cloudstack-marketing mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Toshiaki Hatano" <toshiaki.hat...@verio.net>
Subject [Talk Idea][WIP] VXLAN support
Date Mon, 29 Apr 2013 21:27:34 GMT
Hello Marketing,

 

It’s my first time to post CloudStack mailing-list.

I’d like to ask question regarding CfP for the CloudStack Collaboration Conference.

 

I’m currently working on integration between Linux native VXLAN and CloudStack as my internship
project.

I’d like to talk and discuss in Conference about VXLAN support on CloudStack.

 

Is that suitable topic for conference or not?

Please comment!

 

I already have working demo code, which simply replaces VLAN with VXLAN under cloud-agent
level in KVM hypervisor.

 

 

##) Here is the detail.

 

1.     Background

VXLAN is one of emerging technologies to overcome VLAN scalability limitation, being standardized
in IETF [1].

VXLAN enable Layer 2 tunneling over UDP/IP with VLAN-like encapsulation and allow 16M isolated
networks in the domain.

CloudStack currently have the capability to use VLAN (or GRE) to isolate customer networks
in advanced networking mode. 

But, due to the VLAN ID limitation (there’re only 4096 IDs), only 4K networks (= account)
can be in a zone using VLAN.

4K is not enough for IaaS provider like us, we already allocated much IDs in our DCs and VPC
must be going to accelerate consumption rate.

VXLAN will solve this 4K limitation.

OpenStack already started process[2] to support Linux native VXLAN. 

 

2.     Implementation

There’re three ways to use VXLAN in Linux, Open vSwitch and user land implementation and
Linux kernel (3.7 or later) native support. 

We choose Linux native support because it’s the simplest way. As always, simple is the best.

As proof of concept, I’ve modified “modifyvlan.sh” script which is actual VLAN create/delete
manipulation script called from cloud-agent, to create and to use VXLAN interface instead
of VLAN interface.

Modified script is tested with CloudStack 4.0.1 and 3 KVM hypervisors based on CentOS 6.4
+ 3.8.6 kernel.

And it looks working. (I’m still testing)

 

3.     TODOs

To receive merit from VXLAN, cloud-agent and cloud-server code should be modified to recognize
VXLAN and to allocate VXLAN IDs (VNI) from VXLAN’s large ID space.

There’re difference in parameter between L2 encapsulation (VLAN) and L3 encapsulation (VXLAN)
so it should be handled too.

(E.g. MTU, source IP address, multicast group)

I’d like to commit my code but I don’t know how to so I’d like some help to learn the
process J

 

[1] http://datatracker.ietf.org/doc/draft-mahalingam-dutt-dcops-vxlan <http://datatracker.ietf.org/doc/draft-mahalingam-dutt-dcops-vxlan/>
/ <http://datatracker.ietf.org/doc/draft-mahalingam-dutt-dcops-vxlan/> 

[2] https://review.openstack.org/#/c/26516/

 

Best Regards,

--  

Toshiaki Hatano

Technical Account Manager I

Verio, an NTT Communications company 
E-mail:   toshiaki.hatano@verio.net

AIM:     toshiaki.hatano@verio.net

Phone:  (801)437-7482  Office

        (801)960-6410  Cellular

 



This email message is intended for the use of the person to whom it has been sent, and may
contain information that is confidential or legally protected. If you are not the intended
recipient or have received this message in error, you are not authorized to copy, distribute,
or otherwise use this message or its attachments. Please notify the sender immediately by
return e-mail and permanently delete this message and any attachments. Verio Inc. makes no
warranty that this email is error or virus free.  Thank you.

Mime
  • Unnamed multipart/alternative (inline, 7-Bit, 0 bytes)
View raw message