1

Topic: Loading peaks

Kind day, the colleague! There is a service, is allocated on VDS a hosting. Sometimes there are sharp surges of loading from users, with excess of normal regular loading in times. Thus CPU it is hammered completely and the expenditure of storage of the server leaves in available limit and all starts to brake strongly. That that activity of users non-uniform is an inevitable reality so happens and it is necessary to find the decision how to process activity peaks. It is possible to hold certainly reserve VDS on such cases... But peaks can be not frequent (at all every day)... And because of rather small amount of peaks in a month to hold all capacities standing idle almost time... Sees unreasonable and unprofitable. It would be desirable to understand, whether it is possible and how the decision with dynamic provision  amounts of resources - kernels/memories at loading growth? For certain during an "overcast" epoch there are any decisions?.

2

Re: Loading peaks

Hello, MadHuman, you wrote: MH> It would be desirable to understand, whether it is possible and how the decision with dynamic provision  amounts of resources - kernels/memories at loading growth? MH> for certain during an "overcast" epoch there are any decisions?.  one from cases when   can be more favourable

3

Re: Loading peaks

Hello, Adnin, you wrote: A> Eto one from cases when   can be more favourable On money more favourably? And that I heard that clouds in few times more expensive.... <<RSDN@Home 1.0.0 alpha 5 rev. 0>>

4

Re: Loading peaks

Hello, Adnin, you wrote: A> Hello, MadHuman, you wrote: MH>> It would be desirable to understand, whether it is possible and how the decision with dynamic provision  amounts of resources - kernels/memories at loading growth? MH>> for certain during an "overcast" epoch there are any decisions?. A> Eto one from cases when   can be more favourable not absolutely clearly what exactly can be as a klaud-hosting and as there it works. As far as I know, technologies of dynamic magnification of kernels/memories at VDS are not present... I Know there are decisions at klaud-hostings when it is possible to add/clean quickly resources to , but it manually and is required reboot . It is probably possible and without reboot - but that that manually, it any more a variant. Also it is possible to assume a variant, VDS at once has big kol-in resources, but the essential part their most part of time is not used and not paid. But whether is such in the nature? I did not meet...

5

Re: Loading peaks

Hello, _Raz _, you wrote: _R _> On money it is more favourable? And that I heard that clouds in few times more expensive. Yes,  the hosting can be more favourable on money in such cases, as: - there are big peaks of loading. All time is necessary X servers, but from time to time loading increases and in  time is necessary X+Y servers. - you started to be untwisted and the audience all time increases, it is heavy to predict loading. The normal hosting should be cheaper, if at you + - predicted loading without big . But  all rather also it is necessary to look at specific loadings and rates.

6

Re: Loading peaks

Hello, MadHuman, you wrote: MH> it is not absolutely clear, what exactly can be as a klaud-hosting and as there it works. MH> as far as I know, technologies of dynamic magnification of kernels/memories at VDS are not present... MH> I Know there are decisions at klaud-hostings when it is possible to add/clean quickly resources to , but it manually and is required reboot . It is probably possible and without reboot - but that that manually, it any more a variant. MH> also it is possible to assume a variant, VDS at once has big kol-in resources, but the essential part their most part of time MH> is not used and not paid. But whether is such in the nature? I did not meet... Read that such vertical and horizontal scalability. Look at architecture highload systems, for example further: https://ruhighload.com/post/%D0%90%D1%8 … 0%BE%D0%BA In AWS there is concept EC2 -  the normal server. They can be lifted so much, how many it is necessary. Also there is concept Elastic Load Balancing https://aws.amazon.com/ru/elasticloadbalancing/ Service Elastic Load Balancing automatically arranges the entering traffic of applications between several  Amazon EC2. It gives the chance to reach fail safeties in applications, effectively selecting the volume of resources necessary for load distribution according to volume of the entering traffic of applications. And Auto Scaling: https://aws.amazon.com/ru/autoscaling/Function Auto Scaling allows to support availability of applications on invariable high level and automatically to scale resources Amazon EC2 both towards magnification, and towards reduction, depending on the conditions set by you. Function Auto Scaling also allows you to supervise start of a necessary amount  Amazon EC2. When it is necessary it simply lifts more servers.

7

Re: Loading peaks

Hello, Adnin, you wrote: A> Read that such vertical and horizontal scalability. Thanks for the information. The variant when a problem dares at the expense of system architecture complication (when it is entered , additional working machines on which  starts up loading) is clear. It is interesting, whether there are decisions on other way - when at "iron" level (more precisely the virtual machine), without complication of the program system. That "Auto Scaling" to resources of one virtual machine (a kernel, storage). Progress reached the such? PS. Interesting for a platform windows/.net

8

Re: Loading peaks

Hello, MadHuman, you wrote: Then you need to google for example "vertical auto scaling". Had no business with , but on  to a topic it is possible to decide that  it is real also what there are problems: https://www.linkedin.com/pulse/introduc … -omomhenle

9

Re: Loading peaks

Hello, Adnin, you wrote: A> Then you need to google for example "vertical auto scaling". A> had no business with , but on  to a topic it is possible to decide that  it is real also what there are problems: A> https://www.linkedin.com/pulse/introduc … -omomhenle thanks! Keywords appeared useful!