HP Networking
Discover how the new HP Networking combines the technologies and alliances of 3Com, ProCurve and TippingPoint into the next networking leader.

UPoE: a solution to a problem? Or a solution looking for a problem?

By Sreeram Krishnamachari, Global Product Line Manager, HP Networking


Recently,
I’ve witnessed some great networking technologies reach the market: 10G on copper, FCoE, switch virtualization, unified management—the list goes on. All these technologies aim to solve one problem or another for network administrators.  


But once in a while
, I come across a solution that’s looking for a problem. Case in point: Cisco’s recently announced UPoE, or Universal Power over Ethernet. A proprietary version of PoE (Power over Ethernet) technology, UPoE is a new module for the Catalyst 4500, which delivers up to 60W/port at the switch. And this module will work only with the (super) expensive SUP 7. 


A Power over Ethernet (PoE) primer


To clarify, PoE enables power delivery through CAT cables for network connectivity. When PoE was introduced about 8 years ago, it delivered up to 15.4W per port. It was used primarily for powering up access points located in ceilings, where it is difficult to build out a power outlet.


Over time, as the price for PoE declined, other applications like VoIP phones and cameras began adopting the technology. Approximately four years ago PoE+ was introduced. It increased power consumption to 30W/port and provided more granular power delivery. Interestingly, since the introduction of PoE+, very few applications have taken advantage of its extended power
.


I recently browsed more than a dozen powered device (PD) vendor websites for IP cameras, VOIP phones & PoE access points, and took note of the peak power consumption.
Out of 104 devices I looked at, only three consumed more than class 3 PoE power. And these devices used only 16 to 18W of peak power. And this is the situation for a standards-based technology (IEEE 802.3at). For the statisticians in you, check out the normal distribution of the peak-power consumption of these devices in the adjacent image:


Is UPoE’s really worth it?


Now, a networking vendor introduces a 60W per port proprietary technology without the backing of ANY standards.

And the proposed use for the technology is to power other switches. The question I can’t help asking myself is why? UPoE sounds like a solution looking for a problem.

Peak Power Consumption of PoE Devices.jpg

  • In general, PoE is not an efficient technology for transmitting large amounts of power. Approximately 18% of power is dissipated through the cable. In other words, for every $1M of electricity I buy, I lose $180K to dissipation! 
  • Switches, unlike phones or access points (APs), are in fewer, more specific locations within a campus – it is much easier, greener and economical to provision a power outlet in a closet vs running long UPoE cables. 
  • Even assuming a plethora of applications to take advantage of this higher power range will emerge, there is no guarantee that these future devices will run on Cisco’s UPoE implementation as it is not standards based.

 

Read about PoE’s implementation issues yourself


For the record, when Cisco introduced its proprietary version of PoE several years ago, it created all sorts of problems in connecting standards-based PoE devices to its switches. It was due primarily to implementation differences between what was defined and approved by the IEEE standards body and Cisco’s proprietary implementation.

You can read about the issues in Cisco’s support forum.


So to summarize, there are few to no applications for UPoE. It is proprietary, costs a fortune and isn’t very sustainable!


Is UPoE really solving a problem? Or is it a solution just looking for a problem?


 

>> HPN Blog: Where our customers win in today’s competitive networking arena
>> SearchNetworking.com: Cisco Catalyst 6500 upgrade leaves classic users behind: So what?

>> >> Before replacing existing Cisco switches & routers with more of the same Pause and consider this
>> Learn more about HP Networking products and solutions for the Instant-On Enterprise

 

          7-28-2011 2-08-39 PM.jpg

 

Comments
Fryguy(anon) | ‎07-29-2011 06:36 PM

What is not taken into considertaion here are remote locations where we need to install remote devices (switches, routers, etc) but do not have UPS power available.  By allowing network admins an alternate to having to install either local UPS (which means increased support/point of failure)  or having to go with an unprotected power source - this is an alternative.

 

 

sreeram(anon) | ‎07-30-2011 01:08 AM

Hello Fryguy,

 

Thanks for your comments. Cannot deny corner case alternative use model for this like the one you just pointed out. Some admins may use this in those scenarios even though the ROI is questionable (given the solution cost and the power inefficiency of enabling it)  - However the uncertainity that comes with the proprietary nature of such a solution  will prevent any meaningful adoption of the technology in the market.

 

Again, appreciate your insights.

