The concept of an all-flash data center is very appealing, but for many data centers an upgrade to even a hybrid array is outside of their budget. Also, implementing a new all-flash array or even a hybrid array has more costs associated with it than just the upfront purchase. There is also the cost to migrate the data, verify that the applications can access the array, learn how to use the new array and, of course, update the array’s data protection procedures. While it may only take a day or two to implement a new storage array, the time to cut over can be weeks or even months.
Compounding this problem is the reality that many data centers considering an all-flash or hybrid array to address performance issues may not be at the point in the storage refresh cycle where it makes financial and logical sense to unplug a running storage system. In most cases, the current storage system has plenty of capacity or can be upgraded to meet the capacity demands. What is lacking is performance. If the performance side of the system could be upgraded to meet current and future demands, the organization could save a significant amount of both hard dollars and operational expenses.
Finally, in almost every data center most of the data is not presently active and hasn’t been for months, if not years. Obviously, this data does not need to be stored on an all-flash array, a hard disk tier or even the cloud is fine. The problem is, of course, identifying which data should be or will need to be on flash tomorrow. Without the right solution to help make data placement decisions most data centers over-compensate with an all-flash array.
The Cloudistics Turbine is an application acceleration appliance that inserts into the existing SAN architecture and accelerates block storage. It does so with minimal disruption, certainly less than the migration process described above, and it requires no changes to backend processes like data protection. Installation should take less than 20 minutes and involves shutting down your VMs, registering those VM’s LUNs to the Turbine appliance, and then pointing your VMs to the appliance.
The Cloudistics Turbine is a fault-tolerant appliance with two fully redundant 1U units, eliminating the single point of failure concern. The flash inside the system is a RAID 10 configuration that mirrors data across both units. It leverages an internal Infiniband connection in each appliance to reduce latency and strain on the existing network.
Once running, the Cloudistics Turbine includes read and write I/O acceleration to tackle performance problems. The write cache is fully protected thanks to the HA configuration of the appliance. It also pre-populates the read cache with the best data.
Cloudistics, thanks to the size of the flash caching area, goes a step further with I/O Reduction, which eliminates redundant write traffic. For example, if a database writes to a record 100 times in a short period, all I/Os are serviced by Turbine and delivered to the SAN array as a single write.
Finally, Turbine also preserves cache capacity for random I/O by automatically bypassing large sequential writes and sending them directly to the SAN array. Based on the load, Turbine determines if sequential data should go straight to the disk or if it should be written to Turbine’s cache.
The Cloudistics Turbine may be the perfect solution for data centers that can’t cost justify the all-flash array plunge or data centers that just don’t think that plunge makes a lot of sense. It allows data centers to address the performance demands being placed on them at about 90% less than the cost of moving to an all-flash array, plus it does not require the IT staff to learn new features or change backend processes.