VSphere Network Questions - BW Limiting / Monitoring

[vargonen]

Hello,

I would like to learn is it possible to limit the upload and download speed of an Virtual Machine in Cluster? (protects this limits while it moves between hosts)

Is it possible to monitor the bw of the VM over cacti or other SNMP tools, and analyze the usage of network resources?

Thanks in advance,

Ali

[RParker]

You are going about this the wrong way. The problem isn't limitations (which is NOT possible anyway) on VM's. The thing you should be looking at is making more Bandwidth available, rather than limiting the VM's that seem to take up the bandwidth.

But I believe the problem is disk, you need to add more disks and better drives (SAS vs SATA), the bandwidth is NOT the problem ultimately, it's the responsiveness and spindle count of the disks you have, that's cause the problem with the VM's.

So improve your network cluster and upgrade the disks to SAS/SCSI rather than SATA or add more disks to the SAN.

[vargonen]

Hi,

We have an very powerful EMC Clariion CX4-120 with 120 drives, connected to our cluster with dual 4gbit/sec fibre links,

so you see there no problem with our spindle count or cluster disk I/O.

Let me explain you again, with an example,

You are a customer of mine and have a virtual machine in my esx cluster, and you are paying for 10mbit/sec uplink traffic

to me, and I need to limit you at this, thus you could not use my uplink more than 10 mbit/sec.

If customer wants to get more traffic, they pay and you upgrade the traffic limit, as we do on our switches,

in cisco or juniper etc.

So this is nothing related with our disk sub-system.

[mcowger]

ESX doesn't really have very strong tools for this.

Most people do this the old fashioned way - on their physical switches using private VLANs (or similar) and netflows/jflows data.

--Matt

VCP, vExpert, Unix Geek, Storage Nerd

[vargonen]

Deat Matt,

Thank you for your responce, so creating different VLAN's on the Physical Switch, and giving this vlans a speed limit is

the solution for right now.

About monitoring the Virtual Switch do you have any ideas, as it's a switch inside vmware, is it possible to

assign a SNMP to it, and look into it's ports connected to the vm's and the uplink?

Do I have to analyze the VLan's traffic from the Physical switch looking into vlan's snmp oids?

ali

[mcowger]

The vSwitch doesn't support any form of management.

You'll need to use your physical switches features for this.

--Matt

VCP, vExpert, Unix Geek, Storage Nerd

[a_p_]

Did you already take a look at the Cisco Nexus 1000V ?

André

[vargonen]

Dear Andre,

Yes I checked the Nexus, but it cost bit high, if you have a cluster of 4 hosts, each have 4 procs, you have 16 procs in hand.

Which makes 16 * 695 (per proc) = 11120 USD for the switch software.

Also you need Enterprise Plus to use this Switch, and need to upgrade your essensials or standart to Enterprise plus, which cost apprx. 40.000 USD.

Totaly you need to make an investment of 51.120 USD :smileyhappy:

[MatthewHeap]

If you only have 1 VM in a port group you can you use traffic shapping.

Even if you use VDS this would start to get very messy after 4-5 instances of this.

Regards,

Matt

[vargonen]

Dear Matthew,

Thank you for your reply, you are right, because we have about 60-100 VM's per hosts :smileyhappy:

Thus it is impossible to create a port group for each of them. on ever host....

Also thanks for your help.

ali.