Stephen Herrod on vCloud
On day 2 of VMworld in Cannes on the 25th of Feb, Stephen Herrod, the CTO and Sr VP of R&D at VMware gave his keynote speech and slidedeck.
Stephen covered three topics, here are the timings.
05:30 to 40:00 vSphere (50 slides)
40:00 to 51:00 vCloud (15 slides)
51:00 to 70:00 vClient (22 slides)
You can see by the timings which products are more developed, only 11 minutes on vCloud.
So what did Stephen have to say, here is a loose transcript, its not word for word, but close.
A world of Internal an External clouds. Using a variety of techniques to bridge these together to provide a private cloud.
The vCloud API enables 3 specific things.
- Federation. How do we connect multiple internal clouds together as well as external ones.
- Management. How do we progromatically take control of these resources no matter where they happen to be.
- Interoperability. Allowing you to move between these and not get locked into one particular implementation.
The vCloud APIs are restful interfaces which are now in private release but will be release more broadly this year.
Connecting multiple data centers together. VMotion and VMFS are the tools used internal to a cloud but a different set of tools will be used between clouds.
Four areas being focused on in the Federation front are Security, Networks, Storage and Monitoring.
Between clouds you need to be sure you are secure as you leverage different resources. How you are going to route traffic as you are brining up things in different places? How to maintain a network quality of service? How do you allow an application to get to its data wherever the compute or the storage actually is? How do you move around big virtual machines? How do you monitor this and ensure that your SLAs are being met?
The first version or example is Site Recovery Manager. Connecting at the storage side with SAN array replication and at the Virtual Center level comparing and sharing policies via the SRM run book.
Can we move things live between data centers? Long distance VMotion. Some are doing it today but it takes a skilled workforce and fairly exotic setups. We want to make this more mainstream. Will be used more between internal data centers. Working with the echo system to resolve the challenges of moving the memory of the virtual machine, moving the disk image and maintaining the VMs network connections. De-duplication and WAN acceleration are important here, extending layer 2 across sites. Not looking at global load balancing but at short or ocean distance. Planned downtime for a data center or a full data center move. If you know a hurricane is coming you can move data centers. In the desktop space if you are using a follow the sun typology moving your desktop even closer to you to have better latencies.
Simple programmatic access to different resources.
Two examples. Cloud vendors have created portals for creating VMs, spawning them off and start running/monitoring them. What we will be doing is creating a completely customizable self service portal to allow others to pick this idea up and run with it very quickly. For the external cloud providers customizable, certainly from a branding front, but every provider has a different from of authenticaion and billing, so it needs to be very flexible in how it plugs into those. We are basing this technology on lab manager which has provided self service test and development labs to customers so far. Its a strong foundation to build upon.
Internal admins, how can you look at your workloads. Bruce comes on stage to discuss the vCenter vCloud Plug-in. Lets us manage a broader range of workloads, clouds internal and external. It manages the authentication information for each of the different surfaces. You can add in another cloud by providing your credentials which can add a whole pile of capacity on which to run. Can drag an application from the internal to the external cloud. Can manage the cloud based VM doing such activities such as power on. This plug will be seen this year.
Having compatible cloud providers. Having a simple way to send machines places and powering them on. For the workload you are moving using OVF which describes the machine and some of the policies which go with them. Differentiated clouds which are more compatible than not. A broker model could come into play to broker based on some criteria.
There you have it.
The new elements since VMworld in Sept 08 is that Lab Manager will be used for some form of self service front end. As Mike Dipetrillo announced the beta application is now available for the vCloud API.
Lets keep our eye on the cloud.