Modern convergent technologies on the market in the Russian Federation - let's try to compare? Import Substitution, Ceph + OpenStack, Nutanix
Today in the Russian Federation an extremely interesting situation has developed in the IT market.
There are (successful or not - a separate issue) attempts to import substitution in IT, there are many products "made in Russia", special lists and so on.
We do not want and will not talk about politics, but let's talk about technology and pricing.
')
Today I would like to talk about one of the most “heated” topics in the world and the Russian Federation ( extremely profitable for manufacturers) - “cloud” platforms for storing and processing data — in other words, storage systems, virtualization, clouds, and so on.
Three options were chosen for consideration: import-replacement products that are widely advertised in the Russian Federation and Nutanix as the leader (or one of the leaders) of the global HCI market.
The overwhelming majority of “stacks made in the Russian Federation” fit into the “Openstack + Ceph”.
Paraphrasing, we are talking about a modern approach to building IT infrastructures on the principle of how Amazon, Google, Facebook and others do, but with local features .
As a slight digression, back in 2014, Gartner predicted that in 2017 most of the largest / key companies would use webscale solutions, and they were right.
More than half of the world's largest companies (from Global 100) already use WebScale (infographics - below).
Percent of Global Enterprises Gartner Says By 2017 Web-Scale
Analysts generally agree that traditional approaches to building infrastructures are becoming obsolete (the so-called “three-tier infrastructures” based on data storage systems), and the re-division of the market is extremely fast.
Extremely interesting infographics and analytics :
Now you understand why the market is so hot even in the Russian Federation and why “import substitutes” behave like in the joke “extinguish the light, they climb into the light”, while (my personal opinion) the real situation with technologies very much resembles this:
In fact, we are talking about the re-division of multi-billion (not in rubles) revenues even in the Russian Federation:
Data processing and storage is now a key problem for almost any business and organization in the world, and public IT purchases of Russian companies amount to hundreds of millions of dollars.
Of course, there are other competitive commercial Western vendors, we can talk about them separately.
The essence of the article is to understand what “Russian products” can technologically offer to a client technologically, whether import substitution is real or paper in this case, how expensive or cheap it is .
Many professionals know that in the overwhelming number of cases all these products were not really developed in Russia, and they either use ceph + openstack (by feeling -> 95% of "import substitutions"), or they pretend that they are developing the product locally, but in reality myself western software code.
One of the interesting examples is how “With a slight movement of the hand, the pants turn into shorts.”
American company (formerly Parallels) Virtuozzo → Rosplatforma → SKALA-R
In order to avoid any misconceptions, I cite a link to the speech of the top-manager of Rosplatforma , where it is actually about “getting purchasing preferences for Russian companies”.
The Virtuozzo software code belongs to an American company .
Versioning / documentation is the same.
What do you think, what company really continues to develop the program code, and which simply uses a dedicated build server?
If we talk about the amounts, then we are usually talking about projects from hundreds of thousands to millions of dollars, but the solutions are also applicable to small companies - the threshold for entering a “new wonderful world” (for the Russian Federation) is about $ 35k.
IBS orients its Skala-R as “from an economic point of view, the SCALA-R platform may be of interest to companies with an IT budget of $ 500,000,” which, to put it mildly, is somewhat surprising.
...
By the way, positive exceptions (from the mass of “ready-made American code”) are definitely there — for example, Raidix data warehouses, which created a specialized solution for recording and playing media streams, and this is very successful, but not really suitable for serious projects building “corporate clouds” , nor for storing generic data (colleagues, please do not be offended - but this is real if you want to make a comparison of the functional - I will explain what basic “features” you have to seriously talk about cloudy enterprise application).
...
Today I would like to touch on at least basic questions about the technical part of the solutions, but let's talk a little about pricing.
I also note that the main arguments of "import substitutes" are in fact often the presence of a magic piece of paper (certification), but again, most understand that all this is in fact a profanation.
It is impossible to check (for the foreseeable time) for errors or bookmarks the program code (millions of lines) of the modern Linux kernel and the entire typical environment (based on which almost all the “substitutes” work), not to mention the mass of the additional code (ceph, openstack, virtuozzo and etc.)
Only the Linux kernel came close to 20 million lines of code .
As life shows, almost all the largest / scandalous holes in the Linux code and model libraries were not found and corrected in the Russian Federation (including heartbleed, etc.) - who are interested, in fact, you can take and see when the patches for “Russian Linux” were released.
