"Crooked" server spoils the entire data centers! What to remember when choosing server memory

Hello, Gictimes ! Non-specialists today are sure that data centers are such “reserves of mighty servers”, in which computers differ from home computers only in performance and some type plates / connectors. And if so, then the upgrade of server computers, it turns out, is an ordinary task? In reality, things are very different, and today we will tell you how many nuances you have to consider even when choosing RAM for data centers.

Popularity is a windy lady, especially when we talk about devices for end users. Even if you pay attention to the means of access to the Internet, it is easy to see how desktops and laptops first gave way to the share of tablets (the “post-PC era!”), And then these same tablet PCs were almost completely annihilated, because their place was occupied by smartphones with large screen sizes. The latter became the new kings of the mountain, with much more Internet activity than all desktops and hybrid PCs taken together. Tomorrow manufacturers will “polish” new ideas in smart clocks or folding tablets – and the pendulum will again swing in the opposite direction.

But with the data processing centers, things are much more unambiguous – the number of Internet users is growing, the content “is getting heavy” , Cloud services are added to the popularity, the load on server hardware is increasing. The data centers themselves become more and more every year – the Internet giants switch from the standard versions to the craft with their own projected data centers (data centers) with equipment, cooling and personnel to their taste. But the data centers “turnkey” are experiencing their boom in Russia – for political reasons, and nevertheless.

We will be realists – the construction of data centers “from scratch” in the open field today is mainly engaged in giant companies, Much more often companies resort to leasing computing power or deploy their own configured server in leased areas. Today we will mention important trifles that need to be considered when choosing the server configuration – we understand what needs to be taken into account at the iron and organizational levels when selecting memory for the server.

Why the server is not a home PC “on steroids”

Since mainstream servers have acquired x86-based processors, they have become, as they say, closer to the people and often experience a second youth in gaming enthusiast computers. Not because they are ideal for gaming tasks, but because obsolete server hardware is sometimes cheaper than the gaming components of “home” computers.

Data centers are not only “Its own atmosphere,” but also iron with completely different from gaming and office tasks

With server computers, server hardware shares only high performance and standardized “family signs” of memory. Only in the server, if we are talking about memory, the memory is not short-term peak loads and simple most of the time, but work at the limit with a huge number of requests and operations per second. At the same time, the difference in energy consumption between several types of memory is not “savings on matches”, as gamers call it, but tangible expenditures of the organization’s budget. And the cooling of the components is also designed for “marathons”, and not according to the principle “it’s not throttling in my scenario of use – that’s good.”

Finally, excessive reliability (it sounds strange, but in everyday scenarios of work morale will come earlier than the server hardware resource is exhausted) and memory capacity is not a question of prestige in any form, but a consequence of constantly high loads and consumption RAM, for example, SQL databases, virtualization and, especially, in-memory databases, in which RAM is the main data store.
From the above, it may not be completely understandable why large companies prefer not to contact the purchase of extremely cheap memory with “seemingly normal chips” with the maximum capacity reserve and prefer “loud brands”. A few reasons:

Stress test of Kingston’s memory before its release for sale

Fault tolerance is the most important

If you can still take a risk in your home computer, buy a “noime on normal chips” and, in the worst case, get disappointed with BSOD in the game or when working in the browser and the amount spent on the purchase, then in a situation where the server “fell”, the company starts losing money Immediately (especially if the server is running for e-commerce needs). And in such a situation, no one will console Vasya if he saved his memory at a bargain price – even if he quickly replaces memory, “lifts up” the server and “gets on”, wasted money for low-quality RAM, the hassle of technical support staff and an unscheduled “hole” In attendance, nothing good is promised.

For this reason, the memory of brands is almost always used in the data center, which not only “pastes chips on a printed circuit board”, but controls the creation of memory and checks the products for availability troubleshooting, pre de than to send it on sale. Kingston, for example, manages the entire production cycle of RAM – from the plate to the chips, and employs 35 quality tests (including a 24-hour stress test at 100 degrees Celsius at increased voltages). And only after successful passage of all tests modules are sent to the hands of buyers. Such testing is not paranoia and not marketing, but a harsh necessity, because in this way Kingston guarantees that, for example, every cell (and there are more than 136 billion in modules) will be intact in a 16-gigabyte module.

