Calculate WAN Requirements

[David_Mitchell]

Good morning all,

I have a nice easy one for somone out there i'm sure......

I have been asked to bridge and existing LAN to another site office at the other end of the country:

LAN <> WAN <LAN>

Currently the existing LAN has no external connectivity and everything is on the same subnet, servers and users.

I have no NETFLOW or network management or monitoring software as this LAN started off very small and grew over time and was thrown together by those using it and not anyone working in the network field.

All i have is the stats from the two main ESX servers hosting the VMs at the existing site showing throughput in KBps (attached).

I need to be able to calculate what the WAN bandwidth requirments are based on this information, I have had a go (shown below) but I am not sure if I am working this out correctly and I have no one I can ask in my company to verify this for me.

My working out

Server ending .100 shows avg throughput of 10,000KBps (10MBps)

10MB (10,000KB) * 8 = 80,000Kbps (8Mbps)

So on averaege this server uses 8Mbps throughout the day peaking at 40MB or 32Mbps in the early hours (which I suspect to be a backup)

Server 2 ending .101 is the same workinging out which = avg 16Mbps

8Mbps + 16Mbps + overhead of roughly 30% (8Mbps) = 32Mbps bandwidth required on the WAN link for this service

Does this sound about right to any of you, I really need some help and you guidance is greatly appreciated

Thanks in advance,

David

[mks]

Dear David,  What I understand from your scenarios that

1) you have two vmware ESX servers running, hosting numerous guest machines which are accessed over the LAN.

2) The throughput you provided is likely the mrtg statistics of the interfaces of the ESX server

3) Another LAN is coming up at other location needs to be bridged to this LAN over  a point to point WAN link.

4) You are interested in the bandwidth requirement of the WAN link between these two lans.

What is not clear :

1) Do the guest VM's itself needs to be accessed or only the service running on the these VM needs to be accessed. for example RDP sessions, X sessions vs. services like web,smtp,directory service,VoIP etc or data over smb/cifs/nfs etc..

2) whether both the ESX severs will remain at the first site or distributed between the two for some sort of DR.

3) total number of users at the second LAN and nature of services they will be accessing at first site and vice versa.

4) will the wan link will be a dedicated point to point vs shared ... 

The WAN requirement needs to be calculated based upon the type of services that needs to be accessed between the two lans, cost (bandwidth requirement) of each such service and total number of concurrent instances for each of them (in average/peak conditions).  for example 50 RDP sessions each of ???kbps + 30 VoIP sessions each of ??kbps ..... = sum of all + projected increase ......  

from lan1 to lan2 + from lan2 to lan1 . 

In case of data synchronization over smb/cifs/nfs etc. it depends upon the QoS needed ... Hope this will help.

---

mks

[David_Mitchell]

An outstanding response Mahender, thank you

You are correct in your assumptions of the scenario

What is not clear :

1) Do the guest VM's itself needs to be accessed or only the service running on the these VM needs to be accessed. for example RDP sessions, X sessions vs. services like web,smtp,directory service,VoIP etc or data over smb/cifs/nfs etc..

     Ans. The primary function would be for developer staff to connect to an application shortcut on their desktop that links to a hosted database application on the VM server at the existing site. The database is not designed for access over a WAN but we are looking at a software add on that would accommodate this via database synchronization. We are also considering the option of HTTPS web browser access to the database backend which is an alternative bolt-on product from the software provider. In terms of RDP there is some for Admins to manage servers but there wouldn’t be any constant sessions or multiple sessions running at the same time.

2) whether both the ESX severs will remain at the first site or distributed between the two for some sort of DR.

     Ans. I am basing this on the VMs and the Backup respositories remaining at the existing site, but I have to also consider the possibility of adding an additional ESX host at the new location which could spread the load for Windows Domain services (AD, DNS, DHCP ) etc and provide DR for the VMs using VMotion, so I guess VMO bandwidth requirements need to be considered also.

3) total number of users at the second LAN and nature of services they will be accessing at first site and vice versa.

    Ans. We are looking to have 10-15 staff at the new site, the current existing site has 300

4) will the wan link will be a dedicated point to point vs shared ... 

     Ans. This is the big question, we have a number of networks and I have been looking at which could be leveraged to accommodate this service.  I have identified one existing network that (if permitted) we could potentially tunnel through and share the existing WAN link which is 100Mpbs MPLS at each end.  If I can't piggy back due to the bandwidth being too much additional weigh on the existing network I am considering a dedicated link.

Thanks again your a life saver

David

[rfalconer.sffcu]

Can your vendor provide you an idea of the bandwidth utilization for the DB bolt-on that makes it work over the WAN? Mahender spells it out pretty well about figuring out traffic patterns and types then adding them together to figure out your peak usage.

As for vmotion between sites, make sure to check the requirements of that before you expect it to work. Long distance vmotion needs very low latency, lots of bandwidth, proper storage, and properly set up L2 domains on each side. You'll need some pretty advanced and expensive gear to make this work and offices that are relatively close together.

[mks]

Dear David,

As it comes out

1) their will be around 15 developer staff at the new site and will be connecting teh database over wan link using some sort of VPN between sites.

2) their could be few RDP session, for management purposes.

3) an additional windows domain services (AD/DNS/DHCP) (will likely be synchronizing with primary at the first site)

------this much is apparantly fine-------

and couple of Mbps (for example ~10 Mbps) of low latency link should be fine. unless bandwidth requirement of the database is actually too high... which you need to check with your development team/vendor as pointed by Robert.

4) DR with Vmotion. between sites.

As Robert pointed out for this to make work... you will be needing good bandwidth as well as extremely low latency apart from proper setup. However the 100 Mbps MPLS link can surely make it work provided it also have a very low latency (recommended single digit i.e. less than 10ms) and your ability to schedule it at off-peak hours to get maximum out of it. You can also think of only the database synchronization from primary to secondary rather than whole VM's, if it serves the purpose.

Are also planning high availabilty using the DR setup

i.e. if a ESX host in primary locations goes down willl all the 300 users be routed to the guest VM hosted at second location ???

Just joking.

I personally feel, you can go ahead with point 1-3 with a modest bandwidth. for point 4 you need to plan many many more things.

thanks and regards

---

mks

[Joseph W. Doherty]

Disclaimer

The   Author of this posting offers the information contained within this   posting without consideration and with the reader's understanding that   there's no implied or expressed suitability or fitness for any purpose.   Information provided is for informational purposes only and should not   be construed as rendering professional advice of any kind. Usage of  this  posting's information is solely at reader's own risk.

Liability Disclaimer

In   no event shall Author be liable for any damages whatsoever (including,   without limitation, damages for loss of use, data or profit) arising  out  of the use or inability to use the posting's information even if  Author  has been advised of the possibility of such damage.

Posting

(NB: darn, my prior reply got dropped by the system when I attempted to post it - so this is an abbreviated version.)

Simple approach: obtain the lower of two to three times your current average bandwidth utilization, or your current LAN bandwidth.

If you expect to use QoS, obtain two to three times your current average bandwidth utilization, for that traffic, and whatever bandwidth you need for your bulk traffic to allow it to complete within whatever time window is required.