Considering that ceph and virtuozzo work at the host / kernel level of the OS, this really is a serious problem of reliability and security.
So let's go. For now - about basic things.
Ceph + Openstack.
There is no single product as such; there is a mass of disparate projects (OpenStack components) in different status - from sufficiently high-quality development to completely abandoned.
The situation is reminiscent of “cancer + swan + pike” - commercial providers of OS at all costs try to tie the client to their decision, painless migration even between the products of one vendor is extremely difficult, between different ones is almost unreal.
The primary hypervisor (and in fact the only one applicable for ceph) is KVM.
As SDS (software data storage) - the real option for use is only ceph, a separate opensource project.
Theoretically, you can use ESXi and HyperV via iSCSI, but in practice, due to the lack of support for VAAI / ODX, this is the maximum laboratory option.
As a digression, I sincerely do not understand how you can run any projects on ceph for serious use - it suffices to study the bug tracker.
For example, for the last “stable” release of Jewel, there are about 200 critical (immediate, urgent, high) errors.
All this reminds me very much of walking on a minefield.
Bug tracker ceph
Let me remind you that at the end of 2016, ceph.com "lay" for two days due to the fall of the ceph-based storage . Very ironic.
The stories about how “powerful teams of Russian programmers” (usually it’s really about a maximum of 3-10 people) fixed all the ceph errors and created a Russian product - we leave it on the conscience of the sellers of such solutions, but a competent IT specialist should have questions - why hundreds / thousands of developers of the main ceph "could not", but the micro teams from the Russian Federation did it all.
Technologically, the architecture has a lot of bottlenecks - dedicated server metadata, extremely high RAM requirements, etc.
Interesting and relevant article on this topic (how to install crutches):
Ceph RBD performance issues
The fate of Openstack as a project as a whole also raises many questions.
To begin with, a couple of quotes from the Mirantis blog (despite the fact that in 2016, for example, they introduced OpenStack in the largest bank of the Russian Federation, already realizing that the OS is really “not a long-timer”):
" Infrastructure Software is Dead "
“Mirantis OpenStack software is better than everybody else. Everybody's OpenStack software is equally bad . There is software-defined networking, software-defined networking, cloud management platforms, platforms-as-services, container orchestrators, you name it. It is nightmare to operate. It's all bad. "
Today, almost all the leading “drivers” have abandoned the development - Mirantis dismissed a lot of employees from the OpenStack division and openly admits that they need to go to microservices, HPE sold the entire division, Rackspace, as one of the open-step icons, made a delisting from the stock exchange - it remains, perhaps, only RedHat.
Next, let's talk about the "free" of such solutions.
The price of commercial support for openstack in the Russian Federation is an average of $ 4-5 thousand (of those American dollars) per year for each server.
OpenStack cost
This coincides with the fact that our clients called us in the Russian Federation when the same Mirantis visited them, so we can take as a base for calculations.
There is also always an option to hire engineers to your team, but to ensure the normal functioning of the project, there are usually a few dozens of people who need to pay salaries. An additional problem is that it is practically impossible to find specialists, even those who were (let's take the same Mirantis) are already not many in the Russian Federation (due to the closing of the Moscow office).
By the way, commercial openstack support usually does not include ceph support, you have to pay for it separately.
Certification solutions - confusion and vacillation, some of the localized options are certified.
Starter PACK - from 45 thousand US dollars - at least 6 servers (3 for ceph, 3 for KVM and OS). The configuration is really laboratory, for industrial use it is required to distribute the services ceph.
Each server is about $ 5k ($ 30k per server), plus OpenStack support (about $ 5k per server per year).
Virtuozzo / Rosplatform / SKALA-R.
Parallels -> Virtuozzo - in fact, a respected, fairly common solution in narrow circles of service providers, sharpened for specific tasks (ensuring the operation of SP, usually - low-cost mass hosting).
They used their hypervisor as a virtualization stack, but in the 7th version they decided to switch to KVM, which " works 40% faster than regular KVM " (I immediately recalled an advertisement about "ordinary laundry detergent :)")
Virtuozzo storage is part of the solution, it’s also “p-storage” - actually distributed storage with a lot of restrictions (about it below) and a minimum of functionality, sharpened specifically for the needs of service provision, which is mentioned, among other things, in documentation interesting positioning solutions in the Russian Federation as a universal).
Ready PAK (software and hardware complex) from 4 simple servers starts at minimum from 7.8 million rubles (~ 130 thousand US dollars - you have to pay for import substitution)
“Skala-R” and “Rosplatforma” - it is not entirely clear from the official site - “Product safety is ensured by a complex of information security tools and readiness for certification at the FSTEC.”
