Yandex.Metrica or specialized monitoring system - what and when to choose?
No matter how much cost is spent on the development, design and promotion of the site - all these efforts are in vain if the site for some reason suddenly becomes inaccessible to customers. Everyone solves this problem in his own way: someone out of habit is constructing structures of varying degrees of efficiency on his knee, and someone is looking for ready-made solutions on the network. One of the most popular solutions in Runet is the monitoring function in Yandex.Metrica. Another option is specialized monitoring systems. This post is devoted to the study of their differences, as well as the experience of practical application. By a completely coincidental coincidence, for comparison, the HostTracker site monitoring service was chosen.
Most importantly - the site is checked. This is what a person understands, even without going into details. Somehow statistics are collected on the availability of the site and alerts come when the site is broken. Even configured in a similar way:
If you delve into the settings - then, of course, it is clear that the tools are somewhat different directions. The metric studies customer behavior more, and HostTracker the technical aspects of site operation. But this function seems to be the same.
')
So what is the difference? As we know, to configure Yandex. Metrics, you need to embed some javascript code into your page. Thanks to him, Metric will receive some information about each client. On its basis, conclusions are drawn about traffic and site use scenarios. Also, indirectly, it is possible to draw conclusions about accessibility: if there are no customers longer than usual - perhaps this is a cause for concern.
HostTracker offers another mechanism - monitoring the site by simulating real users who regularly visit the site. However, Metrics has a similar mechanism: checking with special bots of Yandex. With that, it is configured both from the Metrics and from Direct . But little is known about these bots. Their number and addresses are not disclosed anywhere (it is directly stated that this is a secret), and the check interval remains secret. The report indicates the average interval. And although the description says that monitoring works effectively with traffic above 100 views per week (which hints that the main tool is still the script executed on the page), practice shows that the online store is “missing” from 1000 visits per day is only seen after an hour or more. The site, where 200 views per day, lay 3 hours before the alert came from Metrics. In both cases - alerts come only with a continuous idle site. If it “slows down”, or it appears / disappears (for example, due to increased server load), then the alert may be completely absent. Also, no alerts do not come if the page loads, but the connection with the database, file server and other elements necessary for the site to work is broken.
At the same time, active monitoring by compulsory generation of queries, practiced by HostTreker, does not divide sites by the number of visitors, and at the selected monitoring interval, say, 10 minutes, it checks the site every 10 minutes, regardless of other factors.
Specific comparison:
As you can see, the metric for the year noticed only one fall. For the same site, HostTracker has only noticed 4 in the last month. The reason is a timeout, that is, clients really cannot get to the site or experience discomfort. It simply ignores the metric, either because of its short duration, or because someone manages to get to the site despite the big delay and, thus, the Yandex script does work.
This publication was born as an answer to repeated questions (including in the comments on Habré) about the differences in the work of HostTreker and other similar and not very systems.
Convenience Metrics - "all in one." This is also its flaw. It seems that the developers of the monitoring function were understood as a pleasant bonus, but far from the main task of the Metrics (in fact, this is not hidden). It is accurate with a sufficiently large number of visitors (from practice - several thousand per day) and solves the basic task: notify if the client for some reason cannot go to the site. And turn off with this Direct, saving advertising funds. However, in practice, such savings do not always work.
HostTracker is accurate regardless of the number of visitors and has many additional functions, as well as features for basic diagnostics: monitoring using different protocols, checking content, and so on. Pleasant buns are reminders of the expiration time of the domain and the site certificate . The algorithm is clear and transparent. There is also a minus - you have to pay for most of the functions.
Quite a few of our clients use two or three systems, both paid and free. The results of the work do not always coincide. We will find out about this, as in such cases, as a rule, clients are asked to explain the differences in indications. Often this leads to the identification of hidden bugs and "features" of the site, hosting, or the work of a monitoring system, which reduces the chances of blinking the real problem to almost zero.
As you can see, each task has its own tool.
Just yesterday, a client wrote to us about problems with Google Adwords. Google also has a mechanism for checking sites that host ads. But the problem is the same as in Metric - for the time interval between the fall of the site and the next test of the Google bot, a real amount of money is burned. Therefore, we received a request for integration with the Google API, in order to realize the ability to automatically disable advertising when a site crashes. Most likely, soon this function will appear - then we will write about it in more detail. This case perfectly demonstrates the principle of developing our service - we solve specific problems of specific people. From the very beginning . Therefore, we will be grateful for the criticism and suggestions.
