Hosting and dedicated servers: answer questions. Part 1

In this series of articles we want to consider the issues that people have when working with hosting providers and dedicated servers in particular. Most of the discussions we conducted in English forums, trying to help users primarily with advice, rather than self-promotion, giving the most detailed and impartial answer, because our experience in the field has been over 14 years, hundreds of successful solutions and thousands of satisfied customers. Nevertheless, our answers should not be perceived as the only correct answers of the first instance, they may well contain inaccuracies and even errors, no one is perfect. We will be grateful if you add or correct them in the comments.

Hi, we are now using servers in OVH for storage, streaming and conversion.

When we had 500 online visitors who watched something, they did their job well and there were no delays in watching the video, but we quickly grew and sometimes the traffic began to reach 1,200 users. We noticed a big delay in buffering the video, previously it was buffered completely quickly to the end, but now it is not. We begin to notice problems as soon as 1000+ visitors.

What could you suggest / advise? Perhaps you need to increase throughput? I saw the option to increase the bandwidth to 1 or 3 Gbit / s, now we have 1 Gbit / s but without a guarantee, and I’m not sure about how much is guaranteed.

Indeed, your main problem is that you are using a shared connection, that is, a connection where the resulting bandwidth that you can generate on the Internet will depend on how much your rackmates consume at one time or another . Typically, such a connection is organized as follows – a rack-mount switch is connected to a distribution switch and, respectively, to the Internet at a speed of 10 Gb / s, and this bandwidth is distributed between 20 servers. Thus, everyone can be guaranteed to receive at least 500 Mbps if all the rackmates start consuming 100% of the channel. But sometimes this is not so. There is statistics on how much the servers actually consume. Not everyone consumes even 500 Mbit / s, because sometimes such a connection is divided into 40 servers and even more, thereby saving on the switch ports for connecting to distribution switches, especially in cases when the servers are single-unit and they can be put 40+ in one rack or when a higher stand is used.

For each stream, in a good way, you need to provide at least 2 Mbps (approximate HD bitrate in H.264 and .mp4 encoding format). Thus, for 1200 users, you need at least 2.4 Gbps and some kind of% for buffering. This% will depend on the streaming settings, network settings, and can be from 10 to 30% of the capacity you need. And if you are not able to do anything with the settings on the switch side (change the packet size and traffic), then you can save on buffering if you set a limit in the seconds of the video that will be buffered.

So, you need to have at least 3 Gb / s bandwidth for 1200 users who watch videos online with a bit rate of 2 Mbit / s, so that you do not have delays with buffering or even playback.

The second question is the performance of your storage. If you use exclusively SSDs, that’s good. But remember that most drives, especially in such cheap servers as in the OVH data center, can be connected via the SATA 2.0 interface instead of SATA 3.0. Thus, you can not get more than 2.4 Gb / s from each of the drives. And if all this is connected directly to the motherboard using the raid controller, which is included in the low-speed interface, then you will not be able to generate more than 2.4 Gb / s from all your drives.

Remember also that if the interface speed is 3 Gb / s, then in fact, you will not be able to transfer more than 2.4 Gb / s of data through it. Now we will not go into technical details why this is so, just take a word!

So, in order to investigate your situation fully you need to provide us with information about the technical characteristics of your server. Indicate the storage requirements (total file size, file bitrate or average duration and volume occupied by them), as well as which method you use for streaming.

In addition, do not forget that if you use this server not only for storage, but also for video conversion, you will surely get some bottlenecks. It can be not only the interrupt queue on the processors, but also the degradation of the throughput of your storage due to the encoding process, which also requires a certain number of input / output operations (IOPS).

