VMWare ESXi 4.1.0 and Windows Server 2008 64bit Vm installations

Without understanding the coding of the website, it will be difficult to give detailed information.

But you'll need to create a CIFs (Windows Share) on your SAN, and in this windows share, "store the code for the website".

I'm sorry I cannot be more specific that this, as I need more detailed informartion, on the Web Application that's being created and developed.

As hanccocka pointted out, not sure what you mean.  If you're going to run ESX hosts you can configure another LUN and use MCSC clustering to another LUN.  

Diagram what you're trying to do, perhaps it will clarify what you're trying to accomplish.

Looks like you wanna setup MS Clusters. I will post my thoughts if you can explain your setup a bit further.

Regarding the diagram:  Attached here

The setup is to host a website that is distributed over multiple VM's that run Windows 2008 Server R2 (IIS). The distribution is done by a Barracuda load balancer. Then the VM's talk to one beefy SQL Server and there is a failover SQL Server in place.

As explained before the SAN is to be mounted on the VM's as we don't want copies of the website to have to exist on multiple disks. We want all the VM's to see the website on the SAN.

At the moment there are two physical servers to run the VM's but we may well add another two.

The specs of the servers are:

2 x DL360 G5 (Dual quad core, 16GB RAM, 5 x 73GB SAS drives)

2 x DL360 G5 (Dual quad core, 32GB RAM, 2 x 73GB (mirrored for OS) and 2 x 146GB (one for data and one for logs)


Additional Information:

    No Microsoft Clustering to perform
    Vmware ESXi 4.1.0 Used

   What servers are you purchasing?
       We have Two HP DL360 G5's with 16GB RAM, dual quad core xeons
       5 x 73GB HDD's for running the VM's on.

    Do you require a SAN (Fibre Channel or iSCSI) or NAS (CIFs/NFS)?
      Ans: iSCSI


I’ve found these two devices. The netapp is larger but I think is more suitable.
Which is best to go for considering the setup?

http://cgi.ebay.co.uk/StoreVault-S550-NetApp-iSCSI-Storage-Array-6TB-/200556555942?pt=UK_Computing_Networking_SM&hash=item2eb21a2ea6

http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&item=140440290517&ssPageName=STRK:MEWAX:IT

setup-draft.png

The NetApp FAS range are better than the StoreVaults.

I wouldn't purchase a used NetApp from ebay, from a recycing company. Purchase a NetApp from a NetApp partner or Reseller, that can provide you with the correct NetApp support. You'll need specific NetApp support to set it up correctly.

Either purchase the HP NAS or iomega NAS.

Are you going to run the VMs from  the NAS via iSCSI?

Without clustering, how is each server going to run the code from the NAS?

What technology is the website coded in?

ASP, HTML, PHP, PERL?

Without clustering in place, data cannot be simultaneously accessed by multiple servers at the same time.

@arunraju: I concure.

Here's the deal.

If you want to have a single website space then you need to cluster your VMs using MS Clustering, otherwise the barracuda does nothing for you since you will have two seperate servers with their own web directories.

VMWare clustering works in a way where if one VM crashes it restarts on another ESX host.  It does not do any load balancing as your barracuda would provide.

VMware HA is the functionality about which paulsolov has commented about.

1. Website is coded in the following technologies:
      .net (C#) aspx pages, ajax, html etc and sql server 2008.
      One of the vm's will be a state server to handle sessions.

2. I understand the issue of buying from ebay and its better not to buy from it.

    If the following devices are used, I am assuming that we don’t need a separate head/controller for either of these. Am i Correct?

    http://cgi.ebay.co.uk/StoreVault-S550-NetApp-iSCSI-Storage-Array-6TB-/200556555942?pt=UK_Computing_Networking_SM&hash=item2eb21a2ea6

    http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&item=140440290517&ssPageName=STRK:MEWAX:IT

3. Finally, can you guys summarize the option go for?

You should look at VMware FT (Fault Tolerance). Will require an upgraded vmware license, but it basically keeps two servers in lockstep with each other, if one stops responding to requests the other machine will take over at the point the previous server quit working. Its basically one step up from VMWare HA.
Also as far as a SAN goes take a look at Scale Computing. A little pricier, but it is on the vmware compatibility list.
http://scalecomputing.com/products/the_n05/

both those devices can be used directly, but please see my warning about purchasing from eBay, espec ially for the NetApp device.

You will need to store the VMs on the NAS.

You will need to use MS Clustering or another technology to provide availability, scalability and resilience.

