Yes, I have taken the plunge and I am on to a new challenge. I had a great ride at NetApp and I will cherish every moment I spent there, learning a lot which helped me grow both personally and professionally.
I think it is important to take on new challenges and adapt to changes because nothing ever stays the same, which means unless you can adapt and change too, you will be stuck doing the same things, which could then make your life a lot harder than it needs to be.
So, what makes Tintri so special?
Tintri VMstore was born out of two major disruptions in the market – Flash and Virtualization.
Flash on one side, has changed the way vendors think about product and filesystem architecture. It has also impacted how the storage solutions are architected in a datacenter. Virtualization on the other side, has completely disrupted the traditional thought around solution sizing, expected behavior, as well as how the infrastructure is monitored and managed. Not only that, it has also had a big impact on the IT teams especially around the skills that an IT Manager needs to plan to have in the team.
Although, vendors have tried to retrofit, both Flash and Virtualization need an architecture designed from the grounds up. While there are a lot of vendors who have brought focus on Flash and have come out with products designed for Flash from the grounds-up, no one has focussed on one of the biggest challenge that the storage admins face in a modern Datacenter today – Server Virtualization
Tintri VMstore is a Flash Storage designed from the grounds up to take advantage of Flash and be deployed in virtualized datacenters bringing in simplicity, predictability, quality of service, sub-millisecond response times, VM awareness (around IO and functionality), in-built data protection and WAN efficiency at an affordable cost from both Capex and Opex standpoint, without trying to oversell provisioned Vs usable capacity.
Tintri VMstore uses a mix of Flash and SATA which does make it hybrid but because of its Flash First design that is VM aware, it is able to deliver 99% of the IO from Flash. While doing that it ensures that the traditional problems around using Flash as a first citizen are taken care of, enabling it to deliver a predictable performance and sub millisecond latency. The filesystem design is such that if required, Tintri can drop the SATA disks at any point in the future and go all Flash. The whole idea behind having SATA disks in the system is that there is a big difference in the cost of Flash Vs SATA and the customers have a very high percentage of data that is not used for a very long time (Our Autosupport data validates that). So it makes sense to use the SATA for things like snapshot copies, sequential pattern of data or any blocks that are not needed on Flash. Also, since the storage has inline dedup and compression enabled, it amplifies the effective Flash capacity and is therefore able to serve 99% of the IO from the Flash. In case there is a workload that needs that sustained sub-millisecond latency, it can be pinned to flash as well. Here is graph on how % age of IO served from Flash has an impact on the latency and where Tintri VMstore plays-
The other thing that really helps Tintri, which is also one of its biggest differentiator is that it is VM aware. The VM awareness is not limited to just the functionality (Snapshot, clones, replication/vaulting etc.) but it is extended to IO. Every VM gets it’s own queue to various resources (CPU, Network, Flash, Memory, NVRAM) which enables QoS, fair share and takes care of noisy neighbors, thus ensuring that a throughput sensitive application is not impacting a latency sensitive app. The working set is calculated per VM as well ensuring that a wild VM doesn’t take away the Flash resources.
All of this is complemented by the fact that the system is self managed and the only knob exposed to the user is the ability to pin the VM to Flash. So no Storage pools, RAID groups, no LUNs to be created, no volumes to manage and no tunings to be done. And if you ever get into trouble, there is a very powerful GUI to help you troubleshoot and pin-point the problem at VM, network and/or Storage. You don’t have to be a VCDX or a Storage Guru to manage the storage.
The best part of the GUI is the Capacity and Performance Meter which tells you exactly how much headroom is available on the system. In my experience, 80-90% of performance problems are caused by users not being aware of how much headroom the system has. The result is a system that was sized for ‘X’ workload is now loaded 2X. The free capacity in the system because of dedup and other efficiency technologies further amplifies the problem.
You can find a lot of details in the videos here from the Storage Field Day and here on the Tintri website (Check out the Troubleshooting and Tintri Global Centre Demo) . Rob Waite, my colleague in Australia, has written a detailed post here as well.
I am very excited about the opportunity especially when I look at the great engineering and leadership team, a solid vision and upcoming enhancements. In case you would like to join us on this great journey, I invite you to visit Tintri’s career page. We are hiring and are looking for great talent. Feel free to reach out to me at firstname.lastname@example.org and I’ll take care of directing you towards the right person.
Thanks for reading…