My website is currently hosted on Bluehost, and I am not satisfied with the quality of their services (the site’s loading time and emails with a long delay are not particularly comfortable). My website is a design portfolio, it’s quite small, I don’t have any e-commerce solutions or anything unusual and resource-intensive, there are very few visitors on the website. I’m thinking about changing my hosting provider soon and am counting on a little help, because I am not an expert and I don’t know what some of the functions in the description of tariff plans mean, the price is not a problem.

https://hosterion.com/compare-hosting – here I would choose the cheapest rate, 3 € / month just because I don’t think I need something more than this is (translation into English), where I would choose Hosting * BusinessWEB 2,50 € / month (third from left to right).

Can you advise something? Thank you

It is better to choose the second option, it indicates that the storage is SSD. In the first option – it is not clear what type of storage is provided, a large quota is a signal that resources are being resold multipartically, that is, that there is an explicit overbooking. And how the provider ensures the availability of resources to everyone is not obvious.

How to choose the best web hosting?

  • It is very important that the hosting provider does not severely limit resource consumption, since in this case, if the load on your site exceeds the permissible level, your visitors will receive a web server error that will notify you that it cannot temporarily process the request, which will negatively affect the behavioral factors and positions of your site in search engines. Check how your provider considers resource consumption and what are the real limits. Some providers like to measure the load in CP – “parrots”, which are not representative (often you don’t know which processor on which of the nodes is used, because parameter 100 CP says little). 100 CPs on a high-performance server can allow you to serve thousands of visitors per day, while on nodes with a weak processor you will begin to experience problems even when the number of visitors does not reach even a thousand. The scripts that you are going to use are important, the same WordPress with modules can create a critical load even with one visit, if something is installed inappropriately or incorrectly, but no one is safe from this anywhere.
  • Try to contact the hosting support service and ask to create a test page with the generation time (by contacting it at different times it will be possible to see how much the result varies, strong differences in the generation time will indicate that the node is overloaded, and at peak hours your sites will be slower).
  • It is better to use hosting with a clean SSD storage, this will help to avoid delays in work due to the fact that one of the neighbors began to actively use the resources of the storage and its saturation was achieved for I / O operations, which began to become a queue. Do not forget that a conventional HDD-drive can provide only 70-120 random IOPS and even a RAID array of several disks often does not allow providing more than 1000 IOPS for output, and input is at least 2 times less, unless RAID0 of course used, which is unsafe, while an SSD can provide tens of thousands. Some providers save and use hybrid schemes when all demanding IOPS applications are hosted on the SSD, while rarely requested data on the HDD. This is good and appropriate, but not on the hosting. Today, the cost of SSDs is low and in most cases the resources of the Severus (CPU, RAM, IOPS storage, the latter is rarely the case with SSDs) are the first to end, rather than the quota, because the phrase “disk quota costs 0, costs money resources” is so popular .

Page generation time is the most important parameter for you and your users. And for most scripts, in the case of optimally configured nodes and the absence of a huge load from other users, working on a hosting will not present any problems. Recently, many of the providers have been using Cloudlinux, a solution that allows you to more clearly limit resources and separate users from each other, creating additional comfort and security, which increases the stability of the service as a whole. However, “mega-servers” are still popular, when thousands and tens of thousands of hosting clients are hosted on a single server, this is much more dangerous than the lack of the most advanced solutions in the form of Cloudlinux. Do not hesitate to ask your service provider how hosting is organized, the stability of your projects depends on this and even without being a specialist, you can understand if you are dealing with professionals in your field, or with schoolchildren who recently bought a server in Hetzner and they got a cool design and website that could be even better than the websites of reliable time-tested companies that provide services to tens of thousands of customers.

But not everything depends on the server, provider and settings. Of great importance is the data center, its location and transit providers, through which traffic will be delivered to your visitors. Therefore, no matter what the provider seems good, it is not superfluous to check the connectivity – the quality of the connection to the web server from different points of the network.

