For more details, please see ourCookie Policy.

Data Center

B-agile: Disaster Avoidance & Mobility in the Cloud

by Patric.Chang on ‎08-30-2012 11:22 AM (4,222 Views)

                                                                                                                                  Photograph by Don Vu

In the drive to get to a better cloud, administrators and managers of the next generation data center need the ability of mobility.

Pat Gelsinger and VMware is charging down the road to fuel the construction of next generation or software defined data centers by abstracting, pooling and automation of resources. If this is translated into an industry trend, this will bring about massive consolidation resulting in much larger data centers capable of running hundreds of thousands of virtual machines or more for millions of users.

This scale of consolidation has three significant implications. First, the data center cannot experience any downtime. Second, the infrastructure supporting the applications has to be ultra-efficient. Third, the massive pool of resources within the data center has to be optimally utilized even outside business hours and by users located outside the data center’s vicinity.

The means to address these implications were covered by two conference sessions this week by Jonathan Hudson on “Disaster Avoidance for Mission Critical Applications using Metro vMotion” and Chip Copper on “Building the Next Generation Data Center for the Cloud”.

Disasters happen more often than people think. According to the Centre for Research on the Epidemiology of Disasters, 79% of natural disasters were caused by the weather with power and other outages due to hurricanes, snow storms, heavy rain and other perils. An unhappily example was this week’s Gulf Coast hurricane. In addition, 70% to 90% of downtime is planned. But in both the natural disasters and the planned downtime, administrators usually have at least hours of notice to take action. Therefore, with prior preparation, downtime can be avoided using available technologies and solutions available today. How?

To support continuous uptime, the data center architecture needs to have the following:

  • Data synchronization between data centers, allowing applications, no matter where they are running, read and write access to their data at all times.
  • Network infrastructure that provides high performance, highly reliability, and layer 2 extension capabilities to interconnect the data centers.
  • The ability to redirect the client session to wherever the required application might be running.

Or in Jonathan Hudson’s words: “We need to take two data centers and make them look like one.” An example solution is the following Disaster Avoidance solution, built using VMware Metro vMotion and EMC VPLEX. Most data centers already have at least some of the infrastructure (e.g. the Storage networking and extension) already in place to implement the solution. The solution allows applications and VMs to be moved using metro vMotion for up to a 10ms latency round trip between the data centers.

Brocade is also working with other partners to implement other disaster avoidance solutions. Another solution, for a 5ms latency round trip, is available with IBM SVC built with the components illustrated in this diagram.

Another issue in the data center today is that only 50-60% of the data center applications are virtualized. In order to get the full benefits, there is a need to virtualize the business critical applications such as SAP, Oracle, Microsoft Exchange and SharePoint. Running these applications on the above data center architecture (which we also call the Active-Active Data Center) increases the ability to provide the required high availability.  An example solution is the EMC solution with Brocade networking that supports running SAP on vSphere with a stretched Oracle RAC database across two data centers.

As we embark on massive aggregation of resources in a data center, it is mandatory to have a highly efficient infrastructure network. The Brocade VCS Ethernet fabric is an intelligence choice for building a network infrastructure can be operated, managed and extended effectively. The Brocade active-active data center architecture provides the underlying infrastructure to allow applications and VMs to be freed for mobility with the following features:

  • VPLS provides robust, scalable, standards based Layer 2 IP Virtual Private Network.
  • Brocade ADX enables non-disruptive Live vMotion using vCenter integration to keep clients connected to the applications
  • FCIP with Fast Write reduces response time by 50% or increases distance by 2X for VPLEX Metro by reducing latency.

Moving ahead, there is another option to implement vMotion over metro distance using VXLAN on existing networks by using the ADX as a gateway as discussed in this blog.

Should you b-agile? Well why not? If Pat Gelsinger is right, we are in for a long haul of massive consolidation. In your journey of virtualization, you have already stockpiled much of the technology required to implement a disaster avoidance solution for your company. By augmenting what you already have, you can become more agile as you free your applications and VMs for mobility in the cloud. By taking the action now, you will also gain the potential to increase your ROI with additional benefits such as balancing workloads across data centers, improving your service levels as well as the ability to bridge between private and public clouds.

on ‎09-03-2012 09:34 AM

Hi Patric,

This is a nice overview of the methods to achieve data center agility. Here is a link to a design guide for a dual active/active data center using Brocade networking with VMware SRM and EMC VPLEX

It's one of the publications at the Strategic Solution Lab forum where folks can find reference architectures, design guides, deployment guides and validation tests so you can B-Prepared :-)



on ‎02-28-2013 10:53 AM

Thanks Brook for the pointer. This is a really useful document for a SRM solution to implement an Active-Active data center. I wonder if there are cloud service providers who would implement a service using this design to allow customers without a second data center to have an active-active capability with the remote DC in the cloud.