Hosting and dedicated servers: answer questions. Part 3

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 and not 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.



Hosting and dedicated servers: answer questions. Part 1
Hosting and dedicated servers: answer questions. Part 2. Why such an expensive Internet in the data center?

Please tell me which of the two servers is better to choose? On which sites will I run faster?

2x Intel Xeon E5620> 16C / 32T – 2.40 Ghz, 24 GB RAM, 2x 2 TB SATA Storage and 1Gbit Port

vs

Xeon E3-1220 v3> 4C / 4T – 3.10GHz, 16 GB RAM, 512 GB SSD Storage and 1Gbit Port

In order to answer your question fully, first of all you need to answer the question – what determines the speed of the website?

  • on how optimized the site code is;
  • how optimized the server software is;
  • on the quality of communication channels between the server and the visitor;
  • from server configuration.

As you can see, the configuration of iron is the last thing that can affect the speed of your sites.

Code optimization

There are many articles you can find on code optimization on the Internet. Useful found on Habré and in Medium. It was interesting to find 1998 article, where “to increase the speed of downloading network pages to a computer” could be due to increased data transfer efficiency between the browser and the web server. Moreover, the software cost not a little money:

If SpeedServer promises such a gain in speed, then why not get it soon? However, there are reasons to not rush, and one of them may be the price of this package. For nodes with access speed T1, it costs 75 thousand dollars, which, in our opinion, is quite expensive.

That was the time! And what are the amazing results of the software:

Armed with a stopwatch, we measured the speed of loading pages from three Web sites on which the Sitara accelerator was installed: sitara.net, jcrew.com and sportsnetwork.com. The gain in speed varied greatly from node to node.

The results obtained for jcrew.com were the best: its main page loaded without the SpeedSeeker accelerator in 16 seconds, and with it in just 6 seconds. Repeated visits to this page took 5 and 3 seconds, respectively.

The worst results were demonstrated by sportsnetwork.com, where the main page loaded without an accelerator in 45 seconds, and with an accelerator in 40 seconds. During our tests, the final games of the basketball championship were held, and the site was heavily overloaded. We were somewhat disappointed that the accelerator practically did not give a gain in download speed.

During the test period (they were carried out during the week), SpeedSeeker program reduced the average load time by half. Sometimes the gain in speed was not observed at all, and sometimes the speed of exchange with the Web tripled.

Regardless of which line (a modem with a speed of 28.8 kbit / s or T1 with a speed of 1.5 Mbit / s) your clients access the node, the SpeedServer package will increase the speed of their connection. You just have to decide whether to save a few seconds to lay out 75 thousand dollars.

How everything has changed in 22 years! Today we are able to use high-speed connection almost everywhere and the page load time is on average 0.3 seconds (300 ms) even for large pages, and the average web page size has long exceeded 750 KB. In this case, it is necessary to treat the loading time and page generation time. After all, the first consists of the time the pages were generated by the web server + the data transfer time to the visitor. And the first and second can be optimized. By reducing the number of HTTPS requests, the load time can also be optimized, since the HTTP protocol works similarly to the TCP / IP protocol when confirmation of successful confirmation of a data packet is expected, in this case, successful loading of a web page element.

Server Software Optimization

You can speed up the page generation time from the server side by switching the sites to a faster web server and fine tuning the PHP and MySQL parameters, of course, if this type of database is used. Recently, more and more solutions are appearing. We will not dwell on them in detail in this article – since this is the lot of the administrators who accompany the servers. And to describe these points, perhaps, you need to write a separate article.

It is important for webmasters to know that in order for your site to work in the most optimal way and as few neighbors as possible influence its work – it is better to place it on a dedicated server, or at least VPS. Of course, good hosting can turn out to be faster than VPS, due to the large allocated resources at the peak for your project, but here you should see how much your site consumes resources at the peak? Is this peak consumption on the hosting? Does the site need special web server or MySQL settings that are unattainable on the hosting and could speed up its work?

Communication channels and server placement

Server placement is much more important than its configuration. If the server is located in a data center with cheap communication channels, it may happen that from certain regions users will experience significant delays when working with your project. And this is not only about ping or loss on the way, but also about a limited band.

For example, Cogent sells cheap traffic, very cheap, several times cheaper than high-quality competitors. And if the speed from the Netherlands to Kiev and MSC is quite tolerable – over 10 Mbit / s per stream, then to Novosibirsk several times a speed of less than 1 Mbit / s was recorded (according to one of our customers who used this cheap connection in their cheap data center). If you have a streaming project – it matters. Part of the audience may experience a lack of bandwidth as soon as some accident or some network congestion occurs on the route.

And if an expensive data center, which has not only Cogent-clinics available, can solve the problem by redirecting traffic through a more expensive aplnik to a problem direction (thereby providing high-quality access for everyone, paying a small percentage of the audience for access, however the negative impact of this small% on behavioral factors and positions in the posikovoy system), in the cheap you only have to wait until the problem is resolved by Cogent, get a decrease in the position due to poor access quality for several% of the audience for whom connect an alternative, more expensive aplnik. But there was simply no opportunity. And in the end, you have already lost tens of percent of other visitors.

Server configuration

When choosing a server configuration, you need to look at what you really need. Perhaps you do not need so many cores and processors or RAM that are offered in these solutions, but you just need a good server with SSD, which has much less resources and optimization from the points discussed above. Some applications work well with large core frequencies and fewer ones (for example, 1C does not very efficiently parallelize requests), for some, on the contrary, more threads are important, albeit with a lower frequency.

Regarding both options, these are not good options at all. Both configurations are not good and I do not recommend them.

The first option does not have an SSD, the second option has only 1 drive, which is not very safe and fault-tolerant (you need to have an emergency recovery plan and evaluate whether or not downtime is valid). So this is just marketing. Unbalanced servers containing bottlenecks for your projects. You will not use resources efficiently and will get a few bottlenecks.

Of course, an SSD has a positive effect on page generation time because the HDD has a large delay due to rotating parts and only 70-100 random I / O operations per second. At the same time, latency for SSD drives is ten times lower, and SSD can withstand high loads of data input / output (reading and writing). So why not use them in your server? Nevertheless, if you optimize the server software, send the cache and frequent requests to RAM or SSD, good results can be obtained on the server with the HDD. The choice of the optimal configuration depends on many factors. And this is a topic for a separate article.

Can you tell me about which projects you need a dedicated server for? I will help you choose an economical solution for your case, and it will not be very expensive. Perhaps it will cost even cheaper than your options.

To be continued…

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? (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 c using Dell R730xd E5-2650 v4 servers costing 9,000 euros for a penny?

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *