Cloud Insights from Brad Anderson, Corporate Vice President, Windows Server & System Center
Late last night I returned from Cisco’s announcement of their Application Centric Infrastructure (ACI) vision and offering. You can check out a quick recap video of the event at the bottom of this post.
I think what Cisco announced is a game-changing solution that brings simplified/centralized control and deep insight into the network fabric spanning physical and virtual – all in the context of the application. Think about how this contrasts with the infrastructure-centric, box-by-box approach that's prevalent for network provisioning and operations today. This functionality is going to help us, as an industry, accelerate the deployment of all the innovation we are collectively and individually driving. This is a big opportunity to rethink networking to support the cloud-driven transformation that our customers need.
If you have not been following Cisco lately, let me provide a little background.
About 18 months ago Cisco funded a spin-in / start-up called Insieme Networks. The founders of Insieme (Prem Jain, Mario Mazzola, and Luca Cafiero) are three of the most innovative leaders in the industry with an incredible track record – in fact, this is the third time Cisco has funded these three guys, and the two previous projects have been big hits in the market. I’m sure you’re all familiar with their second project, UCS. UCS is Cisco’s server platform and came to Cisco via the second spin-in of these guys (which was named Nuova Systems prior to the acquisition).
What Cisco announced yesterday was a solution that joins the physical and virtual networking infrastructure into a single view with the ability to set policy that is associated with an application. That policy can describe the network configuration required (think VLANs) by the application which will then help deploy and then migrate/move the application. The policy can also configure the network to be optimized for a specific application to do things like provide a guaranteed quality of services for specific users (e.g., the CEO) using an application.
Cisco puts it like this:
ACI is designed around an application policy model, allowing the entire data center infrastructure to better align itself with application delivery requirements and the business policies of the organization. The entire objective of ACI is to allow the data center to respond dynamically to the changing needs of applications, rather than having applications conform to constraints imposed by the infrastructure. These policies automatically adapt the infrastructure (network, security, application, compute, and storage) to the needs of the business to drive shorter application deployment cycles.
ACI is designed around an application policy model, allowing the entire data center infrastructure to better align itself with application delivery requirements and the business policies of the organization.
The entire objective of ACI is to allow the data center to respond dynamically to the changing needs of applications, rather than having applications conform to constraints imposed by the infrastructure.
These policies automatically adapt the infrastructure (network, security, application, compute, and storage) to the needs of the business to drive shorter application deployment cycles.
There is great value there, but it gets even better.
The solution from Cisco also provides incredibly deep insights into how the application is using the network. This kind of information is invaluable when troubleshooting or trying to optimize the network for the application.
To get a sense of just how valuable this is, consider a scenario like this: A customer calls his service provider and says the application he’s hosting in their cloud is running slow. How does the service provider begin to troubleshoot a problem like this? With the Application Centric Infrastructure the cloud admin can actually get a view of that specific tenant and then see a lot of other important factors like: All the other applications that tenant is hosting, where those applications are being hosted at that instant, the route of the traffic through the service providers network (physical and virtual), and the tenants applications are utilizing. Now, that cloud admin will be able to determine the precise reason within the network why the application is running slow. With this kind of insight available, the time to resolution is dramatically accelerated.
This kind of advanced tenant-level telemetry across the shared infrastructure fabric needs deep software and hardware integration, which is where I see a great opportunity for the industry to come together on. Additionally, I see a lot of applications out there that use network paths that span physical and virtual end points - even in highly virtualized environments. We have to ensure that we address such requirements without forcing ourselves into a "virtual bubble." It is great to this issue getting addressed with ACI.
From Microsoft’s perspective, here’s a bit of additional feedback on this.
First of all, I love the focus on the applications. I am an infrastructure guy and have been for most of my career – but I also know that infrastructure exists for one solitary purpose: Supporting apps. All of the innovative work we do in Windows Server and System Center is focused on enabling great apps to be built, deployed, and operated at-scale as efficiently as possible.
We are lucky at Microsoft because we get to work closely with the people building many of the most commonly used applications in the industry. By working side-by-side with these engineers, we get to know exactly what type of infrastructure and management support their apps need in order to perform at the highest possible level.
