To add to "concern 2", you can use pgbench to compare between bare metal
and VM, it's the benchmarking tool from PostgreSQL DB which should be
very handy in your situation.
---
Nux
www.nux.ro
On 2022-11-21 05:54, Vivek Kumar wrote:
Hello Bryan,
For concern 1-
There are couple of ways to achieve this.
1- You can use affinity groups ( host anti affinity ) where it will
make sure that you applications will be powered on separate hosts.
2- You can use HOST TAG to define your hosts in compute offering. If
you want your application to be powered on any specific host.
3- You can create 2 cluster and isolated them by using the TAG either
host and storage tag.
But if you want to use host_tag then I suggest to use at least 2-2
host for same tag, so that VMs can be powered on if any host goes
down, because cloudstack will specifically see the tagged host so if
the tagged host is not available then you won’t be able to power on
you workload.
For Concern 2:
For this, I guess you will have to deploy some testing workload on VM
and compare the performance with the baremetal, You can use tools like
IOmeter, FIO for testing the random and sequential read/write and
throughput. Because Databases required good I/O so underlying storage
would be the critical component.
I hope this helps..!
Vivek Kumar
Sr. Manager - Cloud & DevOps
TechOps | Indiqus Technologies
vivek.ku...@indiqus.com <mailto:vivek.ku...@indiqus.com>
www.indiqus.com <https://www.indiqus.com/>
On 20-Nov-2022, at 7:17 PM, Bryan Tiang <bryantian...@hotmail.com>
wrote:
Hi Guys,
Im a Business Development Manager trying to convince our company to
convert our datacenter into a Private Cloud using Cloudstack.
Currently things are set up in a Baremetal sort of set up. Im not
super technical, more so on a business and risk side of things. And to
have this approved by our management, I will need to address some of
their concerns.
Concern 1:
Some of our applications requires High Availability. On a software
level, we have covered this. But on a hardware level, is there a way
to deploy our applications in a way where each HA Service is deployed
in a different physical server (or cluster). This is to avoid any one
server taking down the entire service (assuming both HA Applications
are deployed as separate VMs, but in the same physical server).
Concern 2:
Is there any benchmarks only that compares the performance of bare
metal and VM? I understand there is a slight drop in performance, but
how much is that? Some of the system we currently deploy on bare metal
are MSSQL Databases and other services that have very high network IO.
Are any benchmarks available online? And can those performance
degredations be restored/matched by compensating with more CPU/RAM
allocation to the VM?
Appreciate the help.
Regardfs,
Bryan