re: Ericsson Thinks Green for RoutersHow about this: watts per capacity but to how many subscribers? the point is that there are different ways of looking at a situation. Watts per sub follows something like ARPU on the revenue side.
Just to be clear, all numbers in this table, other than the SE1200, were derived from publicly available information and Iometrix did not conduct any tests on them, again other than the SE1200.
The power consumption stated by Ericsson for the Alcatel-Lucent 7450 + 7750 SR is completely wrong. Ericsson chose to wrongly assume that two Alcatel-Lucent boxes are required, which has never been the case for subscriber management. Even with two boxes, the Alcatel-Lucent solution would still be under 30mW/sub.
Using Ericsson's own flawed methodology, the correct data for a single Alcatel-Lucent 7750 SR-7 supporting 128k subscribers is a best-in-class 14.7mW/sub.
Separately, the methodology chosen is flawed, since it assumes <2Mb/s per sub, which is barely enough for basic Internet access, never mind IPTV.
If the comparison used a more realistic, but still less than ideal, 5Mb/s per sub and 128k subs, this would need three SE1200s vs one Alcatel-Lucent 7750 SR-12, obviously resulting in a signficantly higher power consumption by the Ericsson solution, as well as additional physical rack space. The difference is due to the more than 2x difference in Gigabit density between the platforms (i.e. both subscriber scale and density/performance matter).
Craig, you missed the winner so green, so obvious that it hurts. With just 0 (zero) mW/sub, the edgeless architecture is #1 in your table. Regardless of methodology.
When do you replace the numbers with the ones from the rebuttals you receive? I'm sure Juniper, cisco, et al, will rush to defend their systems, as ALU has. Don't you think you owe it to everyone to put the correct numbers into your article, as you receive them. so your piece doesn't smack of Ericsson's marketese?
Ericsson's thinking seems very immature and teeters on the edge of a "paid advertisiment".
The basics of energy efficiency at network device level are now well understood and have been described in a number of publications, including VZ.TPR9205, Green Grid "Proxy Metrics" and ATIS TEER draft to name a few.
The main idea is to measure the "actual energy consumption" to "actual work" ratio in a real test. For packet systems, "actual work" equates to bandwidth with lowest-common-denominator features enabled.
Consensus-driven and non-exotic nature of such simplistic test profiles acts as a safeguard against rigging results with Ericsonian-style "typical triple-play configurations" that would otherwise favor test sponsors.
There is also a (somewhat wordly but still good) whitepaper on a difference between rated (worst-case), theoretical and measured energy published on ECR website (ecr.org).
Of course, different devices need different load; ethernet switches are fed with simpler traffic composure (ie stateless streams) compared to BRAS/BNG devices. But as soon as the reference load is defined and every device is tested within correct equipment class and under the same test methodology, the results should remain fully comparable.
Judging by press release, Ericsson managed to take not one, but two steps back from where everyone already stands.
First off, they introduced two new and quite meaningless metrics (energy per subscriber and energy per circuit). It's quite clear that "subscribers" can have different speed and packet touch profiles, and so do "circuits" - the chosen metrics are too vague to start with. The argument about "unique control plane" requirements does not hold - the same can be said about virtually any network device from firewalls to wireless base stations.
Second, Ericsson decided to steer away from fair comparison in the first place.
If they would want to stick to some sensible methodology, they would never resort to comparing one Ericsson device against stacked devices of other vendors. This is not needed becase capacity differences are patently insignificant to normalized metrics.
Overall, looks like a very clumsy attempt to board a green bandwagon.
We've only gotten the one rebuttal. Problem with any green comparison is that not all systems are equivalent, so every vendor can claim that its gear would require a simpler configuration than others (or a host of other exceptions).
Maybe including the table was a mistake, considering the number of caveats and exceptions that come with any kind of "green" measuring (not to mention vendors' aggressiveness at trying to be greener-than-thou). There's always the question of whether the scenario or configuration being considering is the right one. Data-sheet energy numbers probably aren't the answer, since they're all worst-case; every vendor's going to kludge a better "typical-case" figure. Green is a messy area for these vendors, with a lot of loose boundaries.
For more details on the Ixia/Juniper/Lawrence Berkley National Labs "Energy Consumption Rating (ECR) Initiative" please check out: http://www.ecrinitiative.org/
And I thought the Hulk was the only incredible green monster out there..
Watt per sub, but at which capacity? Is this about having green memory?