Some considerations about cloud migration

In some people's mind, the steps to move from a virtual data center to a private cloud are simple, requiring only a little bit of management software and some automation to accomplish. It's not that simple.

When IT managers embark on the path to building a private cloud, in some cases, the defined data center infrastructure may not be appropriate for cloud computing implementations. So they may have to face the test of past assumptions and practices. And vendors are openly claiming that the automation and management of cloud computing cannot be trusted anymore.

In the Enterprise data Center, private cloud provides control over IT resources. It makes workflows more automated and reduces the number of manual configuration tasks, like changing workloads, setting firewall rules, and configuring routers. It consolidates data and management on a single "location", because you can allow data center operators to change the application workload and reallocate memory, storage, and other IT resources, some people consider the cloud computing layer as an administrator.

However, for most data centers, the "need to kill" technology for private cloud automation management requires a reorganization of existing infrastructure and processes. You can't simply install cloud management software on existing servers, storage devices, and network devices as a private cloud. Your infrastructure is not working the way you expect it to.

So let's take a look at some misconceptions about the virtualization infrastructure and consider the changes required by the private cloud environment.

Virtual machine Automation is simple

A business with advanced virtualization technology now takes the next logical step: building a private cloud so that users can dial up to a virtual machine (VM) without a team creating and defining it.

When the application owner has an external vendor, they must be able to dial the connection virtual machine from the private cloud on demand. In this way, administrators are not trapped in a day-to-day solution to the problem of virtual layer operations, while configuring the new virtual machines to reduce the probability of human error generation.

But creating, configuring, and managing virtual machines in the cloud is different from managing practices in existing data centers. In a virtual infrastructure, existing change management routes provide the process of creating new virtual machines that can limit the virtual machine's wanton growth. However, in a cloud computing environment, the challenge is to develop a user-driven environment that does not contribute to the uncontrolled growth of virtual machines. In addition, a virtual machine template for standardizing the hardware and software settings of a new virtual machine seems to include only a basic operating system, service packs, and other patches. Given the performance concerns, most organizations do not advocate the inclusion of complete applications and services in these templates.

However, in a private cloud, one of the ultimate goals is to allow end users to create new applications and services on demand. When end users log on to a cloud portal, they want to see a service directory, not a virtual application that contains only the basic operating system. They need a complete service or application.

So you need to face past assumptions and procedures. For templates, this means "stack," where services and applications are installed in virtual machines. You need to work closely with the people involved in the traditional management of these applications and get permission to configure the virtual machines.

The virtual machines need to be tested and validated to some extent before they are included in the service catalog. Therefore, you need to take appropriate control measures to ensure that the virtual machines that grow in the virtual infrastructure do not become the virtual machines that grow in the cloud. While maintaining the standards of the enterprise, we should also take some balanced measures to give the end users a certain degree of freedom.

One way to achieve this degree of freedom is to provide pre-packaged services that end users do not need to be overly optimized and customized to use.

You can also simplify configuration and configuration processes by creating virtual machine "levels" in the automation engine, such as platinum, gold, silver, and copper level four. With this level of design, IT managers can build and select different virtual machine templates in advance, and users can access templates within a certain range of applications and services. This tiered approach helps to control performance and consumption, and can be a real projection of their resource consumption for departmental units. The tiered model limits the amount of CPU and memory and helps set the stage for the deduction policy.

A Vm¬first policy. While cloud computing does not exclude physical servers, the higher your existing infrastructure virtualization is, the easier it is to implement the cloud. If you haven't done so, use a "vm¬first policy" that defaults to virtualization of new services and new applications. They are then deployed on the specified physical server only if they prove that the services cannot perform virtualization.

Another is to rethink the physical servers that were excluded in the early stages of virtualization. These physical servers may be overly performance-sensitive servers that are considered more difficult to virtualize. With the significant progress of these management procedures, these systems can be moved to the virtual layer.

Finally, the policy and change management procedures that are enforced on virtual machines are re-examined. Are they still valid, or are they a throwback to the work done in the real world? Now that virtualization has proven its potential in data center production workloads, it needs to develop a more aggressive policy.

Configuration Storage is simple

In a cloud computing environment, configuring enough storage resources is a recognized challenge. In a private cloud, storage is multi-tenant, but this pattern will bring technical challenges and war between it forces.

Structural differences. Server Virtualization and Enterprise-class storage technology have different paths to development. So trying to combine them and get the benefits of a cloud computing environment is often not a good idea. For example, a company running a storage area network device of a certain size must be able to access the device directly, or even set up a storage pool to start a separate virtual machine. Compared to a standard virtualized server, this is a single mirror of the virtual disk space that has been embedded in it, assuming that the user operates on a host capable of processing instructions, such as CPUs, and that he can also talk directly to the onboard storage device.

The ideal virtualized host environment is a large, single server with as many cores, RAM, and direct-attached storage as possible. But that's not how the infrastructure and storage Area network works with a single server. However, it does not mean that advanced, expensive, and secure storage devices do not apply to virtualization.