By the way, an interesting point about the work of marketers in the Russian Federation - “The Rock” is a computer that stood at the Chernobyl NPP and generally participated in the monitoring processes. Interestingly, this name was chosen specifically?
Update: as expected, the representative of Rosplatforma came and began to assert that all the conclusions were incorrect, the documentation was old, no connection with the American company (parallels / virtuoso) remained.
The gentleman from “Rosplatforma” apparently does not know that any document can be viewed PDF tags.
Total - May 2017, the documentation is done by Parallels and Virtuozzo employees, while the documentation is of course fresh.
Nutanix
The creators of the HCI / WebScale market, leaders in sales, the mass of implemented projects in virtually all industries.
As I wrote above, more than half of the world's largest companies (from Global 100) already use Nutanix.
Certification in the Russian Federation is possible, there are procedures, at least for one of the largest projects in the Russian Federation, the FSTEC was obtained for the "perimeter".
Almost all key safety certifications in the world have been obtained.
Ready PAK in the Russian Federation starts from 35 thousand US dollars, at least 3 servers.
In fact, all the main infographics are attached:
...
The main technical functionality was compiled into a comparative table, which, among other things, I gave for consideration to Rosplatforma and IBS. They refused to make comments, although at first they showed interest.
The table on OpenStack + ceph checked with representatives of the OpenStack and ceph community in the Russian Federation.
We are talking only about stable functionality, no "beta versions" and "technical previews".
In the table there may be inaccuracies or errors, write in the comments - be sure or correct or explain.
Are there any conclusions? I will not do them, do it yourself. The information you have.
There are (successful or not - a separate issue) attempts to import substitution in IT, there are many products "made in Russia", special lists and so on.
We do not want and will not talk about politics, but let's talk about technology and pricing.
')
Today I would like to talk about one of the most “heated” topics in the world and the Russian Federation ( extremely profitable for manufacturers) - “cloud” platforms for storing and processing data — in other words, storage systems, virtualization, clouds, and so on.
Three options were chosen for consideration: import-replacement products that are widely advertised in the Russian Federation and Nutanix as the leader (or one of the leaders) of the global HCI market.
The overwhelming majority of “stacks made in the Russian Federation” fit into the “Openstack + Ceph”.
Paraphrasing, we are talking about a modern approach to building IT infrastructures on the principle of how Amazon, Google, Facebook and others do, but with local features .
As a slight digression, back in 2014, Gartner predicted that in 2017 most of the largest / key companies would use webscale solutions, and they were right.
More than half of the world's largest companies (from Global 100) already use WebScale (infographics - below).
Percent of Global Enterprises Gartner Says By 2017 Web-Scale
Analysts generally agree that traditional approaches to building infrastructures are becoming obsolete (the so-called “three-tier infrastructures” based on data storage systems), and the re-division of the market is extremely fast.
Extremely interesting infographics and analytics :
Now you understand why the market is so hot even in the Russian Federation and why “import substitutes” behave like in the joke “extinguish the light, they climb into the light”, while (my personal opinion) the real situation with technologies very much resembles this:
In fact, we are talking about the re-division of multi-billion (not in rubles) revenues even in the Russian Federation:
Data processing and storage is now a key problem for almost any business and organization in the world, and public IT purchases of Russian companies amount to hundreds of millions of dollars.
Of course, there are other competitive commercial Western vendors, we can talk about them separately.
The essence of the article is to understand what “Russian products” can technologically offer to a client technologically, whether import substitution is real or paper in this case, how expensive or cheap it is .
Many professionals know that in the overwhelming number of cases all these products were not really developed in Russia, and they either use ceph + openstack (by feeling -> 95% of "import substitutions"), or they pretend that they are developing the product locally, but in reality myself western software code.
One of the interesting examples is how “With a slight movement of the hand, the pants turn into shorts.”
American company (formerly Parallels) Virtuozzo → Rosplatforma → SKALA-R
In order to avoid any misconceptions, I cite a link to the speech of the top-manager of Rosplatforma , where it is actually about “getting purchasing preferences for Russian companies”.
The Virtuozzo software code belongs to an American company .
Versioning / documentation is the same.
What do you think, what company really continues to develop the program code, and which simply uses a dedicated build server?
If we talk about the amounts, then we are usually talking about projects from hundreds of thousands to millions of dollars, but the solutions are also applicable to small companies - the threshold for entering a “new wonderful world” (for the Russian Federation) is about $ 35k.
