26

Re: Azure - how to avoid a bog?

Hello, Cyberax, you wrote: AK>> And how to make 1  the RAM by the physical machine? Already there is such iron? The C> Is. Cheaply, only $10 at an o'clock at Amazona. In  x1.32xlarge costs 2 to storage and 128CPU. On their spots it is possible to buy more cheaply sometimes. The question was about the physical machine, not cloudy VM.

27

Re: Azure - how to avoid a bog?

Hello, Artem Korneev, you wrote: AK> AK>> And how to make 1  the RAM by the physical machine? Already there is such iron? AK> the C> Is. Cheaply, only $10 at an o'clock at Amazona. In  x1.32xlarge costs 2 to storage and 128CPU. On their spots it is possible to buy more cheaply sometimes. AK> the question was about the physical machine, not cloudy VM. And what difference? One or more VM are by one physical machine so the answer to your question - yes, is such iron.  on Yandex. A disk

28

Re: Azure - how to avoid a bog?

Hello, Artem Korneev, you wrote: AK>>> And how to make 1  the RAM by the physical machine? Already there is such iron? The C>> Is. Cheaply, only $10 at an o'clock at Amazona. In  x1.32xlarge costs 2 to storage and 128CPU. On their spots it is possible to buy more cheaply sometimes. AK> the question was about the physical machine, not cloudy VM. So it is the real physical machine, simply access to it through Amazon. Even I will tell in secret,  type x1.32xlarge all work on selected iron so on a physical piece of iron of anybody another will not be.

29

Re: Azure - how to avoid a bog?

Hello, turbocode, you wrote: a C>> And it is already physically unreal. Nobody cancelled a velocity of light -  between different machines always will be more slowly, than local access. T> all that I wanted it GRID at operating system level that it was possible to fulfill monoliths on GRID and in that case it would be possible to launch SQLServer on 10000 processors and 100TB storages and all would be not necessary these noSQL. It is impossible basically. Without a difference on what to do - from physical restrictions not to leave. Well that is, to launch it it will be launched, but here because of overhead projectors on lock will work more slowly, than on IBM PC XT. That worked faster - it is necessary to arrange at logic level, and there already and the CAP-theorem comes.

30

Re: Azure - how to avoid a bog?

C> It is impossible basically. Without a difference on what to do - from physical restrictions not to leave. From what restrictions?> Well that is to launch a C it will be launched, but here because of overhead projectors on lock will work more slowly, than on IBM PC XT. That worked faster - it is necessary to arrange at logic level, and there already and the CAP-theorem comes. Now SQLServer works on 8 kernels why cannot work on 100000 kernels?

31

Re: Azure - how to avoid a bog?

Hello, turbocode, you wrote: T> Now SQLServer works on 8 kernels why cannot work on 100000 kernels? . It is not pleasant - to Brjuer.

32

Re: Azure - how to avoid a bog?

T>> Now SQLServer works on 8 kernels why cannot work on 100000 kernels? S> Amdal. It is not pleasant - to Brjuer. Did not see vital issues because of which SQLServer could not work on 100000 kernels.

33

Re: Azure - how to avoid a bog?

Hello, turbocode, you wrote: a C>> It is impossible basically. Without a difference on what to do - from physical restrictions not to leave. T> from what restrictions? A velocity of light and throughput of communication channels. On frequency 2 for one clock period light transits approximately 15 centimeters (one luminous nanosecond is 30 centimeters). If machines are allocated for 10 meters from each other during passage of a signal to one side transits about 30 clock periods (really even more). Throughput is restricted by frequency of light which transits through an optical fiber and above an ultraviolet not to rise. So  between nodes it is impossible to scale infinitely.>> Well that is to launch a C it will be launched, but here because of overhead projectors on lock will work more slowly, than on IBM PC XT. That worked faster - it is necessary to arrange at logic level, and there already and the CAP-theorem comes. T> now SQLServer works on 8 kernels why cannot work on 100000 kernels? Because it is scaled not linearly. So since any moment adding of kernels will do it MORE SLOWLY. And this "any moment" comes even to 128 kernels. With it it is possible to struggle, but on an output this or that form NoSQL inevitably turns out. So to wave a magic wand and to receive at once ACID on 100500 nodes and with the instant operation - it does not turn out.

