Over the past six years, server virtualization has greatly simplified the deployment of applications in the enterprise. Applications are no longer locked into a single server and can be migrated to other servers, in real time, to satisfy dynamically changing requirements or host maintenance windows. The challenge is that traditional storage systems have become increasingly more complex to operate and tune to support these virtualized environments.
There are two primary problems that contribute to today’s current storage complexity. First, legacy storage was designed for the days of one application per server, not the dozens of applications per virtualization host we see today. They assign storage at a LUN or volume level and typically can’t adjust storage I/O performance or data protection levels on a per-VM granularity. Second, legacy storage infrastructures require a dedicated set of switches and network adaptors. Storage networking introduces additional management complexities such as zoning, LUN masking, etc.
Maxta promises to address these shortcomings with a software-defined storage solution that runs on all the virtualization hosts in a virtualization cluster. Maxta creates a global namespace, abstracts and aggregates all the servers’ attached disk drives and flash drives to create a shared storage pool accessible by all the VMs across the cluster. The global namespace enables all capabilities that are delivered by shared storage such as VM mobility. The storage in the servers can be a combination of hard disk drives and either PCIe SSD flash or solid state drives.
Since the storage software runs on the same servers that are running the hypervisor and the application VMs, the need for a separate external shared storage infrastructure and the need for storage networking is eliminated.
Storage performance, assuming a correctly sized host processor, should be excellent in the Maxta design. All reads should be from local storage. A large percentage of these reads are cached in the SSD layer, enabling great performance. Writes are sequentially written to the SSD, synchronously replicated to the SSD on another server in the cluster before acknowledging back to the application. At a later point in time, writes are de-staged asynchronously and sequentially to the spinning disks. This approach helps improve performance significantly for virtual workloads. The replication of data across the cluster enables High Availability and no loss of data or access to data even if an entire server fails.
Finally, Maxta has the complete storage data services one would expect in a mid to high end storage array. Their feature set, in addition to others, includes thin provisioning, snapshots, clones, deduplication and compression.
Storage Swiss Take
Maxta is a product to take notice of. It has the potential to consolidate the server, network and storage layers into a single tier which should allow for a simpler virtualized infrastructure that scales automatically as new hosts or disk drives are added to the environment.