Hi Linda,
You give good comments to the draft. But I have some different opinions on VM
migrations.
For DCN, VM migration is a very important feature. When a server breaks down,
the applications or VMs on the server can move to other physical servers
dynamically.
and in a mega data center, server breaking down is very common. VM migration
need transfer large number of information, it will conduct a sharp traffic
requirement
and put great stress on bandwidth.
On the other hand, for load balancing or energy saving, some papers have
proposed methods to rearrange the applications by VM migration in the layer-2
domain. If those methods are used in DCN, VM migration will come to more
frequent than today.
So in my thinking, even if VM migration is not very frequent, it must be often
in future data center. It will be considered at the beginning of design and
deployment of
a data center. The architecture of DCN and its bisection bandwidth do affect
the efficiency of VM migration.
By the way, MatrixDCN is a new architecture for DCN like Fat-Tree. It is easy
deployed and managed.
- Section 3.1 (The Evolution Problems of the Logical Network Topology in VMMI
environments):
I don’t think the following statement is valid: “Since a large number of VMs
and their applications are running in the same Layer-2 domain, it (VM
migration) may be very stressful from bandwidth utilization viewpont of the
data center switching network.”
Applications communicate with each other via IP addresses. Yes, some instances
are placed in the same subnet. VM migration is not the main reason for
bandwidth utilization. There are many reasons for (poor) bandwidth utilization.
“In order to improve the bandwidth utilization, it is required to upgrade the
load balancing capability of the network which has numerous ECMP between
different points.”
This “load balance” has nothing to do with VM mobility.
I don’t understand how “Matrix DCN” have anything to do with VM mobility
either.
ytsun
_______________________________________________
nvo3 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nvo3
