Post - a question
Greetings.
On duty, sometimes you have to evaluate third-party developments, people, and so on. Today I was in a large office, a market leader in my field. Usually my task in such events is like finding a needle in a haystack - I'm looking for costs, holes in budgets and extra employees. It is quite difficult to find flaws in the systems that sane, intelligent people raise and lead. However, today was the exception. From the threshold, the technical manager began to say that their bottleneck is RnR, i.e. backend I clarified whether this is really not a base. However, looking at the htop output (linux on board) really saw that 2 servers (6 gigs of opera and 2 xeonas of the latest generation with 4 cores) are bent under the “litigation” of the project with an attendance of 30k hosts per day and 300k hits per day . And the bottleneck is really RnR, judged by htop. Well, or curve hands of the developer.
Actually, therefore, questions to you, dear colleagues, because I have a space on the part of RnR:
1) Is RnR really capable of eating so muchsoftware processor time, or is it possible to look for bottlenecks in software? And if you look, then what? Are there any debugs?
2) There is a problem, but it must be solved quickly, by December. Are there any accelerators for ruby? Keshi?
3) As a solution to the problem, their technical manager proposed to buy an expensive server 100k + rubles for its further clustering with others. I suggest for the same 100k to buy 4 weak servers, throw RnR on them, Sursen, and load balance. On the one hand, this is an overpayment for rack space, on the other hand, it is clearly better than one powerful server. From my point of view.
I would be very grateful for any comments. with RnR, unfortunately or fortunately, I haven’t yet had any deep business.
On duty, sometimes you have to evaluate third-party developments, people, and so on. Today I was in a large office, a market leader in my field. Usually my task in such events is like finding a needle in a haystack - I'm looking for costs, holes in budgets and extra employees. It is quite difficult to find flaws in the systems that sane, intelligent people raise and lead. However, today was the exception. From the threshold, the technical manager began to say that their bottleneck is RnR, i.e. backend I clarified whether this is really not a base. However, looking at the htop output (linux on board) really saw that 2 servers (6 gigs of opera and 2 xeonas of the latest generation with 4 cores) are bent under the “litigation” of the project with an attendance of 30k hosts per day and 300k hits per day . And the bottleneck is really RnR, judged by htop. Well, or curve hands of the developer.
Actually, therefore, questions to you, dear colleagues, because I have a space on the part of RnR:
1) Is RnR really capable of eating so much
2) There is a problem, but it must be solved quickly, by December. Are there any accelerators for ruby? Keshi?
3) As a solution to the problem, their technical manager proposed to buy an expensive server 100k + rubles for its further clustering with others. I suggest for the same 100k to buy 4 weak servers, throw RnR on them, Sursen, and load balance. On the one hand, this is an overpayment for rack space, on the other hand, it is clearly better than one powerful server. From my point of view.
I would be very grateful for any comments. with RnR, unfortunately or fortunately, I haven’t yet had any deep business.
')
Source: https://habr.com/ru/post/75926/
All Articles