<!--
@page
P { margin-bottom: 0.08in }
-->
I've reviewed the "Proven
Practice: Migration Methodology"
(http://viops.vmware.com/home/docs/DOC-1349) document, as well as
"P2V Decision Tree"
(http://viops.vmware.com/home/docs/DOC-1353), but I'm coming up short
on just how to go about identifying good/bad p2v candidates. The
piece of the analysis that I'm getting hung-up on is the "workload
analysis". I'm not sure exactly what the best path-forward is.
I see that the rule-of-thumb is that a machine would be a "bad
candidate" if it looks like this... >4CPU and <= 50%
Utilization, > 8GB used RAM, > 100Mbps NIC, > 100MBps Disk
IO. However, I'm certain is there's an automated tool that I'm
missing that will do some independent analysis... I expect that
there's some tool out there that will setup some sort of performance
monitoring and tell me if the machine is a good/bad candidate for
virtualization?
As far as background goes, I have a
"small" but very visible Virtualization opportunity. The
existing infrastructure supports a manufacturing environment, and
consists of 8 Windows server (7 2003, and 1 2000). Anecdotally, the
I/O profile of these VM's is mixed. The servers are generally older
(3-7 years) and some have "high" CPU utilizations of ~>50%,
2 have "substantial" disk utilization (500GB of storage on
2 of the VM's), and that 500GB is re-written weekly. The Proven
Practice methodology suggests that most of these would be reasonable
virtualization candidates, but I need to come up with a test plan.
There is no existing virtualization infrastructure within the
environment, so this would be a new VMware installation. At this
point I expect we'll be using ESX 3.5, but I'm not quite there yet
(could be ESXi). In the past, I've done a number of small p2v
projects (NT 4, 2000, and 2003) - primarily with machines that were
on their last leg, and virtualization was the only path-forward.
Since this effort is a planned effort on a working infrastructure, I
want to make sure I'm presenting the best possible analysis.
Any suggestions on the work-load
analysis piece would be appreciated.
If server are really old, also with high CPU usage they could be good candidates.
Just check the number of core or CPU used and the CPU frequency.
For example: if you have a dual P3 with 1000MHz that work at 80% you can virtualize with no problem on 2 core of a Xeon 1,6 GHz and you could have a 50% of vCPU usage (maybe a little more).
Andre
I follow the logic there... but what I'm really looking for is a tool that will analyze the existing infrastructure and make recommendations, or give a confidence level that a particular server is a candidate for virtualization. For instance, the best-practices documents suggest that it shouldn't be an issue. But I'd like to be able to run a tools, and have it indiciate with a high-level of confidence that these are good candidates... that makes the project lower risk, and easier to "sell" to the end-customer. Any ideas?
You can use the vmware consolidation setting used via Virtual Center. It is like a mini capacity planner.
As written before you can use the consolidator plugin (if you have vCenter Server).
Or ask of a Capacity Planner service.
Andre
A few options:
1) get your VM supplier to run VMWare capacity planner
2) tru VMWare guided consolidation (free with vSpehere)
3) Get 3rd party tool like PlateSpin's power Recon.