A study of scientists from the University of Toronto proves that the failure of the module does not pass without a trace, even if the ECC corrected the error

There is no “coolest type of memory” – there are suitable and Not suitable for equipment varieties

In modern servers, DDR3 and DDR4 memory is used, but not all in a row, but of the following varieties:

• Unbuffered ECC DIMM – what the experts call the word UDIMM. Typical memory modules up to 8 GB each – the most budget option in servers and workstations. From the “simple RAM” for home PCs, there is an ECC scheme that detects and corrects errors in individual bits of memory.
• Registered DIMM is the same register memory that, de facto, became a visiting card of servers and workstations of the middle class. The most popular option in a wide retail. Modules of up to 32 GB each, slightly slower than non-registered, but much more reliable. It always supports ECC and is more practical than Unbuffered, because it scales much better (number of modules per channel).
• Load Reduced DIMM or LRDIMM – “the latest squeak” (2013 and newer) in server platforms. Compatible with RDIMM connectors, but it’s different – the modules have a memory buffer for transferring data from the host to DRAM, that is, LRDIMM takes over some of the functions of the memory controller, and the controller operates in a serial mode. By reducing the electrical load, the memory runs either faster than RDIMM with a higher clock speed, or at a much higher capacity than the predecessors at the same frequency. The monitoring tools have also become better.

Register memory Kingston DDR3 KVR-16R11S4 8HA on the Ivy Bridge-EP platform

The most appreciated for the opportunity to pile up an impressive (9-12 modules per processor) number of modules and thus achieve the maximum amount of RAM in the server. ]

The general principles of the selection are similar to those we remember on a PC for home use – sets of the same memory are always faster and more stable than the “vinaigrette” from different in the characteristics of the modules, in the installation of modules with voltage 1.35V and 1.5V voltage on All modules are aligned in favor of 1 .5V. In multiprocessor systems, the memory needs to be configured in the same way. The maximum speed depends on the processor, because the memory controllers are now located just under the CPU cover.

• NVDIMM – a mixture of a horror and a hedgehog of familiar RAM and SSD, Hybrid nonvolatile memory. It is installed in standard DIMM connectors and is compatible, but comes with an independent power source. Do not lose data in the event of a power failure. While not too popular, but already used in cloud computing.

Types of memory modules for servers

Data centers can be configured without support, but sadly

Even in an ideal world with “spherical in a vacuum” platforms and memory modules incompatibilities or bugs occur. Or the company works with branded equipment, in which conflicts the unclean foot will break. And someone just wants to minimize the headache and as little as possible to delve into the “pitfalls” and cases of unstable memory work, without overpaying for individual consultations.
Moreover, the creators of branded servers have a glorious tradition of “rip-off” their customers for maintenance and accessories after the end of the guarantee for the equipment. Fortunately, the server hardware has not yet spread vendor lock-in, which means it’s never too late to change the authorized service with the logic “do not get anywhere and pay as much as you say” to independent component manufacturers.

And even the level of service after such castling will only be higher. For example, because Kingston, unlike most competitors, by the way, has a full-fledged representation in Russia. All guarantee issues are resolved without the slightest stress, and buyers in principle will not have a characteristic situation for communicating with distributors of the situation “I just posted an announcement, for technical advice, please go to those guys in China.”

Any problems with compatibility and memory installation can be overcome quickly and easily. And you can check the compatibility and speed of memory not only with the help of documentation and the collective intelligence on the Internet / phone support (by the way, call 8-800-700-13-50, on weekdays 10:00 to 18:00), but also ” Profit “! Kingston will issue a set of memory for the test before buying, and if it does not “start”, the buyer is guaranteed a refund. Are you skeptical or not, but it’s hard to name even one more vendor in the Russian Federation with the same freedom in selecting server hardware.

Selection The optimal memory for the server starts with the configurator. For everything else, there is the official Russian representative office of Kingston and a huge number of distributors

In addition, with “field control” Kingston delivers memory to customers, in most cases faster than competitors, and a discount on iron when ordering large lots or Tenders can be obtained without problems – this is also a noticeable difference from the practice “I’m a realizer, not the person who determines the price policy, buy it as it is.”

There are a lot of nuances in the server memory that are impossible to cover it in one article, but the preamble to – time, and hardcore – hour. Stay tuned – we still have something to talk about the specifics of server hardware

For more information on Kingston HyperX products and visit the official website!. In choosing their HyperX kit help page with a visual aid.