Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror
×

Comment Re:How about a backplane? (Score 1) 51

It depends.

How about a 1U rack-mounted box, with redundant PSUs, hosting around a hundred of Pi SODIMMs, with a USB-Ethernet chip for each slot, every X (7..24) ethernet lines connected to a switch chip with a gigabit uplink on the back of the enclosure.

Current RPi USB-Ethernet is slow because of the USB-HUB-and-Ethernet chip used. You won't have a USB hub in this configuration, and you can use a better USB-ETH chip.

Also you don't need a single, huge, expensive PCB. You can have multiple small PCBs side by side each one hosting just X (see above) cards.

Put a small switch next to each RPi DIMM slot on the board so you can power them off and replace-them online.

Need more local storage for each node ? - you can have a SD-Card slot next to each Pi slot - there are 46 GPIO available pins for those kind of extras (RTC, temperature, front panel leds & reset button for each RPi DIMM)

And Liz just let slip in the announcement comments that Broadcom might enable some extra functionality (ETH, SATA or other fast interconnect) on those extra pins.

PI-nsta-cluster.

Comment Re:BUT, More RAM please sir... (Score 5, Informative) 51

The BMC chip can access 1GB of RAM, but unfortunatelly 512MB is the largest size currently produced in that form-factor (Chip-on-Chip BGA DDR1).
And the Raspberry Pi Foundations does not buy RAM chips in enough volume to justify to any vendor a custom made memory chip at that price.
So we're stuck with 512MB ... unless this new SODIMM form-factor is so succesful that they have enough volume to get that custom 1GB chip made for them at the same price as the current 512MB one.

Comment Re:systemd Architecture (Score 1) 641

As a sysadmin using RedHat in production environments, i will not recommend upgrading to RHEL7 if it uses systemd as it is now, with no backwards compatibility:

1) We have tons of old proprietary software which only knows about init.d startup scripts, cron & so on. That software will surely not get updated to be systemd-aware by the vendor and we cannot make "hacky" things for compatibility as we lose any vendor support.

2) journald does not support remote logging. When you have servers handling PCI data, you spend more on external hardware & software who make sure all PCI data is protected than you spend on the server itself. No remote logging = no go for systemd.

Also, on a production server, you really don't care if the OS startup scripts run in 1 minute or 15 seconds. You still have to wait 5+ minutes for the BIOS to finish the memory & hardware tests, the [proprietary] DB still takes a lot of time to start up.

Comment Toyota recall ? (Score 1) 305

I have a Toyota, it's traction control and all associated assists are acting crazy under certain circumstances (Check Engine light on due to stupid sensor in exhaust pipe + wet road) but i wasn't notified of any recall.
Could this be just for cars that are still under warranty ?
If that's the case, from where i can download the updated firmware and how do i install-it ?

Comment Re:Not sure what they're doing (Score 2) 62

If it can monitor blood sugar level non-invasively and continuously and can make a iPhone/iPad/Mac do something on set thresholds, i am buying one for my mom, no matter the cost.

I wish you all to never have to witness first-hand a loved one almost dying from an accidental insulin overdose.

Comment New ChomeCast Device ? (Score 4, Interesting) 104

Hope Google releases a better ChromeCast device - with an Ethernet port and support for accepting HDMI-CEC events from the TV so you can use the TV remote to Play/Pause/FF/RW.
The current one is sucky.
And if you are on a metered internet connection, beware: While plugged in, the current ChromeCast pulls lots of large photos to display as the screensaver slideshow. It would be nice if it could be pointed to a local network share to display a slideshow with your own photos.

Comment Re:Yo dawg I hurd U like computers... (Score 3, Interesting) 219

Something like this is already implemented on HP's PA-RISC and (later Itanium) servers since last century. Go onto the system console, type Ctrl+B and you have access to a small computer completely separate from the main OS and CPU, running diagnostic software which has "probes" on all the hardware buses and components. On the newer servers you can use-it to power on and off various parts of the server, and to enable or disable various busses and connections, allowing you to electrically partition a single server in multiple ones (CPU board 0 + I/O board 0 = 1st server, CPU board 1 + I/O board 1 = 2nd one ...) power them on and access their consoles. Like some kind of VMWare implemented in hardware.

Slashdot Top Deals

Friction is a drag.

Working...