Introduction to Hyperconverged Infrastructure
In recent years, it seems like technology is changing faster than it used to in decades past. As employees devour newer technologies such as smartphones, tablets, wearables, and other devices, and as they become more comfortable with solutions such as Dropbox and Skype, their demands on enterprise IT intensify. Plus, management and other decision makers are also increasing their demands on enterprise IT to provide more infrastructure with less cost and time. Unfortunately, enterprise IT organizations often don’t see much, if any, associated increases in funding to accomodate these demands.
These demands have resulted in the need for IT organizations to attempt to mimic NASA’s much-heralded “Faster, Better, Cheaper” operational campaign. As the name suggests, NASA made great attempts to build new missions far more quickly than was possible in the past, with greater levels of success, and with costs that were dramatically lower than previous missions. NASA was largely successful in their efforts, but the new missions tended to look very different from the ones in the past. For example, the early missions were big and complicated with a ton of moving parts, while modern missions have been much smaller in scale with far more focused mission deliverables.
What is NASA?
NASA is the United States National Aeronautical and Space Administration and has been responsible for helping the U.S. achieve success in it’s space programs, from the moon landing to recent high quality photographs of Pluto. NASA has faced serious budget cuts in recent years, but has been able to retool itself around smaller, more focused missions that cost less and have achieved incredible results.
The same “faster, better, cheaper” challenge is hitting enterprise IT, although even the hardest working IT pros don’t usually have to make robots rove the surface of an inhospitable planet! Today’s IT departments must meet a growing list of business needs while, at the same time, appeasing the decision makers who demand far more positive economic outcomes (either by cutting costs overall or doing more work within the existing budget).
Unfortunately, most of today’s data center architectures actively work against these goals, because with increasing complexity comes increased costs — and things have definitely become more complex. Virtualization has been a fantastic opportunity for companies, but with virtualization has come some new challenges, including major issues with storage. With virtualization, enterprise IT has moved from physical servers, where storage services could be configured on a per-server basis, to shared storage systems. These shared storage systems, while offering plenty of capacity, have often not been able to keep up in terms of performance, forcing IT departments to take corrective actions that don’t always align with good economic practices. For example, it’s common for IT pros to add entire shelves of disks, not because they need the capacity, but because they need the spindles to increase overall storage performance. There are, of course, other ways to combat storage performance issues, such as through the use of solid state disk (SSD) caching systems, but these also add complexity to what is already a complex situation.
There are other challenges that administrators of legacy data centers need to consider as well:
- Hardware sprawl. Data centers are littered with separate infrastructure silos that are all painstakingly cobbled together to form a complete solution. This hardware sprawl results in a data center that is increasingly complex, decreasing flexible, and expensive to maintain.
- Policy sprawl. The more variety of solutions in the data center, the more touch points that exist when it comes to applying consistent policies across all workloads. Policy sprawl leads to situations in which IT pros have to
- Scaling challenges. Predictability is becoming really important. That is, being able to predict ongoing budgetary costs as well as how well a solution will perform after purchase are important. Legacy infrastructure and its lack if inherent feature-like scaling capability make both predictability metrics very difficult to achieve.
- Desire for less technical overhead. Business want analysts and employees that can help drive top line revenue growth. Purely technical staff are often considered expenses that must be minimized. Businesses today are looking for ways to make the IT function easier to manage overall so that they can redeploy technical personnel to more business-facing needs. Legacy data centers are a major hurdle in this transition.
So, with all of this in mind, what are you to do?
Hyperconverged Infrastructure from 30,000 Feet
An emerging data center architectural option, dubbed hyperconverged infrastructure, is a new way to reduce your costs and better align enterprise IT with business needs. At its most basic, hyperconverged infrastructure is the conglomeration of the servers and storage devices that comprise the data center. These systems are wrapped in comprehensive and easy-to-use management tools designed to help shield the administrator from much of the underlying architectural complexity.
Why are these two resources, storage and compute, at the core of hyperconverged infrastructure? Simply put, storage has become an incredible challenge for many companies. It’s one of— if not the — most expensive resources in the data center and often requires a highly skilled person or team to keep it running. Moreover, for many companies, it’s a single point of failure. When storage fails, swaths of services are negatively impacted.
