Get best of the week

As we tested the performance of new processors in the cloud for 1C according to the Gilyov test



We won’t discover America if we say that virtual machines on new processors are always more productive than equipment on old-generation processors. Another thing is more interesting: when analyzing the capabilities of systems that would seem to be very close in their technical characteristics, the result can be completely different. We were convinced of this when we tested Intel processors in our cloud to check which of them give the greatest return when operating systems on 1C.

Spoiler: as our test showed, it all depends on the task. We managed to select the product from the entire line of new Intel processors that gave a multiple increase in performance due to the fact that Intel Xeon Gold 6244 has fewer cores, each core has more L3 cache memory and a higher clock frequency is assigned - both the base and and in turbo boost mode. In other words, it is these processors that better cope with resource-intensive tasks in terms of a unit of productivity / ruble. For 1C, this fits the best: with new processors, 1C applications in our cloud began to literally “breathe”.

Now we’ll tell you how we tested. Below are the results of Gilev synthetic tests. You can focus on them, but in any case, you need to check the actual disposal yourself on your tasks.

Test conditions


An important note: we did the comparison without any additional optimizations, not a benchmark. With additional configuration of systems in the cloud, the results will be guaranteed better.

Given: two virtual machines with 8 vCPU and 64 GB RAM with FLASH 10,000 IOPS.

The first virtual machine - with Windows Server 2016 and installed 1C 8.3.10.2580, for the second image of a virtual machine with a database (Centos + Postgresql) was taken on Gilev.ru .

The Postgresql database is not accidental, since its operation is closest to the actual conditions of use of 1C by our customers. Yes, yes, we did synthetic tests similar to typical installations, that is, this is not a universal answer to all the questions of the Universe, namely, a guideline for your own analysis.

The important thing is that if you use file architecture instead of a database, test results are usually higher. But in reality, this type of architecture is used only for very small installations. Here RuVDS tested on file architecture. And here is what Vyacheslav Gilyov himself said in a comment about this :
If we are talking about renting 1C in file mode, then yes, but what catches my eye works exclusively in the client-server version. It makes sense: 1) or to make this clarification in the article; 2) or test the client-server version, because the difference in architecture is significant, and the file version does not have full functionality.
No additional settings for the operating system and 1C product were made.

Processors


  • In the left corner of the ring is the Intel Xeon E5-2690 v2 processor, 3.00 GHz.
  • In the right corner of the ring is Intel Xeon Gold 6254, 3.10 GHz.
  • In the center of the ring is Intel Xeon Gold 6244, 3.60 GHz.

Let the battle begin!

results


Intel Xeon E5-2690 v2, 3.00 GHz:


“Good” for us - the minimum mark that guarantees a comfortable level of customer experience with 1C systems.

The result is 22.03.
 
Intel Xeon Gold 6254, 3.10 GHz:



The result is 27.62.  

Intel Xeon Gold Processor 6244, 3.60 GHz:



Result - 35.21.

Total: even if the virtual machine on Intel Xeon Gold 6244 at 3.6 GHz will cost 60% more than the E5-2690 v2 at 3 GHz, then you should choose it. With a smaller difference in price, the benefits become even greater. But we have a much smaller price gap, so such VMs are noticeably more profitable.

Cascade Lake processor cores demonstrate performance gains not only due to the increased frequency, but also more modern architecture. Moreover, different processor models from this line give different results, which must be taken into account when solving your problem.

In the cloud, we plan to use these processors in Turbo Boost mode, at which the processor clock frequency reaches 4.40 GHz, which will increase its performance gap and make the choice in favor of this product even more obvious.

What does this mean for us


For a long time we lived in the old paradigm when one processor didn’t have very many cores, and therefore not many virtual machines were placed on one server. I had to squat a lot in order to achieve at least some optimality in the tight packing of VMs in these servers. Now, when we get 28 or even 56 cores per socket, the problem with the density of stacking is solved almost by itself. And we have resources to think about other goodies for customers of our CROC Cloud. For example, we filmed a separate pool with 6244 processors under the DBMS.

An additional bonus - all this turned out to be a very suitable architecture for 1C. The point is that if you switch from a 3 GHz processor to a 4 GHz processor, then almost all the tests give you not +30%, but + 15–20% ... And this thing gives you +45%. That is, the frequency increases by 30%, and the increase grows nonlinearly to the frequency. And processors are 40 percent more expensive. As a result, new processors are more expensive, but finally 1C starts working normally. You can go to the cloud without worrying that the wrong processors are there. For many of our customers this is now very important.

More articles:


All Articles