[Beowulf] Hyper Convergence Infrastructure

[Joshua Mora]

Without giving specific names of technologies,
hypervisors can reduce around 5-10% the performance.
I tested linpack and stream and I got small reduction ~5%.
For networking: bandwidth about 5%, for latency about 10% reduction.
On hyperconverged, since data is replicated, the write performance is halved
if replication of 2. writes can be improved if caching is provided.
Wrong setups, specially drivers can really hurt network performance (>50% drop
performance)
Also numa awareness  pinning VMs is necessary if you care about memory
performance.
Same with NIC performance.
When doing migration it needs to be placed in another machine of exact same
configuration in terms of numa awareness in order to work.
On hyperconverged, pay attention to QoS.

Joshua

------ Original Message ------
Received: 08:13 PM CDT, 10/03/2015
From: "Lechner, David A." <dlechner at mitre.org>
To: "beowulf at beowulf.org" <beowulf at beowulf.org>
Subject: Re: [Beowulf] Hyper Convergence Infrastructure

> Hi -
> I started looking at the marketing material -
> Thanks for both the lengthy and not-so-lengthy responses provided so far -
> Here are clarifications on my questions.
> W/ Regards
> Dave Lechner
> 
> 
> 1)      Does anyone have benchmarking test results of an HCI solution
compared to the similar configuration of a Linux/Intel  hardware set with
comparable networking backbone?  (ie same or similar server core count, CPU
count, clock speed, memory speed and amount, similar storage (maybe solid
state disks), and motherboard configuration, and interconnects).    Hopefully
there is a range of benchmark performance results that can be shared - data
intensive, I/O intensive, FLOP intensive (  I used to do several tests when I
ran a benchmark lab in the 2005-07 timeframe)..
> 
> 2)      If the hardware design is similar, how does HCI solution provide
improved performance over a similar commodity system?
> 
> 3)      How does the cost and performance results of the HCI solution
compare to the commodity solution in question #1 ?
> 
> 4)      If the price per FLOP (or pic your metric of preference for a
cost/performance ratio) of an HCI solution is not better (i.e.  lower cost per
similar performance), then are there other solutions in the systems
administration?
> 
> 5)      Is there a comparable open-source solution?  (With thanks for
partial answers so far!)
> 
> 
> 
> 
> From: Lechner, David A.
> Sent: Saturday, October 03, 2015 6:19 AM
> To: beowulf at beowulf.org
> Subject: Hyper Convergence Infrastructure
> 
> Hi
> I am wondering if anyone on this list has benchmarked the impact of an HCI
solution on performance, or how this newest "next big thing" compares to a new
Linux/intel commodity solution?
> Is there some performance penalty fron the virtualization?
> How do price points per FloP compare?
> Is the advantage in the Systems administration, and is there a comparable
open source solution?
> 
> Thanks in advance for  any insights.
> Dave Lechner
> 

> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
> To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf
>