“The modern datacenter is more complex than ever before.” – Captain Obvious
At Tintri, we began our journey by asking a simple question: “What should storage look like if it only serviced virtual machines?”
On its surface, perhaps this seems like an easy question to answer.
“Hyper-converged!” or “All flash!” are among the top 5 answers on the board.
While those are reasonable answers, let’s first come to an agreement about what our fundamental responsibility in IT is.
*Puts IT hat on*
The fundamental contract that we hold with the business is to provide predictable and reliable services to support the technology needs of the business.
*Takes IT hat off*
Virtualization is one technology that we use to make the services our organization provides more predictable and more reliable.
Unfortunately, when it comes to storage, virtualization has the opposite effect. This is most notably because of the misalignment between what virtual machines are and how they are misrepresented as LUNs.
If predictability and reliability are key indicators for your organization, then applying all-flash to traditional block or file storage arrays does very little to improve your performance in these areas.
Similarly, grouping distributed disks across the nodes in your virtualization cluster does nothing beyond shortening the time from box-to-datastore.
Tintri’s approach is based on 4 key operating principles, which are different from other vendors:
1. Every VM should experience low latency regardless of its workload profile and whether that changes over time.
2. Performance should be consistent even if the hardware experiences a failure.
3. If there is a condition that impedes performance, intelligent infrastructure should give you visibility into what is causing the condition.
4. Achieving these benefits should require zero storage domain expertise.
Dynamic and Predictable Performance (100% of the time, every time)
The core of the Tintri experience is driven by a very specific and unique key feature:
Real-time IO profiling of every virtual disk in every virtual machine; and corresponsive tuning per virtual machine to achieve low latency.
This is where you must be saying to yourself, “All flash and NVMe is what gives you low-latency”.
However, that would be incorrect, or is not the whole story.
Storage media is one component in a complicated calculus that results in low storage latency.
Resources like CPU, memory, and network bandwidth all impact latency, and Tintri manages all of these resources per virtual machine to ensure that all virtual disks have equal access to that ultra-fast storage media.
The result is that each Tintri appliance is consumed as a single datastore. Any and every virtual machine stored in the Tintri datastore will receive all the performance they need as they are being used and none of the performance if they’re not being used.
Every Tintri appliance is a dual-node cluster in a box. The active-standby architecture ensures that even during a software upgrade, the virtual machines running on the appliance will experience zero impact.
Tintri appliances are deeply integrated into your virtualization environment. Leveraging a set of API calls into your virtualization manager, Tintri can see and show where in the stack latency lies.
Actionable information is surfaced intuitively and immediately; without having to run complex data collection jobs or dig through documentation.
Low Barrier to Benefit
Tintri appliances provide an extraordinary benefit to our customers. But Tintri understands that there is very little value to benefits that must be painstakingly extracted behind complex spreadsheets and time consuming and expensive professional services engagements.
That’s why, if you choose, any Tintri appliance can be installed quickly by your team, without any engagement from a professional services organization.
From our inception in 2008, all the way to 2020 and beyond, we believe that our approach to enterprise technology will continue to drive improvements in our customers’ operational experience.
And as the datacenter continues to march along towards intelligence-driven frameworks like AIOps, Tintri is poised to define the upper threshold for delivering value to our customers.