i thik celeron j1900 and the new j3160 and n3160 are very interesting to use it with routeros because the low power consumption and relative good cpu performance vs mipsbe,powerpc e500v2, an tilera cpus
will be interesting to see what performance can be achieved with this silvermont 4 core CPU on routerOS
If you guys explain me how to test this, I'll let you know.
Does a regular Test-scheme exist, or a regular configuration for routerOS to use for the speed-tests? Reason I ask - is that on my RB493G's I have around 190 active firewall rules, NAT, Mangle, ~200 Blacklist entries (can go up to 2000 in periods of high script kiddy activities) and some other configurations active. With that they achieve, tested with iperf from one i7 system to one other i7 system 192Mbps max speed if no other activity happens on the network.
I could send you the comparative to my setup, but not to the regular setup's Mikrotik uses.
thats an issue plaguing the industry, on one side are the marketing folks who love to publish pretty numbers about equipment performance, far from real scenario performance, on the other side we have us, the people who has to deal with surprises about performance in real scenarios.
two group of people with different objectives, very different, thats the reason of discrepancy on testing results, vendors do know the real performance of equipment but do not publish it, or at least not in a way to make easy to people to find it, for example i have some document of cisco showing performacne of routers under some scenarios close to real:
for example for isrg2 3945 router:
max performance rfc 2544 = best performance scenario: 8 gbps and 1mppps
ipsec best case scenario = 850mbps
firewall max performance = 3 gpbs
that was the pretty numbers, now go with the interesting ones:
hqos performance with mixed size packet at 75% cpu = 600mbps
pat performance with mixed size packet at 75% cpu = 500mbps
recommended ipsec positioning of product = 180mbps
nat + qos + acl performance with mixed size packet at 75% cpu = 230mbps
recommended WAN circuit for platform = 150mbps around 100kpps
as we can see a 10x performance difference between max firewall performance and the real world scenario performance (3gbps 300mbps (150mbps symetric wan circuit)
is something similar to what mikrtik intend to do with numbers published on routerboard.com product page, for example with CCR1036-8G-2S+EM:
max performance rfc 2544: 27gbps/2.2mpps with 1500byte packets, 27gbps/6.5mpps with 500byte packets and 21gbps/41mpps with 64byte packets
25 ip filter rules: 22gbps/1.8mpps with 1500byte packets, 12gbps/3mpps with 500byte packets and 1.5gbps/3mpps with 64byte packets
64byte packet reveals the same downgrade of 10x+ between this two tests, but this tests are far from real scenarios who pullulate here at the forum where we face the surprise when the performance obtained is far from expected, and the worst thing is that's very difficult to predict, mostly because is hard to find the situation when some guy with the skills, the time, and the equipment to test properly a device capable of multi gigabit speeds, and kind to share the result of this hard work to the community.
Is not easy and vendors take advantage of this to confuse the guy buying the equipment.
will be nice to lay the foundations of more serious real world approach to performance testing on mikrotik/routeros devices to the benefit of all.
I think the first step is to establish some base configurations and testing procedures taken from real world to start gathering results and build an information base useful to all
They are even promising to preinstall vsphere.