What common?
Most importantly - the site is checked. This is what a person understands, even without going into details. Somehow statistics are collected on the availability of the site and alerts come when the site is broken. Even configured in a similar way:
If you delve into the settings - then, of course, it is clear that the tools are somewhat different directions. The metric studies customer behavior more, and HostTracker the technical aspects of site operation. But this function seems to be the same.
')
Theory vs Practice
So what is the difference? As we know, to configure Yandex. Metrics, you need to embed some javascript code into your page. Thanks to him, Metric will receive some information about each client. On its basis, conclusions are drawn about traffic and site use scenarios. Also, indirectly, it is possible to draw conclusions about accessibility: if there are no customers longer than usual - perhaps this is a cause for concern.
HostTracker offers another mechanism - monitoring the site by simulating real users who regularly visit the site. However, Metrics has a similar mechanism: checking with special bots of Yandex. With that, it is configured both from the Metrics and from Direct . But little is known about these bots. Their number and addresses are not disclosed anywhere (it is directly stated that this is a secret), and the check interval remains secret. The report indicates the average interval. And although the description says that monitoring works effectively with traffic above 100 views per week (which hints that the main tool is still the script executed on the page), practice shows that the online store is “missing” from 1000 visits per day is only seen after an hour or more. The site, where 200 views per day, lay 3 hours before the alert came from Metrics. In both cases - alerts come only with a continuous idle site. If it “slows down”, or it appears / disappears (for example, due to increased server load), then the alert may be completely absent. Also, no alerts do not come if the page loads, but the connection with the database, file server and other elements necessary for the site to work is broken.
At the same time, active monitoring by compulsory generation of queries, practiced by HostTreker, does not divide sites by the number of visitors, and at the selected monitoring interval, say, 10 minutes, it checks the site every 10 minutes, regardless of other factors.
Specific comparison:
As you can see, the metric for the year noticed only one fall. For the same site, HostTracker has only noticed 4 in the last month. The reason is a timeout, that is, clients really cannot get to the site or experience discomfort. It simply ignores the metric, either because of its short duration, or because someone manages to get to the site despite the big delay and, thus, the Yandex script does work.
What do I need?
This publication was born as an answer to repeated questions (including in the comments on Habré) about the differences in the work of HostTreker and other similar and not very systems.
Convenience Metrics - "all in one." This is also its flaw. It seems that the developers of the monitoring function were understood as a pleasant bonus, but far from the main task of the Metrics (in fact, this is not hidden). It is accurate with a sufficiently large number of visitors (from practice - several thousand per day) and solves the basic task: notify if the client for some reason cannot go to the site. And turn off with this Direct, saving advertising funds. However, in practice, such savings do not always work.
HostTracker is accurate regardless of the number of visitors and has many additional functions, as well as features for basic diagnostics: monitoring using different protocols, checking content, and so on. Pleasant buns are reminders of the expiration time of the domain and the site certificate . The algorithm is clear and transparent. There is also a minus - you have to pay for most of the functions.
Quite a few of our clients use two or three systems, both paid and free. The results of the work do not always coincide. We will find out about this, as in such cases, as a rule, clients are asked to explain the differences in indications. Often this leads to the identification of hidden bugs and "features" of the site, hosting, or the work of a monitoring system, which reduces the chances of blinking the real problem to almost zero.
As you can see, each task has its own tool.
News from the fields
Just yesterday, a client wrote to us about problems with Google Adwords. Google also has a mechanism for checking sites that host ads. But the problem is the same as in Metric - for the time interval between the fall of the site and the next test of the Google bot, a real amount of money is burned. Therefore, we received a request for integration with the Google API, in order to realize the ability to automatically disable advertising when a site crashes. Most likely, soon this function will appear - then we will write about it in more detail. This case perfectly demonstrates the principle of developing our service - we solve specific problems of specific people. From the very beginning . Therefore, we will be grateful for the criticism and suggestions.
Source: https://habr.com/ru/post/317468/
All Articles