The Wiert Corner – irregular stream of stuff

Jeroen W. Pluimers on .NET, C#, Delphi, databases, and personal interests

  • My badges

  • Twitter Updates

  • My Flickr Stream

  • Pages

  • All categories

  • Enter your email address to subscribe to this blog and receive notifications of new posts by email.

    Join 1,927 other followers

Archive for the ‘ECC memory’ Category

D-1541 with increased speed plus support for SR-IOV and DDR3 memory might end up in Supermicro SYS-5028D-TN4T somewhere in februari

Posted by jpluimers on 2016/01/04

I wrote about TinkerTry’s Xeon D-1540 fueled ESXi 6.0 home lab build begins LIVE! before as I think it is an amazing buy.

In the mean time, there are more some bundles of it available, even one shipping from The Netherlands:

Currently these machines (called SYS-5028D-TN4T) contain a Xeon D-1540 processor and use the X10SDV-TLN4F that also lists the Xeon D-1541 processor which will likekly be available on that board starting from about February. The Xeon D-1541 processor which adds new features: is about 5% faster and adds support for both SR-IOV (which can help with virtualisation, but isn’t supported by ESXi on these processors yet), DDR3 memory and Storage Performance Development Kit support for storage acceleration: Intel Xeon D-15×1 Storage Accelerated SKUs.

You have to choose, as they use the FCBGA 1667 which – like any other BGA – are soldered. So if you need these feaures and can wait 2 more months, then go for the Xeon D-1541. Otherwise, order now.

Some articles to help you decide:

On a different topic: if you want to cool the processor better, consider re-applying cooling paste: Socket FCBGA 1667 aftermarket cooling?? | Page 2 | ServeTheHome and ServeThe.Biz Forums.

And finally two more things:

  1. There is a D-1548 too which compared to the D-1540 has the CPU speed, but adds SR-IOV and DDR3 memory support.
  2. There will be 12-core and 16-cores varieties in the Xeon-D series later this  year: Intels Server-Prozessor Xeon D-1500: 16-Kern-Version tritt 2016 gegen ARM-SoCs an | heise online.
    According to the current SuperMicro Xeon-D information these will be available in boards that seem to fit in the SYS-5028D-TN4T system:

    1. X10SDV-12C-TLN4F.
    2. X10SDV-16C+-TLN4F.

Anyway: I think with either processor, you have a great deal!

–jeroen

Posted in ECC memory, Hardware, Memory, Power User, Virtualization, VMware, VMware ESXi | 1 Comment »

TinkerTry’s Xeon D-1540 fueled ESXi 6.0 home lab build begins LIVE! | TinkerTry IT @ Home

Posted by jpluimers on 2015/06/26

Interesting machine: TinkerTry’s Xeon D-1540 fueled ESXi 6.0 home lab build begins LIVE! | TinkerTry IT @ Home.

It does ECC and IPMI, fits mSATA, 2.5 drives, and 3.5 hot-swap bays. For more specs see Supermicro SuperServer mini-tower ordered with 64GB of memory for $1900 starts the ultimate 24×7 home virtualization lab | TinkerTry IT @ Home.

The drive trays used are these:

Besides the official ways of converting the 3.5″ tray for 2.5″ drives that only allows for 4 drives to be converted, I think there is an unofficial way that allows for more hot-swappable 2.5″ drives.

Converting the unofficial way

The case is an CSE-721TQ-250B. Looking at a picture of the dimensions of the space where the drive trays fit in (thanks Anandtech), removing the CSE-SAS-733TQ backplane and fitting these into the MCP-220-00075-0B should allow for 8 2.5″ drives to be connected:

(all found via Sata Rack Enclosure – on Aliexpress.com)

Converting the official way

Modify an existing 3.5″ tray to fit a 2.5″ drive: please note the drive is mounted up side down!

Note there is another (dual!) 2.5″ tray for a different kind servers: many 2U and some 4U rack mounted SuperMicro servers and chassises optionally have this in the back for boot drives:

These are for instance used in the below chassises and SAS extenders based on them (the extenders all use SFF-8644 HD mini-SAS3 cable connectors):

Note to self: if ever getting these, ensure to get them with PWS-920P-SQ or better power supplies as these are super quiet although the chassis fans can be loud which can be resolved with a PWM controller or carefully selecting the SuperMicro parts from Supermicro | Support | System Fan Matrix as described in Supermicro noise levels | ServeTheHome and ServeThe.Biz Forums.

–jeroen

Read the rest of this entry »

Posted in ECC memory, Hardware, Memory, Power User, RAID, SSD | 1 Comment »

ECC vs non-ECC RAM: The Great Debate (via: Nex7’s Blog). Use the ECC dude.

Posted by jpluimers on 2014/03/30

Read this very nice post on Nex7’s Blog: ECC vs non-ECC RAM: The Great Debate.

There is no debate. Use ECC dude.

Use ECC especially for server side things (storage, virtualization, databases, etc) where you employ some kind of redundancy/correction in the storage (ZFS, RAID, etc) side of things.

And think about using ECC for the rest of your stuff, especially when things stay in memory for a longer period of time (in-memory processing of data can speed up things a lot, but also increase the risk).

Summary:

There is no debate here. None.

[…]

if you think non-ECC RAM can compete with ECC RAM, you are mistaken. If you think there’s a risk/reward analysis here, you’re correct. The risk is not gigantic, and there’s a real cost to alleviating that risk. You have to decide if that cost is worth alleviating that risk.

[…]

If you believe there’s a risk/reward plan where you can take the reward and apply to to mitigate the risk, you are back to being mistaken. The only benefit of non-ECC RAM (and thus the only reward in its choice over ECC RAM) is it will make the solution cheaper. There is not, however, any way (that I’ve heard of, yet) you can use the cost savings to mitigate the risk using non-ECC RAM will introduce.

[…]

If you choose to use non-ECC RAM, you open yourself up to a new vector for data corruption/loss/downtime/errors/etc,

one that could (rarely) even cause you to lose your entire filesystem, and one ZFS does not (cannot) resolve for you. Indeed, one it likely can’t even see at all. If you choose to employ non-ECC RAM, or are forced to do so because of circumstance or environmental constraint, that’s potentially understandable (and even acceptable) – but do not then attempt to validate or explain away that choice with pseudoscience or downplaying the risk you’ve added. You are using an inferior solution with an extra vector for data corruption/loss that ECC RAM solutions simply do not have. It is that simple.

[…]

Hint 3: There’s a reason we’re so gung-ho about using ECC RAM for ZFS, and it’s not just because we’re paranoid about data loss (which goes hand in hand with being a ZFS zealot, really). It is because you likely don’t realize how at risk you are. Due to the nature of how ZFS handles writes, your incoming (write) data is at risk of RAM-related bit errors for likely significantly longer than traditional storage solutions or alternative filesystems. 5, 10, 30, 60 or more seconds in a state where it is at risk.

Read the rest of this entry »

Posted in *nix, ECC memory, Endian, ESXi4, ESXi5, ESXi5.1, ESXi5.5, Hardware, Hyper-V, Linux, Memory, Power User, SuSE Linux, VMware, VMware ESXi, Windows, Windows 7, Windows 8, Windows 8.1, Windows Server 2008, Windows Server 2008 R2 | Tagged: , | Leave a Comment »

 
%d bloggers like this: