e. two cores is the recommended minimum for a multi-role server.

Moot point if you go physical as I think you'll be hard placed to find server hardware with a single core anyway as I mentioned before...