34

Re: Azure - how to avoid a bog?

C> Throughput is restricted by frequency of light which transits through an optical fiber and above an ultraviolet not to rise. So  between nodes it is impossible to scale infinitely. Certainly to send on one ASM to a command through a network and to look forward to hearing any velocity of light does not help, and here if to send  with whole subroutine that it does not load a network - but it should work at operating system level transparently. T>> now SQLServer works on 8 kernels why cannot work on 100000 kernels? The C> Because it is scaled not linearly. So since any moment adding of kernels will do it MORE SLOWLY. And this "any moment" comes even to 128 kernels. Requests  it is good, especially on reading. What there such that can decelerate on 128 kernels? A C> With it it is possible to struggle, but on an output this or that form NoSQL inevitably turns out. So to wave a magic wand and to receive at once ACID on 100500 nodes and with the instant operation - it does not turn out. As though now the instant operation at SQLServer, anybody also does not calculate for it.

35

Re: Azure - how to avoid a bog?

Hello, turbocode, you wrote: a C>> Throughput is restricted by frequency of light which transits through an optical fiber and above an ultraviolet not to rise. So  between nodes it is impossible to scale infinitely. T> certainly to send on one ASM to a command through a network and to look forward to hearing any velocity of light does not help, and here if to send  with whole subroutine that it does not load a network - but it should work at operating system level transparently. And to storage as this flow will address? The C>> Because it is scaled not linearly. So since any moment adding of kernels will do it MORE SLOWLY. And this "any moment" comes even to 128 kernels. T> requests  it is good, especially on reading. What there such that can decelerate on 128 kernels? Lock to check up that anybody does not write another to this moment. Well and even requests about reading  it is bad, if all data is not located on one node.

36

Re: Azure - how to avoid a bog?

C> Lock to check up that anybody does not write another to this moment. Well and even requests about reading  it is bad, if all data is not located on one node. Probably I not correctly expressed the thought but I not against, if inside SQLServer is noSQL (either MapReduce or GRID) provided that exterior API for the ultimate user will look as normal SQL.

37

Re: Azure - how to avoid a bog?

Hello, turbocode, you wrote: a C>> Lock to check up that anybody does not write another to this moment. Well and even requests about reading  it is bad, if all data is not located on one node. T> I am possible not correctly expressed the thought but I not against, if inside SQLServer is noSQL (either MapReduce or GRID) provided that exterior API for the ultimate user will look as normal SQL. And . The infinite scaling SQL with arbitrary join' is impossible, as well as ACID-semantics scaling is impossible.

38

Re: Azure - how to avoid a bog?

C> And . The infinite scaling SQL with arbitrary join' is impossible, as well as ACID-semantics scaling is impossible. Tables and  it only abstraction (and the abstraction cannot be impossible) and as this scaling will be implemented inside to me  - if only worked as normal SQL.

39

Re: Azure - how to avoid a bog?

Hello, Cyberax, you wrote: a C> And . The infinite scaling SQL with arbitrary join' is impossible, as well as ACID-semantics scaling is impossible. Do not argue with the troll, it is useless. The character not behind knowledge here came.

40

Re: Azure - how to avoid a bog?

Hello, _ABC _, you wrote: K>> I it to what: losing control over executing _AB environment> In general, on productivity and reliability they give certain warranties. A question in money. Unless I about warranties? I about control. We tell, if I accurately know that is twisted on the server, I can in addition  easy services for the same cost. Or to deliver firewall for this server where the data for the filter will undertake with reliable  other computer. Who gives such flexibility in a cloud? K>> I.e. in this case we worsen quality of system _AB> Well... I would tell that scenarios different happen. I imagine scenarios when the cloud is more convenient and more favourable. _AB> However, they do not concern my sphere in any way meanwhile. Here! And in my area too there are no hyperrequirements. Roughly speaking, average business (which majority) easily manages the homebrew administrator and 5-10 servers with accurately controlled configuration. The cloud it is simple promises that IF suddenly tomorrow all Bangladesh wants at you , your service does not fall. It is necessary to me?