How To Share Local Storage
Local storage can now provide capabilities like virtual machine migration and distributed resource management.
As I covered in my last column, local storage is beginning to give shared storage some competition as the storage platform of choice in virtual environments. In theory, shared storage should have an advantage since multiple hosts have to have access to the same virtual images for capabilities like virtual machine migration or distributed resource manager to work. Vendors that are promoting this SAN-less data center concept have developed alternative ways to share that data.
In this column we will discuss how those vendors are creating a SAN-less environment that still can provide capabilities like virtual machine migration while at the same time benefiting from the simplicity and performance of local storage. There are two common approaches to accomplish this feat: mirroring/replication and something we call the SAN-less SAN. In this column we will cover the mirroring/replication technique; we will discuss the SAN-less SAN in our next.
More Storage Insights
- Storage Switzerland Developing an Enterprise flash strategy
- Build a Highly Affordable and Scalable Private Cloud
The simplest approach is a technique that leverages a mirroring or replication model. Basically an alternative host is designated and data from the first host is replicated to that host. A mirroring approach means that the data on the target host is 100% in sync with the data on the source host. A replication model means that the target host may be slightly out of sync with the source.
If the source host or a VM on the source fails or needs maintenance, the target host or VM can be moved into production. With the replication technique, the VM on the source has to be gracefully shut down for a clean startup on the target. The mirroring technique should be able to be started up instantly and should not need a graceful shutdown.
Initially, the mirroring or replication technique was popular in smaller data centers that simply needed availability more than they needed all the benefits of virtual machine migration. In fact, as we discuss in our article "For the Small to Medium-sized Company Backup is All About Recovery TIME" some backup applications now provide this functionality. They backup physical and virtual servers to a backup appliance and then in the case of failure or maintenance can host the virtual machine directly on the backup appliance. Essentially this solves two problems at the same time; providing virtual machine flexibility and data protection.
The mirroring concept, as we describe in our article "The Benefits of a Flash Only, SAN-less Virtual Architecture" is becoming more appealing to enterprises because it leverages PCIe SSDs installed inside of hosts and integrates with the fault tolerance software that comes with the hypervisors. This is a particularly interesting approach. It takes full advantage of the performance capabilities of PCIe SSD while at the same time provides automated, 30 second failover of failed VMs or hosts.
The downside of the mirroring/replication techniques is you need another copy of the virtual machine image for each physical server that you want to have the potential to host a particular virtual machine. In the smaller enterprise this may not be problem since capacity is inexpensive and they often get more than they need. Also remember that, as we discussed in our last column, local storage should be less expensive than shared storage.
Another downside is that the target VMs are consuming some level of CPU and memory resource on the target hosts on which they reside. Again, especially in the small enterprise and even in the large enterprise, there is often plenty of excess CPU.
Finally, there is the downside of loss of flexibility. With shared storage any server connected to it can typically host any VM. With the mirroring/replication technique you are limited to just the designated target. This shortcoming is overcome by the second technique, The SAN-less SAN which we will cover in an upcoming column.
In-memory analytics offers subsecond response times and hundreds of thousands of transactions per second. Now falling costs put it in reach of more enterprises. Also in the Analytics Speed Demon special issue of InformationWeek: Louisiana State University hopes to align business and IT more closely through a master's program focused on analytics. (Free registration required.)