Dark Reading is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them.Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Vulnerabilities / Threats //

Insider Threats

11:32 AM
George Crump
George Crump

The Death Of DRAM?

SSD increasingly will be used to replace DRAM in everything from laptops to servers. Here's why that makes sense.

As I flew to the Flash Memory Summit Tuesday, I was thinking about predictions for what the event will mean for the average data center manager. One of the top trends that I think we will see over the next year is the replacement of DRAM with flash-based solid-state drives (SSD) in everything from laptops to servers.

I know SSD is supposed to replace storage, so why start picking on DRAM? Because DRAM is an expensive resource that is hard to install in quantity.

The Laptop Case: The number one reason to consider flash SSD in a laptop, as we have seen in our testing, is that it limits the amount of RAM required in that laptop. Meaning that instead of buying an 8-GB laptop you can buy a 2-GB laptop with SSD. Yes, 2 GB of RAM will mean that the operating system has to use virtual memory and swap pages of memory to storage. Now, though, those page swaps will be to a SSD, not a hard disk drive (HDD). In our testing, the performance difference was almost unnoticeable. If you are looking to upgrade the RAM in your laptop, think about upgrading to SSD instead. Get faster drive performance and faster virtual memory.

Flash memory does not always have to be the size of a mechanical hard drive. Using a flash in a memory-module form factor, instead of a drive form factor, can significantly reduce the size of the system. Drive form-factored SSDs, of course, are ideal for upgrades, but new system suppliers should be considering a memory-module approach. This will allow them to build smaller laptops or use the additional space for a larger battery.

The Enterprise Case: While we almost always focus on the performance aspect of flash SSD in the data center, it should also impact how much DRAM you put in a server. Enterprise SSD is faster than what you would put in your laptop--especially PCIe SSD--so thinking of using it as an alternative to additional server memory can be very realistic. The cost of 64 GB or more of DRAM in a server can add up quickly, and it's also risky since it is volatile.

There is also the physical limit of how much DRAM the typical server can physically hold. Comparatively, PCIe flash cards are all in the 500 GB or more capacity range. If you are counting on your database to leverage memory to improve performance, your chances of a cache hit go up dramatically with a 500-GB flash cache vs. a 128-GB DRAM cache. The same is true for a virtualized infrastructure where the need to send pages of memory to storage is more common and the need for RAM more severe.

Where does this leave DRAM? Ironically it has a significant role in storage. As we discussed in our article The Advantages of DRAM SSD, DRAM still enjoys a significant write-performance advantage over flash SSD, as well as a durability advantage. It makes sense then to leverage DRAM as part of a flash solution so that it can take the brunt of the inbound write traffic and increase the life and performance of the overall storage system.

We have always thought of SSD as expensive storage, needed for high-performance situations. Instead maybe (and especially in the server-based case) we should be thinking of it as cheap and more plentiful DRAM and then use a small amount of DRAM where it makes sense in storage systems managing high-write I/O. In fact, mixing the two memory types (flash and DRAM) may become standard practice, especially in the enterprise.

Follow Storage Switzerland on Twitter

George Crump is lead analyst of Storage Switzerland, an IT analyst firm focused on the storage and virtualization segments. Storage Switzerland's disclosure statement.

Comment  | 
Print  | 
More Insights
Newest First  |  Oldest First  |  Threaded View
User Rank: Apprentice
10/26/2011 | 10:34:39 PM
re: The Death Of DRAM?
There are some new things to do with DRAM, like embedding CPUs in it.

see p.10 for Sandia Labs MapReduce benchmarks.

Russell Fish
[email protected]
Why Cyber-Risk Is a C-Suite Issue
Marc Wilczek, Digital Strategist & CIO Advisor,  11/12/2019
6 Small-Business Password Managers
Curtis Franklin Jr., Senior Editor at Dark Reading,  11/8/2019
Unreasonable Security Best Practices vs. Good Risk Management
Jack Freund, Director, Risk Science at RiskLens,  11/13/2019
Register for Dark Reading Newsletters
White Papers
Cartoon Contest
Current Issue
Navigating the Deluge of Security Data
In this Tech Digest, Dark Reading shares the experiences of some top security practitioners as they navigate volumes of security data. We examine some examples of how enterprises can cull this data to find the clues they need.
Flash Poll
The State of Ransomware
The State of Ransomware
Ransomware has become one of the most prevalent new cybersecurity threats faced by today's enterprises. This new report from Dark Reading includes feedback from IT and IT security professionals about their organization's ransomware experiences, defense plans, and malware challenges. Find out what they had to say!
Twitter Feed
Dark Reading - Bug Report
Bug Report
Enterprise Vulnerabilities
From DHS/US-CERT's National Vulnerability Database
PUBLISHED: 2019-11-14
fs/btrfs/volumes.c in the Linux kernel before 5.1 allows a btrfs_verify_dev_extents NULL pointer dereference via a crafted btrfs image because fs_devices->devices is mishandled within find_device, aka CID-09ba3bc9dd15.
PUBLISHED: 2019-11-14
Scanguard through 2019-11-12 on Windows has Insecure Permissions for the installation directory, leading to privilege escalation via a Trojan horse executable file.
PUBLISHED: 2019-11-14
Microstrategy Library in MicroStrategy before 2019 before 11.1.3 has reflected XSS.
PUBLISHED: 2019-11-14
STMicroelectronics ST33TPHF2ESPI TPM devices before 2019-09-12 allow attackers to extract the ECDSA private key via a side-channel timing attack because ECDSA scalar multiplication is mishandled, aka TPM-FAIL.
PUBLISHED: 2019-11-14
SnowHaze before 2.6.6 is sometimes too late to honor a per-site JavaScript blocking setting, which leads to unintended JavaScript execution via a chain of webpage redirections targeted to the user's browser configuration.