Replacing RISC-machines for x86: I invite you to touch and test tomorrow

The RISC-architecture for quite a long time was uncontested for processing banks and other responsible high-load. As soon as the crisis began, the opinion began to sound more and more often that x86 could also cope. Now x86 are actively probing and in some places are being implemented. Because savings at times.

Tomorrow, on Thursday, October 27, at 18:00, we will mock the x86-machine with the intriguing name Huawei Kunlun. Come poke it with a wand. We will load it with Oracle and bitcoin mining until it, like a chainsaw from a joke, says “Hrrr”.

Below is a short FAQ about RISC machines and a test drive program.

- Why do we need RISC platforms?
')
Most often - for processing of banks, billing of operators, miscalculations of insurance, clearing in exchange companies, etc. In retail and medicine, we have only seen them a couple of times. The meaning of a RISC machine is the ability to process a lot of data streams (fairly small operations) at a time. Very simple, compared to the x86 architecture, lags on core management do not accumulate, due to the fact that RISC cores were originally designed for multi-threaded operation, and x86 cores have only evolved to that relatively recently.

- So why not jump from RISC to x86, if this is the same, and yes even at times cheaper?

First, the solutions are new, they are all less than a year old. In Russia, there is only one test wort (and another system, not Kunlun). There is nowhere to look. Secondly, in spite of the hellish budgets, it is rather difficult to take and transfer the processing of the bank, risking to screw the whole bank. In the bank as: did not close the banking day on time three times in a row - you immediately and without questions from Khan. Well and, thirdly, even despite the latest instruction packs that are rapidly approaching the x86 to the RISC level, there is still an ambush in performance and a number of several unpleasant effects on large loads. This is what we will talk about and help you understand whether this will affect your tasks or not.

- Why are RISC so expensive?

Because these are machines that are made far from those circulations that are common universal x86 cores. And their price is not only in purchase (rather expensive), but also in maintenance (for 3 years - like buying a new one). Plus a question of qualification and training of the administrator.

- What will be on the test drive?

Until 19:00 - registration of participants, a small buffet. You can come at 18:40, just enough time to go through the security and register. But the frequent guests of our events come earlier, because the party and the exchange of live experience from colleagues. Without embellishment, everything is as it is.

19: 00-20: 00 - the main part: test drive. Conducted by our expert department of computing systems Anton Semchishen. There will be Valentin Gubarev, director of the Computing Systems Department, Artur Pärn, director of IT solutions from Huawei, Dmitry Kornev, head of business development with Intel system integrators.

20: 0-21: 00 - entertainment part, free communication. Small snacks, talk "for life", finding out the true size of the discount on the supply from the voiced price list and other delights.

- And the technical part?

In the left corner of the ring is a RISC machine, in the right is a cluster of Huawei Kunlun. An Oracle DBMS is installed on both servers and databases are created. We will run the Swingbench test suite on each server - this is the load generator for the DBMS, which can emulate the load from different applications. Run the tests with the most typical settings, including the selection, modification, insertion and deletion of database elements, but these tests can be configured to best simulate virtually any application. Then we will start the process of bitcoin mining and see how many gigageshas per second can calculate each server.

- Ok, what do you need to know about this Kunlun?

Here is the technical specification , and here is the architecture, the rest of the details are in place:



- For whom this test drive?

For CIO, chief admins, technical directors, development directors, heads of technical departments and operations departments of large companies. We will talk about technical details and financial terms. This time there will be no spectacle with server-side bites and emulation of crashes, but we’ll still be fun, we promise.

- Where to come and how to register?

You should call or write to Svetlana Eremeeva, SEremeeva@croc.ru, +7 (495) 974-22-74, additional 3408.
Or here is a link to the registration form.
The number of seats is limited, and we reserve the right to make a final decision on the confirmation of registration in each specific case. Place - LOFT # 2 Moscow, Leninskaya Sloboda, 26, p. 11.