By now, you’re already familiar with Hyperconverged Infrastructure (HCI) – what it is and its advantages. We’ve also discussed some of the less-obvious benefits customers see once they’ve deployed HCI. In this post, we’ll be presenting some of the benefits that are specific to StorPool’s implementation of hundreds of HCI clouds.
StorPool Ensures More Resources Go to Applications
One of the concerns when considering a hyperconverged deployment model is whether the servers will have the resources for all of your applications and their VMs while also running the storage software. The servers are doing double-duty, unlike servers in a more traditional disaggregated deployment.
This concern is understandable. In some Hyperconverged Infrastructure (HCI) solutions, the storage software consumes up to 40% of the server’s RAM and CPU. StorPool Storage’s software-defined storage (SDS) solution was created with efficiency in mind. StorPool’s HCI deployments use ~12% of a server’s RAM and ~10% of its CPU – far less than required by other solutions, leaving more resources available for your VMs and applications.
StorPool Prevents “Noisy Neighbor” Resource Contention
Anytime a server is used for more than one purpose, the possibility of the “noisy neighbor” becomes a concern. This occurs when one of the workloads starts consuming more and more of the server’s resources, causing performance issues for all the other workloads. With HCI servers running VMs, applications, and storage services, it’s easy to see where the “noisy neighbor” could be a problem.
The StorPool Storage SDS platform was built with proper resource isolation in mind. StorPool has implemented a tool called StorPool Control Groups (storpool_cg) that minimizes the complexity of managing Linux kernel control groups and helps ensure the StorPool storage system is protected from random events like out-of-memory states or latency spikes. It creates a “firewall” of sorts, isolating the resources used by the storage services from the other server resources. The StorPool storage software is unable to use resources beyond its pre-defined limits, and VMs and applications are unable to access any of the resources reserved by the storage services. This eliminates the possibility of the storage and application software fighting for resources.
Reduce Cost by Running Cloud Services as VMs
Many organizations approach virtualization and on-premises cloud infrastructure deployments with extreme caution, perhaps even an overabundance of caution. These organizations will run their cloud management platform (CMP) software on a separate standalone server rather than as a VM. The thinking is that if the cloud were to crash, they would need access to their CMP in order to restart all the VMs. Modern cloud infrastructure deployments can be configured to restart crashed VMs automatically.
When deploying an HCI cloud with StorPool, this function can be executed by the storpool_havm service, not the CMP. The service tracks the state of one or more virtual machines and keeps them active on a node in the StorPool cluster, offloading the CMP responsibility for these virtual machines. If something goes wrong, the service ensures fast startup on another node in the cluster. This means that, for example, in the event of a total host failure, all the VMs running on it would be automatically restarted on another host in the cluster. Later, if the host returns to operation, its VMs can be non-disruptively live-migrated back. This makes running the CMP management server as a VM a possibility and a cost-saving option.
Using this approach, it becomes possible to deploy CMP services and other auxiliary services (e.g., a highly available NFS file service) in a way that makes them less likely to experience downtime.
StorPool Offers Complete Choice in Scaling
One of the great benefits of an HCI deployment is that scaling resources is simple. Simply add an additional server to the cluster and all resources (storage, RAM, CPU, and network) will scale simultaneously.
But what if you don’t need that linear scalability. You might have a workload that requires more CPU and RAM, but not additional storage or, the other way around, a workload that requires more storage capacity.
Many HCI solutions, especially the hardware-based ones, only allow linear scalability. That is if you want to scale one of the resources, your only choice is to scale all of them. This means you’re paying for some additional resources you don’t need and won’t use.
StorPool Storage’s HCI solution is extremely flexible and doesn’t require that all of the server nodes follow the HCI model where both applications and storage services run together on every server. With StorPool you can add single-purpose nodes to your infrastructure seamlessly. Need more CPU and RAM but not additional storage? Add one or more compute-only nodes. Need more storage capacity but not additional compute power? Add one or more storage-only nodes.
An example of this type of Mixed Hyperconverged Infrastructure (HCI) is shown in the diagram below.
StorPool Helps Eliminate Storage Silos
One of the benefits of StorPool Storage’s Hyperconverged Infrastructure (HCI) model is that it’s not restricted to a single type of workload or entire software stack. This makes consolidating storage far simpler, allowing you to eliminate all of your separate storage silos.
For example, if you’re currently using VMware vSphere or VMware Cloud Foundation (VCF) but starting to migrate to Linux KVM with a different CMP (like OpenNebula or CloudStack), you can use the StorPool HCI cluster for both environments. This makes the data migrations and conversions less worrisome than moving data between different storage systems.
Another example would be merging IT infrastructure following a merger or acquisition. If your company has built its cloud based on OpenStack and acquires another company that built their cloud based on Proxmox, both clouds can run on a single StorPool cluster simultaneously.
One last example, in addition to your cloud, your company might be running a business-critical high-transaction database. A StorPool Storage cluster – HCI or not – is capable of providing high-performance to both environments at the same time.
StorPool Eliminates Planned Downtime
One advantage of StorPool Storage’s implementation of Hyperconverged Infrastructure (HCI) is its redundancy and reliability. StorPool’s clustered architecture makes planned downtime a thing of the past.
With StorPool, all maintenance and updates are non-disruptive. Your infrastructure can receive hotfixes, patches, updates, hardware maintenance, and hardware refreshes without any downtime required. You can see how that would benefit your organization’s continued operations.
StorPool Provides Reliable, Predictable Performance
One of the long-term problems plaguing storage is the connection between performance and capacity utilization. Most data storage performs at its best when less than 50% of its capacity is used. Many storage solutions see a slight decrease in performance when the system is above 50% capacity. It gets worse the more capacity is used. For years, people in the IT industry have simply accepted that all storage solutions lose performance when they hit 80% capacity utilization.
The workaround for this has always been the same – purchase more storage capacity. This requires that you purchase additional capacity that you plan on not using. Imagine a real estate agent selling you a house and then telling you that you can’t enter the living room. No one would stand for that. But for years, storage vendors have expected customers to purchase extra capacity that they won’t use, as if the “performance takes a steep dive at over 80% capacity utilization” were a law of physics, like gravity.
StorPool Storage, on the other hand, was built brand-new from the ground up, and thus, it hasn’t inherited the problems that have plagued legacy storage solutions. StorPool consistently delivers and maintains sub-100 μs in-VM latency.
It continues to consistently deliver this level of performance under simultaneous load from hundreds of VMs running on the cluster. We’ve seen our clusters continue to deliver sub-100 μs latency while the cluster was serving a workload of more than 10 million IOPS.
And, StorPool Storage doesn’t fall victim to the 50% capacity threshold or even the 80% threshold. StorPool delivers the same IOPS and latency all the way up to 95% capacity utilization. We deliver performance without requiring our customers to purchase additional capacity they can’t use. This is just another way we help our customers reduce data center costs.
Learn More
You can learn more about our HCI solution here. You can read our customers’ success stories, or contact us today to schedule a meeting or demo with our storage experts.