In “New Approach To Data Protection, Part 1” we cut down the amount of VMs being replicated by separating the VM from the underlying storage array. Now that we are at the VM level, we have to solve two other issues. These issues go hand in hand. In order to effectively remove data from the wire, the receiving system must be intelligent enough to reproduce the VM from a given set of blocks and receive only the blocks of data that it doesn’t already have.
To explain these issues, we return to our friendly neighborhood online retailer. Our retailer has recently recognized that some of their customers are smarter than others. The smart ones are able to leverage raw materials to produce the final product on their own. To better serve the intelligent consumer the retailer now sells kits for the widgets. Each kit is comprised of the materials required to produce the item and a set of instructions on how to put the item together. Once the customer receives the entire first shipment, they can assemble their product. Fortunately for the customer, none of the raw materials are consumed in the production of the product, so now to produce another one that is similar they only have to order something such as a different paint color to produce a new version.
By making these changes our retailer has transitioned from a traditional mail-order retailer, selling only finished goods, to a modern online retailer of the future. They now leverage the intelligence in the infrastructure to better utilize the resources available and nearly eliminate shipping altogether. This is exactly how an OmniStack Federation functions. Each time a replication needs to occur we check to see which blocks already exist at the receiving end. This allows us to send an updated set of instructions (metadata) on how to build the VM and then ship only the blocks the receiver needs.
So what have we accomplished? We have separated the VM from the underlying LUN. This eliminated a portion of the data across the wire. We also leveraged our deduplication by recognizing that blocks at the remote site don’t need to be sent again and again and again.
None of this would be possible if we hadn’t addressed deduplication from the beginning. In order for a storage system to effectively leverage true global deduplication, it must be built into the foundation of the architecture. Retrofitting deduplication produces a ripple effect in the architecture that makes the task impossible for most vendors.
SimpliVity OmniStack was built from the ground up to tackle the data inefficiency problem at the source. As soon as a block of data gets sent to the IO stack to be written out, it is deduplicated. If the block is already in the federation, it won’t ever be written again. This powerful, foundational capability is enabled by SimpliVity’s Data Virtualization Platform that provides a unique encapsulation layer to all VMs. This allows us to backup, recover, and move entire virtual machines between data centers without rehydrating the entire VM. This deduplication carries over from one VM to another and across all data centers and the cloud. It is one of the fundamental technologies enabling our expansive list of capabilities. For a live demo of this VM mobility in action, check out this video showing the efficient replication in action.