Monday, January 13th 2020

Kingston Technology Announces Data Center DC1000B NVMe SSD

Kingston Technology today announced the Data Center DC1000B M.2 NVMe SSD, optimized for server boot drive applications, featuring power-loss protection (PLP). Kingston's Data Centre DC1000B is a high-performance M.2 (2280) NVMe PCIe SSD using the latest Gen 3.0 x 4 PCIe interface with 64-layer 3D TLC NAND. DC1000B offers data centres a cost-effective boot drive solution with the reassurance that they are purchasing an SSD designed for server use. The DC1000B is ideally suited as an internal boot drive for use in high-volume rack-mount servers, as well as for use in purpose-built systems that require a high-performance M.2 SSD that includes on-board power loss protection (PLP).

M.2 NVMe SSDs are evolving within the data centre, providing efficiencies in booting servers to preserve valuable front-loading drive bays for data storage. Whitebox and Tier 1 Server OEMs are beginning to equip server motherboards with one, or sometimes two, M.2 sockets for boot purposes. While the M.2 form factor was originally designed as a client SSD form factor, its small physical size and high performance make it attractive for server use. Not all SSD are created equal and using a client SSD in a server application may result in poor, inconsistent performance.
Boot drives are used primarily for booting an OS, but in many use cases today the boot drive has a secondary purpose: logging application data and/or configured as a high-speed local cache drive. The DC1000B was therefore designed with added endurance (0.5 DWPD for 5 years) to handle the OS workload as well as the extra write workload of caching and data logging. In addition to being developed for long-term reliability, the DC1000B is designed to deliver enterprise-level performance consistency and low latency features typically not found on client SSDs. Available in 240 GB and 480 GB capacities.
  • M.2 (2280) NVMe PCIe SSD Gen 3.0 x 4, performance
  • NVMe for server boot workloads
  • Application-optimised capacities (240 GB & 480 GB) keep costs low
  • On-board (PLP) power loss protection
  • Self-encrypting drive (SED) with AES-XTS 256bit
Add your own comment

4 Comments on Kingston Technology Announces Data Center DC1000B NVMe SSD

#1
ixi
TLC for data centers? Wait a minute... IOPS - low. Read I would say okey, but write is low too.
Posted on Reply
#2
notb
ixi
TLC for data centers? Wait a minute... IOPS - low. Read I would say okey, but write is low too.
There are scenarios that need fast write and there are some that don't.
Many systems today are focused around fast reads, whereas writing data happens very rarely.
This kind of drives will be more cost-effective for data lakes and the like (the whole "Big Data" phenomenon).

Fast writting and high IOPS remain a crucial for transactional systems (frequent writing of small packages).
Posted on Reply
#3
Gungar
notb
There are scenarios that need fast write and there are some that don't.
Many systems today are focused around fast reads, whereas writing data happens very rarely.
This kind of drives will be more cost-effective for data lakes and the like (the whole "Big Data" phenomenon).

Fast writting and high IOPS remain a crucial for transactional systems (frequent writing of small packages).
And endurance maybe?
Posted on Reply
#4
notb
Gungar
And endurance maybe?
If a drive offers slow writes, it won't be used in systems that write a lot. Endurance (understood as TBW) is not important here.

You have stuff like high-performance transaction or data gathering systems, where drives may write even non-stop.
A popular choice like 3.84 TB PM1643 can write at ~2000MB/s and has TBW~=7PB. So it can be killed in roughly 43 days on average. :)

And on the other end you have systems that are fed with data very rarely (lets say: every few days) and then they're just read.
This is quite common in large companies (think about dictionaries, e.g. client data - names and addresses).
It's also very typical in science. You make an experiment which provides huge amount of data. Then you analyze that data for days or months.

Many different scenarios lead to optimized hardware. Never mock an enterprise product. It it exists, it means someone needs exactly that. :)
Posted on Reply