IBS orients its Skala-R as “from an economic point of view, the SCALA-R platform may be of interest to companies with an IT budget of $ 500,000,” which, to put it mildly, is somewhat surprising.
...
By the way, positive exceptions (from the mass of “ready-made American code”) are definitely there — for example, Raidix data warehouses, which created a specialized solution for recording and playing media streams, and this is very successful, but not really suitable for serious projects building “corporate clouds” , nor for storing generic data (colleagues, please do not be offended - but this is real if you want to make a comparison of the functional - I will explain what basic “features” you have to seriously talk about cloudy enterprise application).
...
Today I would like to touch on at least basic questions about the technical part of the solutions, but let's talk a little about pricing.
Spoiler
As you have probably already figured out, it is necessary to pay extra for “import substitution”, and be ready to turn a blind eye to the mass “this is not a bug, this is a feature”.
I also note that the main arguments of "import substitutes" are in fact often the presence of a magic piece of paper (certification), but again, most understand that all this is in fact a profanation.
It is impossible to check (for the foreseeable time) for errors or bookmarks the program code (millions of lines) of the modern Linux kernel and the entire typical environment (based on which almost all the “substitutes” work), not to mention the mass of the additional code (ceph, openstack, virtuozzo and etc.)
Only the Linux kernel came close to 20 million lines of code .
As life shows, almost all the largest / scandalous holes in the Linux code and model libraries were not found and corrected in the Russian Federation (including heartbleed, etc.) - who are interested, in fact, you can take and see when the patches for “Russian Linux” were released.
Considering that ceph and virtuozzo work at the host / kernel level of the OS, this really is a serious problem of reliability and security.
So let's go. For now - about basic things.
Ceph + Openstack.
There is no single product as such; there is a mass of disparate projects (OpenStack components) in different status - from sufficiently high-quality development to completely abandoned.
The situation is reminiscent of “cancer + swan + pike” - commercial providers of OS at all costs try to tie the client to their decision, painless migration even between the products of one vendor is extremely difficult, between different ones is almost unreal.
The primary hypervisor (and in fact the only one applicable for ceph) is KVM.
As SDS (software data storage) - the real option for use is only ceph, a separate opensource project.
Theoretically, you can use ESXi and HyperV via iSCSI, but in practice, due to the lack of support for VAAI / ODX, this is the maximum laboratory option.
As a digression, I sincerely do not understand how you can run any projects on ceph for serious use - it suffices to study the bug tracker.
For example, for the last “stable” release of Jewel, there are about 200 critical (immediate, urgent, high) errors.
All this reminds me very much of walking on a minefield.
Bug tracker ceph
Let me remind you that at the end of 2016, ceph.com "lay" for two days due to the fall of the ceph-based storage . Very ironic.
The stories about how “powerful teams of Russian programmers” (usually it’s really about a maximum of 3-10 people) fixed all the ceph errors and created a Russian product - we leave it on the conscience of the sellers of such solutions, but a competent IT specialist should have questions - why hundreds / thousands of developers of the main ceph "could not", but the micro teams from the Russian Federation did it all.
Technologically, the architecture has a lot of bottlenecks - dedicated server metadata, extremely high RAM requirements, etc.
Interesting and relevant article on this topic (how to install crutches):
Ceph RBD performance issues
The fate of Openstack as a project as a whole also raises many questions.
To begin with, a couple of quotes from the Mirantis blog (despite the fact that in 2016, for example, they introduced OpenStack in the largest bank of the Russian Federation, already realizing that the OS is really “not a long-timer”):
" Infrastructure Software is Dead "
“Mirantis OpenStack software is better than everybody else. Everybody's OpenStack software is equally bad . There is software-defined networking, software-defined networking, cloud management platforms, platforms-as-services, container orchestrators, you name it. It is nightmare to operate. It's all bad. "
Today, almost all the leading “drivers” have abandoned the development - Mirantis dismissed a lot of employees from the OpenStack division and openly admits that they need to go to microservices, HPE sold the entire division, Rackspace, as one of the open-step icons, made a delisting from the stock exchange - it remains, perhaps, only RedHat.
Next, let's talk about the "free" of such solutions.
The price of commercial support for openstack in the Russian Federation is an average of $ 4-5 thousand (of those American dollars) per year for each server.
OpenStack cost
This coincides with the fact that our clients called us in the Russian Federation when the same Mirantis visited them, so we can take as a base for calculations.
