NutanixOS 4.5: Important Update and Big Plans

Finally, after a forced summer lull (very tiring news to know, but not to have the right to tell, believe me) and rapid internal boiling of news in the company, a fresh release was issued on the mountain.

In the spring, we had a big and important update of the “hardware platform” in connection with the transfer to the Intel Haswell CPU from our equipment supplier, and we released the “G4” model, Generation Four, and now, in the first days of October, Acropolis Hypervisor, AHV , and everything connected with it, and new plans for the software part of the product were announced.
')
But, in order.

All the summer long, the company had a lot of work to release a new, large and important release, which was called 4.5. The main news was that Nutanix decided to acquire its own hypervisor, having forked KVM and continued to develop it independently of RedHat. It was decided to call it Acropolis Hypervisor, as you remember, it was the Project Acropolis that was called our KVM Management Toolkit, which appeared at the beginning of this year.

First of all, it was caused by, well, besides a clear bla-bla about customer satisfaction and innovation, the fact that Nutanix has long had complex and forced close relationships with VMware. Essentially, I will say this, a very large half of the Nutanix business is systems under VMware vSphere. And everything, in general, suited everyone, while VMware did only the hypervisor and until it had a product that directly and directly competes with us, I'm talking about VSAN and EVO: RAIL . Inevitably, the question arose about measures to protect the business, as there was a risk and temptation for VMware to promote its product by “twisting the arms” to its main competitor, us. And in this, I must say, both VMware and EMC are still masters, and their dominant position in the market is a great temptation to do so.

Therefore, our developers sat down and wrote their own hypervisor, you know with what. The basis was taken open KVM, which, however, already for quite some time in the company sawed, primarily for stability and sekurnosti, so the choice was obvious.
Thus, we (and you, of course) now have a new hypervisor on the open core. Today, over 1200 nodes have been running on KVM-based Nutanix systems for about two years, such companies as Nintendo, Swisslos (the Swiss national lottery operator), the Asian multi-profile PCCW telecom holding, and others use them. But all was not limited to one hypervisor, whether it was worth raising so much noise because of forknuty KVM. No, everything is much more and more cunning, and the version of Nutanix OS 4.5 that came out a couple of days ago is just the beginning. In the coming months, many chips will appear in minor releases, which were decided to roll out gradually. I can’t talk about them, I’ll just mention that our GUI Prism control is waiting for a big update, and many new features will be integrated into it. There will also be several new uses for Nutanix.

But without him there is something to tell. Now I will talk about updates in a quick and general way, and then, in separate articles, as details arrive, I will talk about each feature in detail.

So, we have our own hypervisor, created on the core and KVM basis, and it is called Acropolis Hypervisor . This, however, does not mean that we stop working on VMware ESX / vSphere or Microsoft Hyper-V, not at all. They are still a big piece of the park of our users, and will continue to be maintained, developed and grow. But at the same time, we had the opportunity to “play our blackjack” and do some things in the way we consider correct. And, perhaps, to convince the correctness of our user approach.

First of all, apart from the AHV itself, two new architectural concepts have become part of the rollout, which, and this is important, is completely “cross-hypervisor”, that is, they will work on top of any hypervisor, not only with AHV, but also with VMware ESXi and MS Hyper-V, and even with future virtual, “cloudy nutanixes” in AWS and Azure.
This is what we have become known as Distributed Storage Fabric (DSF), and App Mobility Fabric (AMF).
Now "Acropolis" will be called all of the above.



Distributed Storage Fabric will include everything related to the “layer” of data storage on disks, for example, within the DSF framework, our Erasure Coding will develop, which I already wrote about when it appeared in Technical Preview, and which was released in production as once in NOS 4.5 published two days ago, and tools for creating and working with snapshots, instant snapshots of the state of disk storage of your system, and backup tools “into the cloud” of external cloud providers Amazon Web Services and MS Azure, and integration into third-party backup software, such as CommVault Simpana / IntelliSnap, and VM Flash Pinning, about which separately below.

App Mobility Fabric will include tools related to cross-hypervisibility and application migration in a Nutanix cluster environment. I already mentioned briefly that Nutanix is ​​busy with cross-browser, and now we already have a special capacity node, NX-6035C, a “disk shelf” for the Nutanix cluster, invented for the task of inexpensive expansion of storage space. At the same time, there are capacious and dense low-cost hard drives on the 6035C, but this all works under KVM, however, unlike other models that require clustering nodes from nodes managed by the same hypervisor, which is the same throughout the cluster, node 6035C c KVM internally can be included in an ESXi or Hyper-V cluster. And it is important that in this case you will not need to purchase licenses for sockets for commercial hypervisors, such as vSphere. Given their price - this is a significant benefit.