Combining storage with compute is in many ways a return to the past, but this time many new technologies have been wrapped around it. Before virtualization and before SANs, many companies ran physical servers with directly attached storage systems, and they tailored these storage systems to meet the unique needs for whatever applications might have been running on the physical servers. The problem with this approach was it created numerous “islands” of storage and compute resources. Virtualization solved this resource-sharing problem but introduced its own problems previously described.
Hyperconverged infrastructure distributes the storage resource among the various nodes that comprise a cluster. Often built using commodity server chasses and hardware, hyperconverged infrastructure nodes and appliances are bound together via Ethernet and a powerful software layer. The software layer often includes a virtual storage appliance (VSA) that runs on each cluster node. Each VSA then communicates with all of the other VSAs in the cluster over an Ethernet link, thus forming a distributed file system across which virtual machines are run.
Figure 1: An overview of a Virtual Storage Appliance
The fact that these systems leverage commodity hardware is critical. The power behind hyperconverged infrastructure lies in its ability to coral resources – RAM, compute, and data storage – from hardware that doesn’t all have to be custom-engineered. This is the basis for hyperconverged infrastructure’s ability to scale granularly and the beginnings of cost reduction processes.
If you’re new to hyperconverged infrastructure or are unfamiliar with the basics, please read Hyperconverged Infrastructure for Dummies, available now for free from www.hyperconverged.org.
Resources to Consolidate
The basic combination of storage and servers is a good start, but once one looks beyond the confines of this baseline definition, hyperconverged infrastructure begins to reveal its true power. The more hardware devices and software systems that can be collapsed into a hyperconverged solution, the easier it becomes to manage the solution and the less expensive it becomes to operate.
Here are some data center elements that can be integrated in a hyperconverged infrastructure.
In order to achieve the most storage capacity, deduplication technologies are common in today’s data center. Dedicated appliances are now available which handle complex and CPU-intensive deduplication tasks, ultimately reducing the amount of data that has to be housed on primary storage. Deduplication services are also included with storage arrays in many cases. However, deduplication in both cases is not as comprehensive as it could be. As data moves around the organization, data is rehydrated into its original form and may or may not be reduced via deduplication as it moves between services.
SSD Caches/All-Flash Array
To address storage performance issues, companies sometimes deploy either solid state disk (SSD)-based caching systems or full SSD/flash-based storage arrays. However, both solutions have the potential to increase complexity as well as cost. When server-side PCI-e SSD cards are deployed, there also has to be a third-party software layer that allows them to act as a cache, if that is the desire. With all-flash arrays or flash-based stand-alone caching systems, administrators are asked to support new hardware in addition to everything else in the data center.
Data protection in the form of backup and recovery remains a critical task for IT and is one that’s often not meeting organizational needs. Recovery time objectives (RTO) and recovery point objectives (RPO) — both described in the deep dive section entitled “The Ins and Outs of Backup and Recovery” — are both shrinking metrics that IT needs to improve upon. Using traditional hardware and software solutions to meet this need has been increasingly challenging. As RPO and RTP needs get shorter, costs get higher with traditional solutions.
With the right hyperconverged infrastructure solution, the picture changes a bit. In fact, included in some baseline solutions is a comprehensive backup and recovery capability that can enable extremely short RTO windows while also featuring very small RPO metrics.
There are critical recovery metrics – known as Recovery Time Objective (RTO) and Recovery Point Objective (RPO) that must be considered in your data protection plans.
Data protection is about far more than just backup and recovery. What happens if the primary data center is lost? This is where replicated data comes into play. By making copies of data and replicating that data to remote sites, companies can rest assured that critical data won’t be lost.
To enable these data replication services, companies implement a variety of other data center services. For example, to minimize replication impact on bandwidth, companies deploy WAN acceleration devices intended to reduce the volume of data traversing the Internet to a secondary site. WAN accelerators are yet another device that needs to be managed, monitored, and maintained. There are acquisition costs to procure these devices; there are costs to operate these devices in the form of staff time and training; and there are annual maintenance costs to make sure that these devices remain supported by the vendor.