Please comment on the following:

Could you explain why we need to store the VM’s on the SAN? We don’t need any kind of high availability that VMware/ESXi provides when storing the VMs on a SAN.


The plan is to have 4-8 VM’s all accessing the same database, and the Barracuda to load balance between these 4-8 VM’s. If a ESXi host dies, we’re still running half the amount of VM’s, so that’s enough HA for us at the moment.

Is there not a simpler way for the VM’s to access the database simultaneously? What would be the drawback using, say, NFS instead of iSCSI?

What are the benefits for us to use a SAN to store the database? Is a SQL database required to be mounted locally (iSCSI) to function


Secondly:
Let's just establish something here and that is the VM's need to access the SQL server via odbc or ole db. This access has nothing to do with the san and sql server will be fine with this.

Its the web app codebase that will be housed on the san that all the vm's need to access at the same time. So aspx, html, css, js, jpg and png files.

So no database on the SAN!

If we do not need block-level access, I.e. files only, then we don’t need iSCSI, and thus, no clustering and also, no expensive SAN, I guess. All we need is an NFS share from a good NAS, which then can be accessed simultaneously by all the VM’s.

Experts. What do you comment on this.

Okay, so this is "simple" then:-

Store all the 4-8 VMs on local disks on the two ESXi servers.

4-8 VMs all access the SQL database via ODBC.

But we've already stated that for the servers to access the codebase simultaneously on the SAN, MS Clustering must be used.

otherwise try it and see, create a CIFs share on the SAN you are going to purchase and put all the codebase on the SAN in the CIFs share, web app codebase that will be housed on the san that all the vm's need to access at the same time. So aspx, html, css, js, jpg and png files.

So just get on and purchase your san, but for test purposes, you can create all the VMs now on the ESXi servers, connect to the SQL DB via ODBC, and use a test CIFs share on a server to test the scenario.

In that case no issues.  You'll have a SQL server available for the VMs to have access and use the Barracuda for load balancing.  In that case VMs can be on local storage.

Now for NFS share you can create Linux box or purchase a NAS for this.  Question becomes how reliable do you want it to be. Netapp NFS is great but it is an add-on to the base package.  Another issue is how are you going to backup the NFS share.

If your SQL Server is a standalone VM or physical machine this will be your single point of failure.  As long as you know this you're ok

@hancocka: regarding your end of the last comment, you mentioned use a test CIFs share on a server to test the scenario, where do you want me to create the CIFs share? which server you mean.


@paulsolov: if load balancer barraduda is used, then MS Clustering needed?
       My first comment above states "The setup is to host a website that is distributed over multiple VM's that run Windows 2008 Server R2 (IIS). The distribution is done by a Barracuda load balancer. Then the VM's talk to one beefy SQL Server and there is a failover SQL Server in place." so there is no single point of failure for SQL Server.


Experts: how to deal with the backup of SAN (web app codebase)?

To test your project, you don't need a SAN, you can create a CIFS (windows share) on any Windows Operating System to test if this will work, all the servers accessing the CIFS share.

What my fellow expert paulsolov was stating is that the Single SQL server is the single point of failure in this implementation. If the SQL server fails, it doesn't matter how many servers, load balancers you have, it will fail. You mention a failover SQL Server in place?

How does the failover occur? via Microsoft Clustering?

Backup of SAN - Tape attached to SAN (via NDMP), Symantec Backup Exec, Robocopy (file copy), SAN Snapshot

In the diagram above:
there are 2 physical servers ( windows server 2008 64 bit OS and SQL Server on it)
If one physical server fails, i want to use the second server.
How to achieve this? i mean this as SQL Server failover.

Good to know this, Many thanks for your suggestion.

Let's nail this down:

Barracuda Load Balancing/Failover: Web Servers/Applications Servers (Front end)

MS Clustering:  Failover for SQL Server (back end)

how long it takes to complete this setup and how much people charge normally to set this up? I mean installation cost and support cost?

@paulsolov: can u explain in depth please.

Please contact us via our profile, and we can talk in the new Year.

I can arrange for one our Engineers to visit and discuss Consultancy.

I'm not prepared to give the details here in this open forum, this needs an initial meeting and discussion of the Project, Project Plans, Solution, Scope of Works. Please checkout our Profile and please feel free to contact us to discuss.

The scope/breadth of this is too wide as hanccocka stated.  If you're in the states I also perform consulting services, you can contact me,  my email address is on my profile.

Experts:  Many thanks for all your suggestions esp. hanccocka and paulsolov