New virtual servers OVH in comparison with domestic state employees
OVH is a world leader in hosting services and needs no introduction. In 2016, the hoster announced a new line of virtual servers www.ovh.ie/vps/vps-ssd.xml . It has a new platform: now, instead of AMD Opteron processors, Intel Xeon E5v3 is used, SSD drives are used as drives, and the OpenVZ virtualization system is replaced by KVM. All this led to the fact that the new virtual servers VPS SSD 2016 v1 in comparison with previous configurations of Classic 2014 have become much faster.
At the same time, OVH promises SLA at the level of 99.95%, unlimited traffic at 100 Mb / s, protection against DDOS attacks, monitoring server status and the ability to use up to 16 IP addresses on the server. But all these "goodies" I would not be interested if it were not for the price ...
For non-EU residents, which include webmasters from Russia, prices for virtual servers start at 3 Euros.
')
At the time of writing the publication, at a cost of foreign currency of about 75..80 rubles for 1 Euro, the price of VDS at the minimum tariff did not exceed 240 rubles per month.
I was very surprised that the virtual server from the world leader in hosting with 2 GB of RAM, SSD disk, unlimited traffic and protection against DDOS attacks is so cheap. Therefore, I decided to test it on the suitability of hosting sites and compare it with domestic virtual servers with a rental price of up to 250 rubles per month.
I did not reinvent the wheel and, for comparison, used companies that had already participated in my previous publication habrahabr.ru/post/282429 . These are Digital Ocean , Flops , Simple Cloud and Vscale . All of them, with the exception of Vdsina, offer virtual servers at a price of up to 250 rubles per month inclusive.
Since the companies represented offer virtual servers in different data centers, I found it necessary to specify the specific locations of the equipment.
I will start the comparison with the review of tariffs at the providers considered:
As you can see, OVH fits perfectly into the company of inexpensive domestic hosts.
To carry out tests on all servers, the CentOS 7.X operating system was installed with a bit of 64 bits with all the updates at the time of writing this review. To configure the web server, I used the free VestaCP control panel with a set of application software:
To test the site, I used the info.net.ru test blog on WordPress 4.5 with the Twenty Fifteen template supplied with this CMS. This system is very popular with bloggers and at the same time quite voracious. Therefore, its use in tests is very vital and will not be a measurement of a spherical horse in a vacuum.
I will test virtual servers for suitability for hosting sites. Therefore, in this review, I will not lay out the sheets of logs of synthetic testing, but I will give only the results in the “for housewives” format. From which you will immediately see how quickly the site will work on the server being tested.
As you know, the server's response time to user requests directly depends on the distance between them. For visitors from Russia, foreign data centers are at a disadvantage in terms of ping compared to domestic sites. But if you use a server for hosting sites, then delays of less than 100 ms can be considered acceptable, since they are not noticed by the user.
I took measurements from Volgograd. This is a good point for measurements, as the quality of the connection to the global Internet is here somewhere between capitals and rural areas. And it can be considered the average for Russia.
During the measurement, the ping value to the Yandex search engine servers ya.ru (93.158.134.3) was consistently 31 ms (I checked it through three providers: Sprint Network , Columbia Telecom and Beeline )
The table with the results, the smaller the obtained value - the better:
Flops pleased. Foreign servers lose, which of course was expected. But the magnitude of the delay is within reason.
This test will indirectly evaluate the performance of the virtual server. The higher the result, the better:
In synthetics disappoint Flops. And the OVH is kept in the middle.
The first real test. For its implementation in the bottom of the page, insert the code in PHP . With its help, we will track the time spent on the construction of the page and the eaten megabytes of memory.
The table with the results, the smaller the obtained value - the better:
The fastest is Vscale. Flops rehabilitated. All go tight in one group.
With the help of the ping-admin service, I will imitate visitors from Russia and check the time and speed of loading the site.
For OVH, the result is:
The table with the results. Time - the less, the better. Rating - the more stars, the better:
In Russia, the fastest - Vscale. OVH organically entered into the group of domestic providers.
With the help of the free PingDom service , I will test the speed of loading pages on the site for a conditional visitor from Sweden.
The table with the results. Time - the less, the better. Rating - the more percent, the better:
In Europe, the fastest - Simple Cloud. OVH in the lead. All go very tight.
All providers did not have any problems at the free tariff from the loadimpact.com service, which simulates 50 simultaneous connections. For these virtual server configurations, this is the “children's” load, so there will be no tables and graphs.
I also did not find anything criminal with the help of the perfload.ru service. It simulates a load of 7,200 visits per hour. Everyone coped with the task, the picture in OVH looks like this:
During testing, the Digital Ocean server behaved strangely. Either I chose an unsuccessful Data Center, or I took measurements at an unfortunate time, or I simply had no luck due to an accidental combination of circumstances.