As a result, private cloud architects should carefully examine how storage devices interact with the overall architecture of the datacenter. If your storage pool is the best, and virtualization is implemented, you don't need to be involved in too much management after setup. When you connect virtual resources to an unknown infrastructure pool through a wider channel, your storage management interface will not be able to collaborate seamlessly with virtual machines.

Storage access. In traditional virtual environments, access to storage devices is tightly controlled, and virtualization administrators are fighting every week and even daily to get the necessary storage resources. In the cloud, end users can access gigabytes or even terabytes of storage with mouse clicks without the same oversight as before. The challenge, therefore, is twofold: herding culture and technological change.

The job of cloud administrators is to deliver storage resources very artistically, both easy to use and the idea of "no free Lunch". When end users choose from the service catalog, the best cloud automation software can help them understand the cost of storage through the debit program.

Today, a number of virtual platform storage management Plug-ins, such as VMware's Virtual Suite vsphere, allow administrators to configure new storage resources directly from VMware's management console. These plug-ins save a lot of time and automation, which is time-consuming even with the help of scripting tools. However, while Plug-ins are a great boon, storage teams may hesitate to give virtualization administrators the right to use the tool, because extending access channels will reduce the power of control over storage arrays.

Configuring the network is simple

For the infrastructure that is about to implement cloud computing, the matching network also needs a major overhaul. This homogeneity presents new challenges to network bandwidth and configuration, although the private cloud masks low-level differences at the infrastructure level to achieve scale flexibility and dynamism.

Bandwidth。 Even if your network is connected to a Gigabit Ethernet and has a few strong links to meet everyone's needs, you may still face bandwidth problems. Therefore, be prepared to invest in monitoring network blocking tools. If you virtualize everything that can be virtualized and start to service all of these resources by the network, and the user has the means to do it on their own, the bottleneck problem is relatively fast.

If virtual machines are growing unchecked, it's a problem for your IT department, and implementing a private cloud will create a bigger problem. You may already have a team that supports a handful of servers, resulting in a lot of disruption to other businesses. Now imagine them doing this at home, blocking your entire running Internet connection until you can control them. If you are also planning to implement desktop virtualization or workspace virtualization, this vexing problem will continue to emerge. The design of the client/server means that the actual work takes place at both ends of the network, and the information is exchanged over the network, while cloud computing means that most of the work takes place in the data center as well as the message to the user.

To address these issues, consider redistributing and expanding bandwidth for users who lack resources before implementing the cloud computing strategy.

Many IT departments have a "fair" subconscious, where all the organizations ' network resources should be shared equally, whether they really need them or not. However, the distinction between different types of users should be planned in place, where resources should be recommended.

A consolidation of many physical servers into a small number of virtual environments does not necessarily increase network traffic, nor does it require special consideration in resource allocation.

But transforming your data center into a private cloud means providing more services, and more services need to be delivered over the web to users who need to be volatile when they rise. You should consider your bandwidth requirements and upgrade issues.

VLAN tag. A virtualized network also requires a separate virtual machine to ensure the privacy of data passing between tenants in the cloud. Therefore, they need to implement a mechanism to ensure that these networks share the same physical network connections without causing information leakage between networks.

To allow access to a physical network, most cloud computing automation software uses a virtual local area network (VLAN) tag model. This method requires a network team to create a VLAN ID pool in advance on a physical switch. When you create a new virtual machine or virtual application, even if a cloud end consumer ends up using these van IDs, you don't have to ask the network team to set them up.

But the VLAN defined on the physical switch is not "infinite". Most physical switches support only a certain number of VLAN definitions, and the VLAN namespaces are run out faster than expected. The biggest change here is to make the network team that created the VLAN think it's a good idea. In some ways, it ignores a generation of best practices that previously recommended that IT managers only make the necessary configuration to protect resources from being hijacked by evil intruders.

Virtual switches. IT managers need a protection strategy for logical configuration and Virtual Switch management, and virtual switches will provide virtual machine connectivity. Virtual administrators may need to re-examine their default settings, which might have been created for server consolidation projects. For example, most virtual switches have a large number of "ports" for the "insert" of a virtual machine. Consider it a traditional physical device, such as a 48-port switch. Of course, in the virtual world, you can have a lot more "ports" than the real world. Most virtual switches use a static model to assign ports to virtual machines. This static port pool will soon run out, so a virtual administrator must always focus on vswitch settings, using a more dynamic pattern or a method for creating and destroying vswitch ports when the port is needed or not needed.

Private Cloud is simple

While vendors are turning their services into cloud computing, touting private clouds as simple and easy to implement (with their help, of course), IT managers should draw attention. This is illustrated by your experience, other IT projects (software migrations or old hardware upgrades), technology changes, and personnel changes.

The private cloud infrastructure is no different. A real private cloud model means re-consider all the infrastructure components that make up your data center, and those who manage IT resources. Don't be afraid to start all over again and come from a vendor challenge. It will make heavy changes (including change management) to achieve the goal.

"Edit Recommendation"

Analysis: Trends and dynamics of cloud computing at home and abroad Gartner: Cloud services middleware leading cloud computing applications as a barrier to cloud computing promotion? The business model of cloud computing must find a new "landing point" cloud computing will cause the IT industry to change the six predictions "editor: Xin Wei TEL: (010) 68476606"