November 2008 Archives

Power Consumption Testing - Some Figures

| 1 Comment | No TrackBacks
| More

Care of my mates at Ixia's test labs in Marlow recently, I took a few D-Link switches, including one from its Green Ethernet range, and ran them through Ixia's IxGreen power consumption measurement suite.

Key here was D-Link's "Green Ethernet" range submission - the snappily-named DGS-1224T Green Ethernet Web SmartSwitch (but none of the other vendors are any better at this naming game either). The smart bit here is that the switch reacts when a device attached to it is turned off, by placing the corresponding port in a standby mode that requires less power. Typically Gigabit Ethernet ports are always in enabled mode and consume power even when unconnected or not passing traffic. D-Link claims this feature can save up to 24% of the power used by a typical Gigabit Ethernet switch - something we put to the test - read on.

Using a real world traffic mix on the Ixia XM12 test chassis, running the IxGreen software, we tested each switch in turn initially at 0%, 50% and 100% loads with default gigabit mode enabled. In each case, we reran the test several times to check for consistency and took the average value - though there was very little variation between iterations, as we would expect. If you see the table below, you can see that the "Greenness" of the DGS-1224T shines through against D-Link's other two switches in the test (which both recorded low consumption figures in their own right, as it happens). The results are in terms of total wattage recorded.

Switch

0% Load

50% Load

100% Load

DGS-3627

32.28

42.15

50.35

DGS- 3426P

33.75

      40.32

46.07

DGS-1224T

20.7

22.63

23.65

The consumption of the DGS-1224T amounted to less than a watt of power required to drive over a gigabit per second throughput - not bad, eh. And in "Green Ethernet" mode, power consumption was reduced from the already impressive 20.7watts (24 ports enabled, no traffic) to a truly outstanding 9.3watts. To put this into some perspective, most PCs consume more power than this in standby mode. These results really do show the benefit of being able to configure gigabit ports to be inactive by default - rather than the industry standard default of being active while ever the switch is powered up.

So, D-Link appears to be setting the pace for power reduction; where are the rest of you vendors? Game on...

Note: A full report on this testing will be available shortly both from the D-Link UK website and the Broadband-Testing website (www.broadband-testing.co.uk).

Power Consumption - The New Performance Metric?

| No Comments | No TrackBacks
| More

Bob Dylan never sang "The Metrics They Are A Changing" but he might well have done, had he launched himself during the era where IT and 'Green' - an unlikely alliance if ever there was one - finally embraced each other.

I'm getting calls from all manner of sources currently re: investigating power consumption of network devices and total consumption of data centres. A recent report on the future of data centres from consultants, MSC (http://www.msc-reports.co.uk/) also highlights the increasing issue of data centre power consumption and how this might lead to a need to relocate in less power-hungry parts of the UK. The reality is that the energy providers are, er, running out of energy when it comes to supplying certain parts of the UK, though the current financial meltdown should mean that issues in the Docklands and the City become less demanding than they currently are...

This power-hungry issue isn't simply an 'enterprise' thing but also an issue for TelCo's and ISPs, something I'm investigating in the new year with a European consultancy, The Cavell Group. Watch this blogging space for more stories from there. Moreover, it's an issue for vendors. Why? Well, let us say that some published vendor power consumption figures are not quite as 'accurate' as they might be.  On behalf of one vendor client I undertook some power consumption measurement work that led me to additionally fall upon a major rival vendor's kit, and, amazingly, some publicly quoted figures for said product. And guess what? Not only were the published figures massively inaccurate, but the device's own stated power usage figures (available at the CLI) were also wildly out! A device that doesn't talk to itself then.

While I realise you're all going to be shocked by the revelation that network vendors are releasing inaccurate figures about their own products, the real story here is the complexity involved in what - on the surface - seems like a very simple task. Power up - attach multi-meter - get measurements for wattage, ampage, voltage (and which of these is the key metric?) - can't be easier? Except that network devices consume different levels of power depending on what the traffic load is, what features and ports are enabled/disabled, what modules are installed (in a chassis-type product), how many power supplies are installed... and so on.

