Server virtualization deployments continue to grow at an amazing rate, which isn't surprising. When properly implemented, server virtualization brings several key benefits, including reliability and a high return on investment. Clients are looking to XenServer as a virtualization platform that will deliver consolidation, increased utilization and workload balancing.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
This article covers best practices that solutions providers should use when designing a XenServer installation and deployment plan.
Using service level agreements for XenServer architecture
When using XenServer for production, you must be able to match the proper design to the service-level agreement (SLA). Clearly-defined SLAs are driving factors behind design and will help you determine if the free version is suitable or if you need the high-availability component that is included in the Citrix Essentials package. There are still many other architecture options to consider, including remote site disaster recovery that is built directly into XenServer.
Testing media and server hardware for XenServer installation
Although it shouldn't have to be said, it is important to always verify the installation media for XenServer. I have successfully built flakey, somewhat operational servers only to find out during the rebuild process that the installation .ISO file was corrupt. Taking five minutes to verify that the installation media is functioning properly will save you time and money.
I can't stress enough the importance of testing the server hardware. You should treat the hardware as though you were going to strap yourself on it and fly to the moon.
I find that many organizations I work with do not perform hardware testing prior to implementation, and it seems that these two main factors are almost always involved:
- The hardware hasn't been stressed to expose weakness. Over the years, I have found that most server hardware is being used at about only 10% of capacity.
- Organizations are conditioned to running Windows and Linux on server systems that have such a low utilization rate that in-depth hardware testing has not been necessary. It is very necessary now.
Testing hardware is just one of a number of cultural shifts that needs to happen. Remember, without virtualization, the hardware is not being put under stress. But, if you are not pushing the hardware up to about a 70% level of utilization, then you are missing a great benefit of virtualization: The whole idea is to leverage virtualization to get rid of as many physical servers as possible, which frees up physical space in the server room, decreases power consumption and reduces operational costs.
Use the vendor's hardware configuration disks to run the hardware diagnostics because if you find a problem, the vendor will ask if you used its software during testing. I recommend that you let the disks run for at least 48 hours, but I prefer 72-hour testing as a standard.
You will also want to make sure that the client's server hardware supports virtualization. Generally speaking, any Intel or AMD chips manufactured after August 2006 that have virtualization capability should work. If you plan on creating a resource pool with multiple physical host machines, the CPUs have to be from the same manufacturer and should be in the same family. They can have slightly different speeds, but if you get chips that are too dissimilar, they will not be able to join the pool.
Storage considerations during XenServer installation
XenServer is designed to support an impressive number of storage repositories by default. And Citrix just announced a new Essentials package that includes StorageLink for XenServer, which allows hundreds of storage arrays to behave like native XenServer storage. Even without the Essentials package, XenServer allows users to directly link into NetApp and Dell EqualLogic storage arrays and use the features, such as snapshots and data deduplication.
Even when you are running virtual machines (VMs), the underlying storage still needs to be configured to meet the demands of the applications running on the operating system inside the VM. Usually you have different RAID levels on your shared storage and place your VMs on the appropriate LUNs, just as you would with a physical server.
Network configuration and XenServer deployment
The network configuration will vary depending on many factors, including the number of onboard network interface cards (NICs), IP-based storage, and server or blade limits. Generally, you want to have as many multiple paths and as much redundancy as possible at each level in the design. XenServer supports up to six single NICs or 12 NICs bonded, and you should add as many as you reasonably can. Depending on if you are using IP-based storage, you may have two NICs for management, two for storage and at least two for VM traffic.
Completing the XenServer installation process
Once you have all of the physical components assembled and tested, building XenServer is about a 10-minute job. Before you start, you will need some basic information: the XenServer host IP address, DNS servers, NTP servers and XenServer host name. Then, a setup wizard will walk you through the installation process. While the server completes the installation, install XenCenter on your client machine so that you'll be ready to start managing the server right away.