Flops despite the weak synthetics showed good results on a real site. Simple Cloud and Vscale alternately ranked first in the tests. And OVH kept in the middle.
Testing with a real site has shown that all companies from this review are suitable for hosting.
During testing, OVH and Simple Cloud did not use SWAP virtual memory. This is due to the large amount of RAM they have compared to other reviewers. Servers with 512 MB of RAM have dropped to SWAP 150 ... 170 MB of data.
OVH has released a viable line of virtual servers. Prices at the level of domestic providers, and the stated characteristics, in particular the amount of RAM, is noticeably higher. In addition, none of the domestic hosts, I have not found official information about the protection against DDOS attacks, which is present in OVH by default. True, I do not know how to test this protection in real life.
The performance of the virtual server was very good and suitable for hosting. The site hosted at OVH opened equally quickly with visitors from both Europe and Russia.
But there are OVH and cons. This is the lack of cloud "buns" in the form of server activation in a couple of minutes and hourly payment for using the server.
I really want that after the publication of this review, domestic providers quickly increased the amount of RAM on their tariffs and implement protection against DDOS attacks.
At the same time, OVH promises SLA at the level of 99.95%, unlimited traffic at 100 Mb / s, protection against DDOS attacks, monitoring server status and the ability to use up to 16 IP addresses on the server. But all these "goodies" I would not be interested if it were not for the price ...
VDS tariffs from OVH
For non-EU residents, which include webmasters from Russia, prices for virtual servers start at 3 Euros.
')
At the time of writing the publication, at a cost of foreign currency of about 75..80 rubles for 1 Euro, the price of VDS at the minimum tariff did not exceed 240 rubles per month.
I was very surprised that the virtual server from the world leader in hosting with 2 GB of RAM, SSD disk, unlimited traffic and protection against DDOS attacks is so cheap. Therefore, I decided to test it on the suitability of hosting sites and compare it with domestic virtual servers with a rental price of up to 250 rubles per month.
Companies survey participants
I did not reinvent the wheel and, for comparison, used companies that had already participated in my previous publication habrahabr.ru/post/282429 . These are Digital Ocean , Flops , Simple Cloud and Vscale . All of them, with the exception of Vdsina, offer virtual servers at a price of up to 250 rubles per month inclusive.
Sites and Data Centers
Since the companies represented offer virtual servers in different data centers, I found it necessary to specify the specific locations of the equipment.
- Digital Ocean - Germany, Frankfurt
- Flops - Russia, Moscow, DC “StoreData”
- OVH - France, Strasbourg, DC SBG1
- Simple Cloud - Russia, St. Petersburg, DC "Miran 2"
- Vscale - Russia, St. Petersburg, DC Selektel Dubrovka
Tariff Plans
I will start the comparison with the review of tariffs at the providers considered:
| Provider | Virtualization | CPU | Memory | Disk | Traffic | Price |
|---|---|---|---|---|---|---|
| Digital ocean | KVM | 1x1800 MHz | 512 MB | 20 GB SSD | 1 TB | 5 USD ~ 350 rub. |
| Flops | KVM | 1x2400 MHz | 512 MB | 16 GB SSD | 0.9 TB (30 GB per day) | 250 rub. |
| Ovh | KVM | 1x2400 MHz | 2048 MB | 10 GB SSD | Unlimited | 2.99 Euro ~ 240 rubles. |
| Simple cloud | KVM | 1x2500 MHz | 1024 MB | 20 GB SSD | Unlimited | 250 rub. |
| Vscale | KVM | 1x2300 MHz | 512 MB | 20 GB SSD | 1 TB | 200 rub. |
As you can see, OVH fits perfectly into the company of inexpensive domestic hosts.
Configuration for testing
To carry out tests on all servers, the CentOS 7.X operating system was installed with a bit of 64 bits with all the updates at the time of writing this review. To configure the web server, I used the free VestaCP control panel with a set of application software:
- Nginx / 1.8.1 web server as light front-end
- Apache / 2.4.6 web server as application back-end
- MySQL 5.5.44-MariaDB database server
- PHP 5.4.45 popular web scripting language with ZendOpCache
To test the site, I used the info.net.ru test blog on WordPress 4.5 with the Twenty Fifteen template supplied with this CMS. This system is very popular with bloggers and at the same time quite voracious. Therefore, its use in tests is very vital and will not be a measurement of a spherical horse in a vacuum.
Testing method
I will test virtual servers for suitability for hosting sites. Therefore, in this review, I will not lay out the sheets of logs of synthetic testing, but I will give only the results in the “for housewives” format. From which you will immediately see how quickly the site will work on the server being tested.
Site response time
As you know, the server's response time to user requests directly depends on the distance between them. For visitors from Russia, foreign data centers are at a disadvantage in terms of ping compared to domestic sites. But if you use a server for hosting sites, then delays of less than 100 ms can be considered acceptable, since they are not noticed by the user.