Ping is by no means the most important characteristic, as many mistakenly believe. For most sites, ping 60 ms or 120 is not so critical, for the eye it is not noticeable, of course, if only the web page does not contain a huge number of elements that require separate HTTP requests, that is, if the page is not optimized. We’ll probably talk about optimization in a separate article. But it’s worth remembering that most Internet pages have an average size of about 700 KB and load over a time of about 0.3-0.5 s (300-500 ms), which is why the difference of 60 ms will not be any significant for these cases . Much more important is the stability of the ping and the lack of losses.

There are several services that help you check availability from different regions, one of them, the most famous, host-tracker.com. Using the 2ip.ru service, you can find out what other sites are located on the IP address of the provider and already check the speed of loading pages on the host tracker from other clients, even before purchasing hosting services.

On overloaded servers or in the case of insufficiently good connectivity, a low page generation speed will be noticeable and, as a result, a longer time, as well as a change in this parameter during the day, so this test will easily determine the quality of the service before receiving any internal information about the hardware, uplinks and in general how and what is organized. The only thing to be careful with is sites that use external elements to generate the page, here the page generation speed may already depend not only on the web server and how the site and the web server itself are optimized for work, but also on external servers . Therefore, you should check the operation of several independent websites at once.

Many people nowadays believe that hosting has “died” and can agitate you to purchase VPS (virtual private server), because the price of the service is not much more expensive, and sometimes even cheaper than hosting. Nevertheless, this service has both pros and cons.

+:

  • Software or even hardware isolated environment (in the case of hardware virtualization);
  • root access, the ability to install specific software that is not accessible to hosting users;
  • the possibility of tuning the web server for a specific task, which speeds up page generation time and improves behavioral factors, as a result – your project is growing and you can always predict growth;
  • ease of migration, you can always transfer the project to another VPS or a dedicated server, saving all settings, mail and much more;
  • the ability to monitor and connect round-the-clock support from the provider, which will ensure a healthy sleep and save from hacking, prevent most problems and improve the site and search position by tuning and tracking work on the trashholds of many parameters.

-:

  • The amount of allocated resources may be less than the hosting of the “right” hosting provider, and the service, in the end, can be more expensive;
  • not every provider provides a server control panel for free (the same ISP Manager 5), as a result, the cost of the service increases, sometimes many times (CPanel lovers will be forced to pay for the panel several times more than the cost of VPS, which in itself is nonsense, but they buy !);
  • administration and around-the-clock support costs an average of $ 30-50 per month, and only a few can afford to offer it at a loss of a price of $ 10, which is comparable to or less than the cost of the initial VPS in order to develop their users in the future;
  • a dedicated IP address with no neighbors can be either a plus or a minus, some SEO followers believe that this is a problem and that Google still “fills” site grids only by IP address;
  • for a business card site or even a small portal VPS doesn’t need a nafig, this is an excess that only leads to the consumption of IPv4 addresses, which are already in short supply, web hosting is not dead!

The sequel will be very soon …

A bit of advertising 🙂

Thank you for staying with us. Do you like our articles? Want to see more interesting materials? Support us by placing an order or recommending to your friends, cloud VPS for developers from $ 4.99, A unique analogue of entry-level servers that was invented by us for you: The whole truth about VPS (KVM) E5-2697 v3 (6 Cores) 10GB DDR4 480GB SSD 1Gbps from $ 19 or how to divide the server correctly? (options are available with RAID1 and RAID10, up to 24 cores and up to 40GB DDR4).

Dell R730xd 2 times cheaper at the Equinix Tier IV data center in Amsterdam? Only here 2 x Intel TetraDeca-Core Xeon 2x E5-2697v3 2.6GHz 14C 64GB DDR4 4x960GB SSD 1Gbps 100 TV from $ 199 in the Netherlands! Dell R420 – 2x E5-2430 2.2Ghz 6C 128GB DDR3 2x960GB SSD 1Gbps 100TB – from $ 99! Read about How to Build Infrastructure Bldg. class using Dell R730xd E5-2650 v4 servers costing 9,000 euros for a penny?

Similar Posts

Leave a Reply