[erlang-questions] trouble with erlang or erlang is a ghetto

[Jon Watte]

A rack in each city? Is the Erlang kernel getting more latency tolerant?

Btw, when it comes to "cost effective hardware":

With regards to hardware: The price/performance curve right now favors two
dies with 6 hyperthreaded cores on each, for a total of 24 hardware threads.
Intel will soon be selling (if they aren't already) 10, 12 and higher count
cores, and you can buy motherboards and core logic with support for four-way
and eight-way dies, but the price/performance isn't quite there yet.

When it comes to RAM, you get 32 GB of high-quality, high-performance ECC
RAM for about $600. That's almost less that some companies charge for
shipping :-/

When it comes to disk space, you can pay $100 for a SATA spindle, or $200
for a SAS spindle, or $400 for a SSD disk. However, that SSD disk will have
20x the transaction throughput of the other spindles, so again, there's
really no question that any new box should have an SSD. Even large, database
server type boxes generally should have SSDs, except for long-term storage
nodes (I'm talking dozens of terabytes and up) where transaction throughput
simply doesn't matter.

We're also getting to the point where 10 Gbps Ethernet is palatable. It's
certainly cheaper to get one 10 Gbps port than 10 1 Gbps ports in switching
and network hardware. I imagine that a year from now, you'll start seeing 1
Gbps being pushed down to the "connectivity" part of the segment, and 10
Gbps will be default for any new switches, on-motherboard networks, etc.

So, the *current* best-value hardware looks something like:
32 GB of RAM
24 hardware threads (in 2-way NUMA, btw -- does BEAM pay attention to memory
affinity?)
240 GB SSD, 1 or 2 (RAID-1 for redundancy)
Probably 10 Gbps networking

Next year's best-value hardware will probably look something like:
64 GB of RAM
40 hardware threads (still with 2-way NUMA)
240 GB SSD, 1 or 2 (RAID-1 for redundancy) (it will be cheaper than this
year, but still the "sweet spot" unless you're building RAID 6 volumes or
something)
Definitely 10 Gbps networking


We're all living in the future! :-)

Sincerely,

jw


--
Americans might object: there is no way we would sacrifice our living
standards for the benefit of people in the rest of the world. Nevertheless,
whether we get there willingly or not, we shall soon have lower consumption
rates, because our present rates are unsustainable.



On Wed, Jul 27, 2011 at 3:21 AM, Hendrik Visage <hvjunk@REDACTED> wrote:

> On Wed, Jul 27, 2011 at 11:07 AM, Thomas Lindgren
> <thomasl_erlang@REDACTED> wrote:
>
> > ----- Original Message -----
> >> From: Ulf Wiger <ulf.wiger@REDACTED>
> >
> >> ...
> >> I guess it's fair to say that the Erlang/OTP doesn't have a
> >> "dangerously sexy" approach to SMP scalability. They recognise that
> >> their sponsors [1] would have their heads if BEAM started hanging or
> dumping
> >> core in the sole interest of scaling to ridiculously many cores on
> hardware that
> >> none of the sponsors are using.
> >
> >
> > Furthermore, IMO it's something of a mistake to obsess over
> single-process SMP. I think distributed erlang is pretty sexy.
>
> And *that* my dear friends, is where the "clouds" are hanging out ;)
>
> A rack in each city centre with a bunch of 1u/2u boxes, using
> something like GoogleFS, and we have close to a petabytes of
> distributed disaster tolerant storage and processing, and then I also
> say: Bring me my Erlang :)
> _______________________________________________
> erlang-questions mailing list
> erlang-questions@REDACTED
> http://erlang.org/mailman/listinfo/erlang-questions
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://erlang.org/pipermail/erlang-questions/attachments/20110727/0f63332b/attachment.htm>