There is also always an option to hire engineers to your team, but to ensure the normal functioning of the project, there are usually a few dozens of people who need to pay salaries. An additional problem is that it is practically impossible to find specialists, even those who were (let's take the same Mirantis) are already not many in the Russian Federation (due to the closing of the Moscow office).
By the way, commercial openstack support usually does not include ceph support, you have to pay for it separately.
Certification solutions - confusion and vacillation, some of the localized options are certified.
Starter PACK - from 45 thousand US dollars - at least 6 servers (3 for ceph, 3 for KVM and OS). The configuration is really laboratory, for industrial use it is required to distribute the services ceph.
Each server is about $ 5k ($ 30k per server), plus OpenStack support (about $ 5k per server per year).
Virtuozzo / Rosplatform / SKALA-R.
Parallels -> Virtuozzo - in fact, a respected, fairly common solution in narrow circles of service providers, sharpened for specific tasks (ensuring the operation of SP, usually - low-cost mass hosting).
They used their hypervisor as a virtualization stack, but in the 7th version they decided to switch to KVM, which " works 40% faster than regular KVM " (I immediately recalled an advertisement about "ordinary laundry detergent :)")
Virtuozzo storage is part of the solution, it’s also “p-storage” - actually distributed storage with a lot of restrictions (about it below) and a minimum of functionality, sharpened specifically for the needs of service provision, which is mentioned, among other things, in documentation interesting positioning solutions in the Russian Federation as a universal).
Ready PAK (software and hardware complex) from 4 simple servers starts at minimum from 7.8 million rubles (~ 130 thousand US dollars - you have to pay for import substitution)
“Skala-R” and “Rosplatforma” - it is not entirely clear from the official site - “Product safety is ensured by a complex of information security tools and readiness for certification at the FSTEC.”
By the way, an interesting point about the work of marketers in the Russian Federation - “The Rock” is a computer that stood at the Chernobyl NPP and generally participated in the monitoring processes. Interestingly, this name was chosen specifically?
Update: as expected, the representative of Rosplatforma came and began to assert that all the conclusions were incorrect, the documentation was old, no connection with the American company (parallels / virtuoso) remained.
The gentleman from “Rosplatforma” apparently does not know that any document can be viewed PDF tags.
Total - May 2017, the documentation is done by Parallels and Virtuozzo employees, while the documentation is of course fresh.
Nutanix
The creators of the HCI / WebScale market, leaders in sales, the mass of implemented projects in virtually all industries.
As I wrote above, more than half of the world's largest companies (from Global 100) already use Nutanix.
Certification in the Russian Federation is possible, there are procedures, at least for one of the largest projects in the Russian Federation, the FSTEC was obtained for the "perimeter".
Almost all key safety certifications in the world have been obtained.
Ready PAK in the Russian Federation starts from 35 thousand US dollars, at least 3 servers.
In fact, all the main infographics are attached:
...
The main technical functionality was compiled into a comparative table, which, among other things, I gave for consideration to Rosplatforma and IBS. They refused to make comments, although at first they showed interest.
The table on OpenStack + ceph checked with representatives of the OpenStack and ceph community in the Russian Federation.
We are talking only about stable functionality, no "beta versions" and "technical previews".
In the table there may be inaccuracies or errors, write in the comments - be sure or correct or explain.
Are there any conclusions? I will not do them, do it yourself. The information you have.