In the future, our task is to provide complete transparent cross hyper-visibility for applications. Your application, database, web service, it will be completely unimportant on which hypervisor it runs. An application is what matters to you, on what it works and how it doesn’t matter to a business if it works well.



Currently, within AMF, we are developing tools for easy migration from the ESXi environment to the Acropolis Hypervisor environment, as well as High Availability tools that allow restarting and maintaining the availability of virtual machines when the node-server on which it was running failed.

One of the interesting features that appeared in 4.5 released, and on which I promised to stay above, was VM Flash Pinning. As the name suggests, this is the ability to “pin” a VM and its vdisk at the Flash storage level.



As I mentioned earlier, one of the features of the Nutanix disk subsystem is its two-tier and transparent storage levels for applications. All active operations go on SSD, and due to this, speed and performance are provided, and for “cold”, voluminous, or written-read sequential data, a capacious and inexpensive level made up of SATA drives is available. As the data cools, they are transferred to SATA, transparent to the application. When data is accessed again, it migrates back to the SSD. It all looks transparent to the application, and it does not control it, it just has a large amount of SSD and SATA system disk capacities, and it is always fast. Or almost always.

Problems begin when the application needs quick access to the entire volume of data. When access time and latency operations are critical. When the situation when access time to neighboring data blocks can be significantly different, because the system decided that no one touched this data block for an hour, and let me demolish it on SATA, no one will notice.
For tasks, when access time is critical, and it should always be constant (and predictably submillisecond constant) in our time it is accepted to buy All-Flash storage. Such tasks, in all honesty, are few. For them, most of the hundred-vendor vendors have created systems called All-Flash Storage, we also have them, this is the NX-9000 series. However, the widespread use of AllFlash is quite expensive.

It is good when there is a lot of money and the IT department can spend it without considering, when you, say, a “out of the first hundred” bank, a large stock broker or, for example, a large ticket booking service, where delay in a split second is really a matter of business success. Overall, just buy Violin and forget about the problem of the year for three. The rest still have to choose the option to achieve the result as efficiently as possible with the available means. And one of them is to fix VMs with applications that need SSD speeds on these SSDs, forbidding them to migrate to a slower Tier. As a result, you get high All-Flash response speed for a particular critical application, and high storage efficiency for all other applications, plus flexibility in spending expensive Flash space.

Very briefly on the features of App Mobility Fabric.
One of the most interesting, in my opinion, was the mechanism for migrating virtual machines from external, third-party VMware vSphere installations to the Nutanix KVM environment (AHV, Acropolis Hypervisor).



Despite the fact that KVM some time ago had the opportunity to work with VMware virtual disks files (structures * -flat.vmdk), the question remained with the drivers of “virtual hardware”. Now, Nutanix has an automated service for embedding VirtIO drivers in Windows-based virtual machines, so now, with our Imaging Service, this process of creating and converting a virtual machine image from one hypervisor to another hypervisor has become even easier.
The functionality of this system is not limited to this opportunity, and will continue to expand, up to the complete victory of our idea about the Invisible Infrastructure, IT infrastructure, which is “not visible” when you use the result of its work, and for you all the same IT tools deliver this result to you.

Finally, without a special pomp, Nutanix developed integration with OpenStack , through a special integration mechanism-driver that allows you to embed Nutanix systems into the OpenStack infrastructure, recently wrote our architect Steven Poitras, the leading Nutanix Bible blog. For those involved in OpenStack, it may also be interesting to see what's there.



This is all we can talk about in October, we are waiting for further announcements in minor releases, with which functional modules will be added, and remember that the update NutanixOS on Nutanix systems occurs non-disruptive, without stopping the system and the availability of applications and services, is done from Prism web interface in your favorite web browser. The other day I just updated our test system, which we demonstrated to the customer. Just within an hour, during the break, everything happened, the update package was downloaded, which turned around and rolled on all four nodes of the cluster in turn, and the VM migration from the node to the cluster node, initiation and exit from maintenance mode, and other things. All applications continued to work.