I took measurements from Volgograd. This is a good point for measurements, as the quality of the connection to the global Internet is here somewhere between capitals and rural areas. And it can be considered the average for Russia.
During the measurement, the ping value to the Yandex search engine servers ya.ru (93.158.134.3) was consistently 31 ms (I checked it through three providers: Sprint Network , Columbia Telecom and Beeline )
The table with the results, the smaller the obtained value - the better:
| Provider | Response time, ms |
|---|---|
| Yandex | 31 |
| Digital ocean | 58 |
| Flops | 31 |
| Ovh | 74 |
| Simple cloud | 37 |
| Vscale | 41 |
Flops pleased. Foreign servers lose, which of course was expected. But the magnitude of the delay is within reason.
UnixBench Synthetic Testing
This test will indirectly evaluate the performance of the virtual server. The higher the result, the better:
| Provider | UnixBench value |
|---|---|
| Digital ocean | 1035 |
| Flops | 530 |
| Ovh | 1200 |
| Simple cloud | 1450 |
| Vscale | 1165 |
In synthetics disappoint Flops. And the OVH is kept in the middle.
WordPress Page Generation Rate
The first real test. For its implementation in the bottom of the page, insert the code in PHP . With its help, we will track the time spent on the construction of the page and the eaten megabytes of memory.
SQL - <?php echo get_num_queries (); ?> | <?php timer_stop (1); ?> . | <?php echo round (memory_get_usage ()/1024/1024, 2) ?> The table with the results, the smaller the obtained value - the better:
| Provider | Page generation time, s |
|---|---|
| Digital ocean | 0.108 |
| Flops | 0.080 |
| Ovh | 0.086 |
| Simple cloud | 0.083 |
| Vscale | 0.075 |
The fastest is Vscale. Flops rehabilitated. All go tight in one group.
Site loading speed for visitors from Russia
With the help of the ping-admin service, I will imitate visitors from Russia and check the time and speed of loading the site.
For OVH, the result is:
The table with the results. Time - the less, the better. Rating - the more stars, the better:
| Provider | Load time, with | Download speed, MB / s | Rating |
|---|---|---|---|
| Digital ocean | 0.40 | 0.362 | **** / ** |
| Flops | 0.72 | eleven | *** / ***** |
| Ovh | 0.70 | ten | *** / ***** |
| Simple cloud | 0.67 | ten | *** / ***** |
| Vscale | 0.32 | eight | **** / **** |
In Russia, the fastest - Vscale. OVH organically entered into the group of domestic providers.
Site loading speed for visitors from Europe
With the help of the free PingDom service , I will test the speed of loading pages on the site for a conditional visitor from Sweden.
The table with the results. Time - the less, the better. Rating - the more percent, the better:
| Provider | Load time, with | Rating,% |
|---|---|---|
| Digital ocean | 1.8 | 75 |
| Flops | 0.752 | 94 |
| Ovh | 0.582 | 96 |
| Simple cloud | 0.434 | 98 |
| Vscale | 0.821 | 93 |
In Europe, the fastest - Simple Cloud. OVH in the lead. All go very tight.
Stress Testing
All providers did not have any problems at the free tariff from the loadimpact.com service, which simulates 50 simultaneous connections. For these virtual server configurations, this is the “children's” load, so there will be no tables and graphs.
I also did not find anything criminal with the help of the perfload.ru service. It simulates a load of 7,200 visits per hour. Everyone coped with the task, the picture in OVH looks like this:
Test results
During testing, the Digital Ocean server behaved strangely. Either I chose an unsuccessful Data Center, or I took measurements at an unfortunate time, or I simply had no luck due to an accidental combination of circumstances.
Flops despite the weak synthetics showed good results on a real site. Simple Cloud and Vscale alternately ranked first in the tests. And OVH kept in the middle.
Testing with a real site has shown that all companies from this review are suitable for hosting.
During testing, OVH and Simple Cloud did not use SWAP virtual memory. This is due to the large amount of RAM they have compared to other reviewers. Servers with 512 MB of RAM have dropped to SWAP 150 ... 170 MB of data.
findings
OVH has released a viable line of virtual servers. Prices at the level of domestic providers, and the stated characteristics, in particular the amount of RAM, is noticeably higher. In addition, none of the domestic hosts, I have not found official information about the protection against DDOS attacks, which is present in OVH by default. True, I do not know how to test this protection in real life.
The performance of the virtual server was very good and suitable for hosting. The site hosted at OVH opened equally quickly with visitors from both Europe and Russia.
But there are OVH and cons. This is the lack of cloud "buns" in the form of server activation in a couple of minutes and hourly payment for using the server.
I really want that after the publication of this review, domestic providers quickly increased the amount of RAM on their tariffs and implement protection against DDOS attacks.
Source: https://habr.com/ru/post/301624/
All Articles