 

Cheers,

_Sreeram

Networking Nerd(anon) | ‎07-30-2011 10:40 PM

A few thoughts:
1.  Yes, there isn't much of a market for delivering 60 watts per port today.  Why's that?  Probably because the devices that might benefit from having that much power available aren't capable of being run over PoE, at least today.  When 802.3af topped out at 15.4 watts, everyone started scrambling to reduce power consumption due to the ceiling that had to be obeyed.  The original 802.11n pre-standard access points that I saw from several vendors required 18-20 watts to run over PoE, and since that was more than the standard could deliver, we were faced with the options of running an electrical cord to them or using mid-span injectors to provide additional power.  Now that we have 802.3at (PoE+ is a vendor term), we have up to 30 watts per port available for the last two years, people have started using it to drive things like high definition video cameras and even PCs.  Many of the newer thin client devices that I've seen can be powered via 802.3at, reducing the reliance on power cables.  Now, with the advent of 60 watt technology, which I believe is standards-tracked under 802.3at Draft 4.1, I can power something like a monitor via PoE.  This means that my desktops can completely eliminate power cords and have some flexibility in deployment, only requiring two Category 5e cables to run the monitor and the PC.  In addition, since those devices are now a part of the power infrastructure provided by networking equipment, the environmentally friendly architectures put in place to power off phones and access points during idle times can now be extended to places where computers may be left on 24/7, like schools or businesses.  Control of the power infrastructure now extends beyond a simple power cord.


2.  Standards are important in any industry, but so is innovation.  I'm going to speak specifically to Power over Ethernet here to stay within the theme of the post.  Two vendors originally came up with designs to provide electrical power via Category 3 copper wires.  This could lead to dangers from connecting a powered device to the wrong power sourcing equipment.  Damage to the equipment, or harm to the user could occur in extreme cases.  Thus, the need to develop a standard for the wiring pattern.  While this did force the non-compliant vendors to conform to the new standard, it recognized their innovations and development in the area and built upon them.  Similarly, when additional wattage was needed, vendors worked to provide more power (PoE+) despite the lack of a standard.  Now that 802.3at has been ratified, the standard can provide the appropriate power for all standards compliant devices.  With 60 watt power appearing to be on track for ratification via 802.3at Draft 4.1 it is only a matter of time before it becomes standard as well.  This appears to be the pattern for many developments in networking, whether they be MPLS or link aggregation or power over Ethernet.  A vendor takes the chance of development of a technology to address a new need, usually not concerned with interoperability during development.  This is the engineering mentality of "Just make it work".  Once the technology has proven itself, then formalization and standardization can help the technology reach a wider audience.  This is the idea "Okay, now that it works, write it down so we can do it again".  I firmly believe that proprietary design and standards-based technologies can co-exist in the same areas without being mutually exclusive.  Based on the history of technology, I would almost argue they both have to exist to allow for innovation and adoption.

rangello(anon) | ‎07-30-2011 11:44 PM
What about VDI and UPOE powered zero clients? Pretty cool stuff - http://www.cisco.com/en/US/prod/collateral/switches/ps5718/ps4324/solution_overview_c22_676338.pdf Also, pre-standard versions always come before industry standards, and I can't think of a greater innovator or another vendor that's more involved in the standardization process..
Sreeram Krishnamachari(anon) | ‎08-01-2011 07:22 PM

Hi Networking Nerd,

 

Thanks for taking the time to elucidate your thoughts on the subject.

 

I agree with you that some customers might like the aesthetic appeal of not having any power outlets and just using network cables to power the monitor and PCs. There might also be some other corner case applications for this. But  as you pointed out, lot of typical office monitors today can be powered within 802.3at specifications. 

 

But let me address some of the other aspects:

 

First, on standards, currently there is no official activity within 802.3 for anything beyond current 802.3at. Very few within the standards body are convinced of a real market need for this. So, per the sources, 60W power is not being ratified through any forum whatsoever.

 

Second, the savings from powering off PoE devices needs to be examined against the cost of power dissipation. Most of the end point devices today go into some sort of standby state when not in use (Its true on my laptop, my monitor & the VoIP phone I have on my desk). Several APs also do this. So, most of these devices at idle are consuming very little if any standby power. Granted, if we are powering through PoE we can completely turn it off and even save on this small standby power dissipation. (And in fact, HP supports this). But what price are we paying when the device is active? We are dissipating a significant amount of power due to transmission losses and greater the power transmitted, greater are the losses! So in effect, the savings of turning off a standby device is more than offset by much larger power dissipation on an active device! The ROI is questionable.

 

If dynamic power savings is the goals, technology like Energy Efficient Ethernet (802.3az) which reduces power consumption based on true traffic pattern is more beneficial and we will see real savings.

 

 If green is a criteria and there is a need to deliver lot of power then it is best to provision an outlet where possible.

