What would you recommend - server refresh

[nutcase]

We know we need to change how our servers are put together, but we're not sure what route to go. Currently we have this:

Box1: DC; DHCP; DNS; Exchange; File server; Content filtering; Backup Exec (DL380 G7)- Server 2008R2

Box2: RDP (DL380 G5) – Server 2008R2

Box3: Hyper-V host running VM1: DC; DNS; Legacy Database (DL380 G5) Server 2003

Box4: RDP (DL380 G8) – Server 2008 R2

NAS: Backup repository; Archived file access (clearing space from Box1) (WD Sentinel) – Server 2008 R2 fileserver

Main Site: ~60 seats Server rack fibre MPLS access

Secondary site: ~20 seats Server rack (not currently in use) fibre MPLS access

Tertiary sites (4: ~20 seats per site EFM MPLS access

Internet breakout in London Datacentre through Sonicwall CGSS

Cost effective recommendations for virtualising all servers and retiring Boxes 2 and 3. Ideally using Secondary site as a DR site – either load balancing or failover (connection between the two sites is currently fibre 10MBPS but can be increased in 10MBPS intervals to 100 MBPS if required (at a cost).

Our new IT consultants have recommended using the G7 as a Vcenter; re-purposing the G8 as VM Host and adding a second G8 as another VM Host. Linking the two hosts would be a P2000 G3 MSA SAN with dual fibre channel switches.

The hardware alone works out to be very expensive for their option – is there a cheaper alternative that still gives good availability?

Any thoughts?

thanks

 

[#Chri5#]

Quick post...

For a simple 2-node vSphere setup, you could ditch the Fibre Channel and go for the SAS MSA P2000 with dual HBAs in each VM host. Like this.

There's also a newer MSA, the MSA 2040 which I presume will replace the P2000 G3 at some point.

 

[Yamahahahahaha]

I would say it's a good time to do a few things;

1) Figure out which services need Fault Tolerence (usually email, file servers) and which need High Availability (Non-critical business applications) and which just need to be recoverable (Non-critical non-BAU applications).

2) Put a price tag on downtime for each of these services. If you have had any failures in the past few years, what was the average time to recovery? Quantifying how much emails are worth depends how much staff rely on emails to do their jobs.

3) Look at cloud products like Mimecast to potentially offer a hybrid email solution which offers extremely high availability while also being relatively cheap (compared to Microsoft licensing).

4) Let the costs and applications dictate the appliances. I don't know how much storage you need, or how many emails are going through your company, or what is critical to your business. What your consultants have suggested is a fairly standard N+1 VM setup which is perfectly adequate for most SMBs.

I always look at the Dell and IBM equivalents of HP servers (and vice versa for all). The Dell R720s are very nice boxes and are very capable and versatile machines. Also the HP G9 stuff is coming out soon, you may want to hold on for that. Also being colocated you want things which are compact and energy efficient, so that is a consideration. I remember one client had to rent an entire extra rack to get enough power for their 16U of beastly kit...

Sorry for vagueness but it's difficult without specifics!

 

[nutcase]

Quick post...

For a simple 2-node vSphere setup, you could ditch the Fibre Channel and go for the SAS MSA P2000 with dual HBAs in each VM host. Like this.

There's also a newer MSA, the MSA 2040 which I presume will replace the P2000 G3 at some point.

Thanks for the link - I'll have a proper read later but it's good to know even if we do go with the SAN route there's a lower level we can go in at

I would say it's a good time to do a few things;

1) Figure out which services need Fault Tolerence (usually email, file servers) and which need High Availability (Non-critical business applications) and which just need to be recoverable (Non-critical non-BAU applications).

2) Put a price tag on downtime for each of these services. If you have had any failures in the past few years, what was the average time to recovery? Quantifying how much emails are worth depends how much staff rely on emails to do their jobs.

3) Look at cloud products like Mimecast to potentially offer a hybrid email solution which offers extremely high availability while also being relatively cheap (compared to Microsoft licensing).

4) Let the costs and applications dictate the appliances. I don't know how much storage you need, or how many emails are going through your company, or what is critical to your business. What your consultants have suggested is a fairly standard N+1 VM setup which is perfectly adequate for most SMBs.

I always look at the Dell and IBM equivalents of HP servers (and vice versa for all). The Dell R720s are very nice boxes and are very capable and versatile machines. Also the HP G9 stuff is coming out soon, you may want to hold on for that. Also being colocated you want things which are compact and energy efficient, so that is a consideration. I remember one client had to rent an entire extra rack to get enough power for their 16U of beastly kit...

Sorry for vagueness but it's difficult without specifics!

1 & 2) A few weeks back just after the consultants gave us their all singing all dancing proposal I thought I ought to do my own - first thing I did was to do a spreadsheet along these lines for the company directors to fill in their thoughts on how long we could afford to have various services unavailable. We have been a bit unlucky recently with short outages. Culminating (hopefully not due something worse!) both terminal server boxes going down last weekend leaving us without office applications for the other sites for the whole of Monday while I went home for my microserver and built a temporary one (that coped amazingly well considering!). I've chased the response to the spreadsheet as so far nothing. Hopefully this will be a wake up call...

3) Cloud is something I've looked at and will be an option for my proposal - but historically as a company we like to own kit and breaking that is VERY hard. Additionally some clients have very strict requirements for their data which makes hosting it internally easier to accomplish. Not impossible to overcome but not straightforward to persuade either Probably doesn't help that I'm not a big cloud fan either!

4) We've reached the low terabyte storage requirements so not huge - but of course it is increasing so need capacity for expansion. I want to get the NAS back to being just for holding backups for one thing so we do need to invest in drives for whatever solution we go for. Fortunately we're not huge email users but we have to retain a lot so archives are quite big. Compact is less of an issue - for a small company we have 2 42U racks (and a half height one in another building at head office!) but energy efficiency is of course an issue.

Ref HP vs A.N.Other - We do have 2 Dell servers running a database (separately maintained by a 3rd party so outside the scope of this refresh) - one's a fairly basic R410 that sits as a backup in the other building, and the main one is a highly specced R420. Perfectly decent bits of kit - no need for them to be 1U but that's just what was supplied. Touch-wood we've had no hardware issues beyond the odd failed drive and controller battery, so the only thing we've got to go on with preference is build quality and ease of use. HP wins that one for us but not so much that for a decent saving Dell would certainly be considered.

Perfectly understandable you can only give a bit of a guide rather than full on advice - but it's very much appreciated

 

[Caged]

The Dell VRTX sounds pretty much perfect for that size environment.

 

[nutcase]

That looks a bit restrictive for flexibility? We've got no shortage of rack space so don't need to go for that crazy contraption

 

[Quartz]

Make sure you have at least one domain controller that is not virtualised - two by preference.

 

[nutcase]

Not really discussed how the hardware would be used, but I'd assume that the vcentre box would also be a DC. Second one possibly would be a VM (as currently) but again it's not something discussed as yet.