| SCALA-R / Rosplatform / Virtuozzo | Nutanix | CEPH + OpenStack | |
|---|---|---|---|
| Software architecture In the case of operating at the host OS / kernel level, potential safety issues and fault tolerance. | Kernel / Host | Isolated Virtual Controller | Kernel / Host |
| “Clearance” of the solution | Service Providers | Enterprise (corporations), government services military, medicine, industry, resource companies, financial structures. | Service Providers |
| HCI (hyper-convergent) solution. Date Services and Virtualization working on servers at the same time. | Yes | Yes | not |
| No bottlenecks / points failure (centralized services, for example metadata server or management server) | not | Yes | not |
| STIG Policy (Security Technical Implementation Guide) | None, manufacturer's recommendation - manual rootkit search and hacking detection | Yes | There are no, there are a lot of disparate recommendations and techniques |
| Location of the company that developed the main part of the software | USA (Parallels -> Virtuozzo -> Rosplatforms -> ”SKALA-R”) | USA | USA (basic development), produced by many companies. Full source code publicly available. |
| Built-in full monitoring (all hardware and software components) and self-diagnosis | Partially, Scala-R uses additional monitoring tools | Yes | Partially |
| Intelligent Automatic Data Tiering - moving data blocks between cold and hot levels in case of a change in the frequency of requests for this data (“heating” or “cooling”) | not | Yes | not |
| Built-in self-service portal | not | Yes | Yes |
| Localization of VM data (data locality) - the active VM data is on the same server where the virtual machine is running. The cardinal acceleration of read operations and reduction of the load on the network at times. | not | Yes | not |
| VM level replication | not | Yes | not |
| Data integrity recovery | Manual start in case of loss of a large domain | Automatic start | Automatic start |
| Data deduplication | not | Yes | not |
| Data compression | not | Yes | not |
| Erasure Code (noiseproof coding) | not | Yes | Yes |
| All flash support | not | Yes | Yes |
| Mix All Flash and Hybrid in a single cluster | not | Yes | Yes |
| Failure handling of SSD drives with metadata | Temporary loss of part of data nodes and long recovery | Automatic testing, does not affect cluster performance, all data nodes continue to work | Automatic testing, may affect cluster performance |
| VAAI and ODX support (protocols for unloading I / O operations to the storage system) | not | Yes | not |
| Application Consistent Snapshots support (VSS provider for Windows Server and implementation for Linux) | not | Yes | not |
| Unlimited number of VM snapshots, without affecting the performance and the possibility of manipulation (including deletion) of snapshots at any level | not | Yes | not |
| Shadow clones (shadow disk) - creation of a local copy of VM disks for cardinal acceleration of loading and operation | not | Yes | not |
| Embedded backup on Amazon / Azure | not | Yes | not |
| Best Practice availability for key sample applications (Oracle RAC, MSSQL, Postgres, SAP NetWeaver, MongoDB, Microsoft Exchange, Cisco Unified Communications, and others) | not | Yes | not |
| Distributed fault-tolerant management stack (management plane) without using centralized databases | not | Yes | not |
| No dedicated management servers are required. | not | Yes | not |
| Automatic upgrades of hardware firmware (BIOS, firmware drives and flash, controllers, etc.) | not | Yes | not |
| Transparent update without restarting client services when changing the main software version | not | Yes | Yes |
| Hypervisor Support | KVM (non-certified patches), Virtuozzo (outdated, in the new version switched to KVM) | AHV ( KVM compatible , certified by Microsoft, SAP and others), XenServer, vSphere, HyperV | KVM It is possible to use vSphere and HyperV with an iSCSI connection; it is not recommended for production due to the lack of support for VAAI / ODX. Not a HCI solution . |
| Mainstream Desktop Virtualization Stacks (VDI) Support - Citrix, VMware | not | Citrix, VMware, Workspot and others | not |
| Ability to run Microsoft Windows Server with full Microsoft support (SVVP - Server Virtualization Validation Program) | No, SVVP certification is present for the old version of the American product with another hypervisor. Windows Server Virtualization - at your own risk | Full SVVP certification | There are for RedHat OpenStack and Canonical (Ubuntu) OpenStack, missing for most other options. |
| Container Virtualization | Virtuozzo | Docker | Docker, LXD |
| Built-in backups (without the use of third-party software) with file recovery and self-service | not | Yes | not |
| Metro cluster (distributed geo cluster with synchronous data replication) | not | Yes | not |
| Built-in support for emergency multi-recovery data centers (many to many DR) | not | Yes | not |
| Cross-hypervisor DR support | not | Yes | not |
| Automatic conversion of the hypervisor and all VMs in a cluster (for example, ESXi-> AHV / KVM) | not | Yes | not |
| Built-in SDN stack with hardware networking integration | not | Yes | Yes |
| Affinity / Anti-Affinity support, for optimization of software licensing. With no support unjustified licensing costs may be millions of dollars | not | Yes | Yes |
| RESTful API support (industry standard) | not | Yes | Yes |
| Free software upgrade to new versions (minor and global version updates) if available existing basic technical support | Scala-R - paid Rosplatform - free Virtuozzo - paid | Yes | Yes |
| Pricing (for the Russian Federation) for ready-made commercial solutions (PAK - hardware-software complex) | Starting price from $ 136,000 Cluster of at least 4 nodes | Starting price of $ 35,000 3- | 45000$ : 3 ceph, 3 , ~5$k ( OpenStack). ceph . |
| 95%, NPS score 92 | , |
Source: https://habr.com/ru/post/331306/
All Articles