Looking back in history a bit, Windows Server really caught fire with Windows Server 2000. The rapid adoption was not driven just because it was a great operating system, but because of two killer apps: Exchange 2000 and Active Directory/Group Policy. Windows Server 2000 was ultimately a better product because the team building it interacted so closely with the Exchange and Active Directory teams. SQL has also been a tremendous partner for Windows Server, and SQL has pushed us harder than anyone else to provide high scale at a low price.
This same kind of internal learning is now driving the public cloud. Windows Server 2012 is a better product because of the incredible things we have learned from being the platform that hosts Bing, Azure, Outlook.com, Communicator and more than 200 cloud-based services. Our close collaboration with the teams building those services has made the infrastructure we ship to our customers world class.
One more note about Cisco’s Application Centric Infrastructure:
Prem Jain first called me about a year and a half ago to discuss their vision for Insieme and the kinds of customer value they planned to build. The customer value was immediately apparent, and we agreed right then to work on this together. Since that time, our teams have been working together to ensure that clouds built on the Microsoft Cloud OS were first class citizens within what Cisco was building.
I am proud to say that at yesterday’s launch event Microsoft was front and center as one of the most important partners. It is also worth noting that VMware was noticeably absent from the event.
Looking ahead, here’s what you can expect from Microsoft and Cisco:
One of the key things the Cisco ACI requires in order to be able to offer the kinds of insights into tenant use of the network is that it has to know which VMs belong to which services and to which tenants, with built-in security and isolation. We worked with Cisco to expose the interfaces in System Center Virtual Machine Manager (SCVMM) that enables Cisco ACI to understand the VM-to-tenant relationship. This helps us combine the flexibility of Microsoft's infrastructure solutions with the scale and performance delivered by Cisco ACI to support the large-scale multi-tenancy requirements that are becoming such a common part of my discussions with both enterprise and service provider customers.
One of the key components of the ACI solution are application-specific models that instruct the ACI infrastructure how to configure the network to best deliver on the application's needs. This really enables application owners to focus on their applications versus having to worry about infrastructure details. Those of you familiar with SCVMM know that we introduced service models several years ago that enable IT to describe an end-to-end application service with associated service levels – i.e. the various tiers, their relationships, the compute, storage and network needs of each tier and associated scale up/down rules. I see the model introduced by ACI as being additive to the work we started in SCVMM, and I look forward to this integration.
Similarly, I look forward to integrating common constructs around quantifying application health and surfacing them in other parts of the Microsoft System Center solution, like having a holistic view of application health across the infrastructure fabric (physical and virtual).
A policy-driven approach like this will help us jointly enable a datacenter without boundaries for our customers as these application-centric network policies travel with the application, thereby setting up frictionless application mobility within and across datacenters/clouds. As a part of this, we will also jointly deliver gateway solutions in our platforms that will support flexible and seamless connectivity at the networking layer.
As a part of our ongoing collaboration, our goal is to deliver speed and simplicity for customers while keeping a laser focus on optimizing Microsoft applications and workloads on Cisco ACI infrastructure. To do this we will be building pre-integrated Cisco ACI stacks for our workloads – just like we have been delivering services models and the automation to deploy other workloads in our clouds through the Building Clouds blog (like they did previously with Oracle and SharePoint, for example). We will be delivering ACI models for our workloads like Exchange, SharePoint, SQL and more.
This integration builds on the work that we have been doing for a number of years around Cisco UCS. UCS and the Microsoft Cloud OS are very well integrated, and Cisco has developed more than 2,500 PowerShell commandlets for UCS that enables administrators to configure and automate just about every aspect of it. The result is a pretty useful set up: When running on UCS you can use PowerShell for automation from the app, through the OS, down into all the hardware. That’s pretty cool.
Cisco and Microsoft have been key partners of all of you for many years (there are very few enterprises that don’t use technology from both companies), and the new solutions and functionality from Windows Server, System Center, UCS, and ACI bring together some incredibly forward-looking value. These offerings will enable you to really accelerate the deployment of innovations happening within your organizations – and it will support the source of so many ongoing innovations: New applications and updates to existing applications that can be deployed into your clouds!