Tonight the Patriots face the Texans, and I am excited to see the Pats pull out another victory. And after the Patriots beat the Dolphins without their star quarterback or the backup quarterback, I am starting to think that maybe the Super Bowl (which is the equivalent of total world domination for American football, even though it is only played in the US) might be possible. This led me to a re-read of last week’s article from Chris Mellor in The Register titled “Is hyperconvergence about to take over the enterprise data centre?” Mellor’s article also looks at the topic of total world domination, albeit in a completely different industry. It was a most interesting read featuring the perspectives of luminaries in the industry such as John Abbott from 451 Research, Marc Logen from Citihub Consulting, Tony Lock from FreeForm Dynamics, and Chad Sakac from Dell EMC.
One of the topics addressed in the article is why customers are moving to hyperconverged infrastructure. The benefits cited for adoption by Abbott include fewer systems to manage, scalability and performance, agility, and VM centricity. Sakac focused on the trend towards consumerization of IT. Certainly, SimpliVity is seeing similar reasons for embracing hyperconverged infrastructure from our customer base with one notable difference. A recent study of our customer base completed by IDC found the following reasons cited as top areas of improvement through the use of SimpliVity hyperconverged infrastructure:
Interestingly, neither Abbott nor Sakac mention what IDC found as the number one area of improvement selected by SimpliVity customers: improvement in backup/recovery and/or disaster recovery. That’s because hyperconverged infrastructure is traditionally viewed as the convergence of only compute, storage, and SAN layers.
SimpliVity takes a different approach than most hyperconverged infrastructure vendors since our solution includes built-in data protection that is capacity- and bandwidth-efficient with guaranteed restores of a 1 TB local VM in 60 seconds or less. This is why IDC found that 90% of SimpliVity customers are using the built-in data protection capabilities of our platform, and that a majority of those customers retired the use of existing third-party backup or replication software in favor of SimpliVity’s capabilities. Travelport is a great example of an enterprise that transformed its data center and improved its data protection by implementing SimpliVity hyperconverged infrastructure.
Another area where SimpliVity’s approach is different than other vendors in the industry is our support for compute nodes, which addresses one of the perceived limitations cited by Abbott: fixed scaling or the requirement to purchase compute and storage together rather than independently. With compute nodes, SimpliVity customers can connect existing x86 servers to their SimpliVity hyperconverged infrastructure systems should they require additional compute capacity. This enables customers to leverage more of their existing infrastructure and add more compute capacity into their environment as needed.
The article also covers which technology advances are supporting the move to hyperconverged infrastructure. Lock mentions virtualization and storage pooling as key, while Logen cites software and automation as the “magic sauce for hyperconvergence.” Lock and Logen’s points are spot-on, especially since solutions like SimpliVity’s enable the creation of a shared resource pool of compute and storage resources that is tightly coordinated across data centers and remote sites. The infrastructure can be fully automated, including the provisioning of new VMs and policies for data protection, via orchestration tools or REST APIs.
Sakac mentions faster networks as another notable technology advance that is contributing to the adoption of hyperconverged infrastructure. This one is more debatable and has perhaps led to some unwise choices in terms of how some vendors have architected their hyperconverged infrastructure platforms. A common architecture for hyperconverged infrastructure systems is a RAIN (Redundant Array of Independent Nodes) approach. This type of architecture relies on fast networks since I/O is distributed across multiple nodes, and that same network has to be utilized should a system outage or loss of one or more drives occur. The challenge is that it is based on the assumption that the network will be dramatically faster than the storage media. This was true with HDDs, but is no longer the case with solid state media.
The approach taken by SimpliVity is quite different. SimpliVity’s architecture is based on RAID (Redundant Array of Independent Disks, or how many nodes are required to sustain a failure with no loss of data) and RAIN (how many copies of data are being stored minimize or prevents loss of data in the event of node failure), which reduces latency as read I/O is not sent over the network since the complete set of data is maintained on the local node. This method improves the speed/performance of the infrastructure as a whole. A SimpliVity prospect recently evaluated our solution against a vendor with a RAIN-based approach. In their evaluation, they simulated several drive failures. When a fourth drive was pulled out of a competitive system, it failed completely and data loss was evident. The SimpliVity hyperconverged nodes continued without incident. So the bottom line is, yes, we agree with Sakac that networks are faster; however, they are not so fast that abandoning RAID altogether is the way to go.
Another important factor to note is the importance of data locality. Locating the data and compute resources of a VM on the same host can be a distinct advantage since it will minimize latency, and thus provide the most predictable performance of the storage layer. SimpliVity utilizes the OmniStack solution to maximize CPU and memory resources, but also provide it with full awareness of the location of the data within the cluster. This way, the data does not follow the VMs within the configuration – an inefficient process, all things considered.
Where I think we can agree, however, is that hyperconverged infrastructure will eventually take over the data center and become the primary alternative to hosting workloads in the public cloud. Although whether the Patriots will achieve world domination or not is still too early in the season to tell!