[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200811145827.GA6641@openwall.com>
Date: Tue, 11 Aug 2020 16:58:27 +0200
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: Performance of 5 days demo of the cloud which was mentioned in the message with subject John in the cloud
Hi Johny,
Performance varies between AWS instance types a lot. On the new "John
the Ripper in the cloud" homepage there's a screenshot (now with ALT tag
giving more detail) showing a p3.2xlarge instance's NVIDIA Tesla V100
GPU achieve ~27M c/s at md5crypt-opencl for "Many salts". That's a very
good speed, but this instance currently costs $3.06/hour on-demand or
$0.918+/hour spot, plus $0.64/hour for our Bundle (beyond our 5-day free
trial). You appear to be asking about cheaper/free instances, so for
comparison a t2.nano (the smallest instance they have) achieves ~70k
short-term at md5crypt for $0.006/hour and a t2.micro achieves the same
(but has more RAM) for $0.012/hour on-demand or currently $0.0035+/hour
spot. These tiny instances are also free for the first year for new AWS
users, and we don't charge for usage of our Bundle on those.
So by using these tiny instances, you could get something like 0.26% of
the performance of our recommended instance type, but it's free or
cheap, and in absolute terms (not relative to a high-end GPU) the speed
might just be sufficient for some focused attacks or detection of the
weakest passwords. Those tiny instances would also fare relatively
better for hash types that are GPU-unfriendly (e.g., bcrypt) or for
which there's no GPU support yet.
For WPA PSK in particular, I am getting 2200 to 2300 c/s short-term on
either t2.nano or t2.micro.
But there's a catch:
Performance on these tiny instances might become lower with long-term
100% CPU load unless a paid option called "T2/T3 Unlimited" is enabled.
Without that option, t2.nano's performance reduces to 5% and t2.micro's
to 10% of original in my longer benchmark runs. Specifically, I get
only 214 c/s long-term at WPA PSK on t2.micro without "T2/T3 Unlimited".
Some of my tests on t2.nano fail because its 0.5 GB RAM is too little,
but many other tests work. I recommend at least t2.micro with its 1 GB.
You obviously need larger instances for serious usage, but these tiniest
ones are suitable for some educational uses (such as learning how to run
a password cracker on AWS at all).
5 days free trial applies to our software only, and only for larger
instances (on these tiny ones, our software is free without a time
limit). If you're new to AWS and choose one of these tiniest instances,
as I understand you can use it for free for the first year since your
AWS account creation. If you're no longer eligible (have an old AWS
account) or use larger AWS instances, you need to pay for their usage
(and there's no free trial for that).
The prices I mentioned here are for instances in a US region and for an
AWS account with a certain US billing address. There might be sales tax
or VAT added depending on choice of region and/or billing address.
I hope this isn't too much detail for you.
Alexander
On Tue, Aug 11, 2020 at 02:42:50PM +0200, Johny Krekan wrote:
> Hello, I would like to ask what is the estimated performance of the
> cloud (like passwords/second for example for WPA PSK hashes).
>
> Five days means that one user who is testing this cloud will have some
> hardware resources remotely at his disposal for 120 hours?
> Thanx for info
> Johny Krekan
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