Now, this uncertainty doesn't wear well with the network vendor community, which simply loves its 'standards'. Consequently, in conjunction with one of our test equipment partners, Ixia, we are looking to create a standard methodology for power consumption testing of network devices based around  wattage - that wot most people understand the most and wot most companies are billed against usage thereof by energy providers. The test methodology (based on RFC2544 if that means owt to any of you) also doubles as a performance test suite, so we get two lots of info for the price of one. Which is nice.

My next entry will describe a very example of how this works in practice. Or is that in practise? Sustainable IT anyone? Talking of which, if anyone wants to attend two days of sustainability debate (can it be sustained for two days?) at the next of next week, I can point you in the direction of the Future of Europe Summit 2008 in Andorra (http://www.europesummit.org/) which I will be attending as it is in my personal back yard and the government invited me, so I can hardly say no...


802.11meatballs

| No Comments | No TrackBacks
| More

Why is it that all the furniture in Sweden looks like it was obtained during a raid on IKEA and that, everywhere, you are offered meatballs?

Even at the launch of ProCurve's newly acquired Colubris WLAN technology in Stockholm, meatballs figured heavily (literally) in many conversations. So, what's special about - no, not the meatballs - the Colubris, sorry ProCurve, technology? Well, at first base it follows the time-honoured tradition (of the past five years) of managing the wireless network from a centralised controller. Then, instead of feeding dumb APs (a la Trapeze, Aruba, Meru etc) it power active APs that work with the central controllers to provide obvious scalability advantages over the totally centralised alternative.

So why did the others not think of this? Well, of course they did, but they saw a security issue, a la Fat APs, in that when they get nicked, there's lots of juicy information on them to be extracted. But not in this case, according to Colubris', sorry, ProCurve's Carl Blume, Worldwide Director of Mobility Solutions who states that the AP plays very dead indeed when disconnected from its power source (AKA Ethernet cable). Of course, most of the other solutions were originally designed pre-PoE ratification so maybe that's where the really problem lay.

Carl Blume (who despite being American actually likes walking, which is a great sign) told me that, internally, Colubris (not ProCurve) had tested with their new 802.11n APs and achieved up to 270Mbps on a 1:1 connection, not far shy of the theoretical 300Mbps limit of the .11n technology. Given that, for example, the alleged 54Mbps max throughput of the .11g standard actually was more like 24Mbps, this sounds like something of an achievement.

But ours IS to wonder why, so at Broadband-Testing we'll be putting the Colubris, sorry ProCurve, solution to the test in the next few weeks. Watch this space for the Bluming truth.

And here's the real issue. Will a WLAN technology that finally (perhaps) does perform at modern LAN speeds (100Mbps+) be seen as a wired replacement technology and not just in the branch office? I welcome your thoughts on this one.

I also note that Meru is claiming a first in the WLAN area with virtual AP technology that for a reason I personally cannot ascertain, is different to that presented by the likes of Symbol (now Motorola) years ago. Answers on a virtual postcard at the back of the virtual room please...

Finally, with Trapeze having been acquired earlier this year, now Colubris, and with Juniper still needing WLAN technology it looks like it's a straight scrap as to who they acquire between Meru and Aruba? Unless someone out there knows something different...

Introducing our new networks blogger, Steve "Blogger" Broadhead...

| No Comments | No TrackBacks
| More

steve.jpgSince the departure of John-Paul Kamath for pastures new, we've been searching for a new blogger with the expertise to take on the networks generation... And now, we've found him: Steve Broadhead.

Some of you may recognise the name on the back of several previous appearances in Computer Weekly and might wonder why on Earth he's been allowed to start a blog... Well, Steve claims he has several points in his favour:

a) He's from Yorkshire and therefore demands the right to shout his mouth off and get away with it (watch out Clarkson)...

b) He knows what a punch card machine is and how to use it...

c) He remembers finding Snake on the BBC Model A quite exciting...

d) Likewise Load-Balancing at Layer 7...

e) He knows a man who did get fired for buying IBM...

f) He knows a lot of people who used to work for Cisco - and some who still do...

g) He has sampled a lot of wine, and written a book about it (along with fellow computer hack, Jim Hayes). Amazon might even still have a few copies in stock, and Christmas is just around the corner (Ed: enough).

h) Sometimes, he actually knows what he's talking about.