 

Finally, your comments on standards are well thought out - However,

 

1. Networking being the plumbing of IT, relays heavily on working with a host of other devices.

2. The life cycle of a deployed network is continuing to get longer.

 

Both of these make it a prerogative for network managers to design solutions that will last for next several years or a decade. With standards, customers can be assured that at least the technology they are implementing will be supported and adopted by the broader market.

 

IMHO, doing standards based design doesn’t preclude innovation - in fact, it is the opposite, it is innovating by involving all the key constituents and building out solutions that can be broadly adopted in the market place. It is ensuring that the end user or customer has maximum choice and flexibility. It fuels competition and helps the consumer in the end.  It empowers the customer instead of a specific vendor.

  

Cheers,

Sreeram

PS:  you have a cool pseudo name  :smileyhappy:

Michaelw | ‎08-11-2011 02:07 PM

I think I'm going to xPost my Linked-in comments here as I feel they are relevant..

 



 I kind of wonder if Cisco has something against Green IT. This is an example of yet another technology being released by them that makes me think that Greenpeace should camp outside their office. Whilst HP (and others) is working on releasing more efficient equipment (such as the MSM46x WAPs, which are - AFAIK - the only dual-n radio APs capable of being run on standard PoE), they seem to be focussing on the ability to pump more power across the wire or keeping hopelessly inefficient equipment in service. 

The 3750x switch is probably the only 1RU switch around that can use dual-1Kw+ PSUs; I don't think even most File Servers require that much juice. Maybe if an environment just happens to have a shedload of PTZ cameras within a 100m radius of the switch this might make sense but otherwise...? 

At the other end of the spectrum, take their recent release of the Sup-2T for the 6500 series in order to prolong its use. The 6500 was a great switch and still is a quality device; unfortunately it was never designed with power efficiency in mind. 

There are many other examples...10GbE over copper on Cisco requires > 10w per link on Cisco (vs. 4w X2 or <1w SFP+) and yet they push it as an "advantage" over other vendors. HP didn't release this technology until they could get the power consumption down to at least X2 levels. 

I don't set out to "bash" Cisco myself; I've worked with their kit most of my career and they really are the foundation of modern networking. Unfortunately, their complacency for nearly the last decade has put them perpetually into "catch up" mode lately, and they seem more intent on releasing poorly thought out solutions with great marketting spin then in actually engineerin fundamentally better products. Other than their routers, I've been very disapointed in what Cisco has delivered for nearly the last decade.

severud | ‎08-29-2011 04:33 PM

So then HP will not be supporting any higher power output than specified by IEEE 802.3at-2009? In other words, if 802.3at draft 4.1 is finalized HP will not support it in any products?

Sreeram Krishnamachari(anon) | ‎08-30-2011 12:05 AM



Hi Severud,

 

Thanks for your comment.

 

There is NO 60W specification for either a PSE or a PD in any of the IEEE drafts. The standard shows a number of PSE and PD configurations, NONE of them are 4 pair. “The IEEE802.3at draft 4.1 specifically states ‘PSEs shall not operate both Alternative A and Alternative B on the same link segment simultaneously ‘. Any statement that a four-pair PSE device complies with IEEE 802.3at is false and misleading. So any 4-pair implementation is a proprietary implementation and is susceptible to interoperability issues. 

 

HP innovates based on the customer demand for solutions - if we find a strong demand for this, we will help drive the appropriate standards to deliver a solution.

 

Cheers,

-Sreeram

GreenIT(anon) | ‎09-16-2011 08:09 AM

Hello Sreeram -

 

Thanks for this article as I was looking myself for some education on UPoE. I did not know it was proprietary, but I'm not concerned as good innovations become standards anyway. Look at Cisco's "tag switching" who became MPLS for everyone's benefit.

 

Despite the 18% loss, I found the possibility to use EnergyWise quite appealing to me, particularily in comination with UPoE devices. I can measure, manage and control in a very efficient maner, and have total IT power usage visibility under a single application / pane of glass. How would you do that with a separate power plug that can't be measured or controlled easily?

 

When I look UPoE, IP Phones, VDI, UPoE Monitors, etc... I really see a way to streamline and optimize my expenses. And we can use EEE on the top of that, as this is independant.

 

The other problem with traditional outlet is that they all require a transformer. What the efficieny of many transformers, and how much power do they vampire on the grid even when the device if off or even disconnected? Pretty sure this is significant.

 

In term of redundancy, the ability to use StackPower (the power supplies are shared amongt a stack, so you need less power supplies). No vampire power.

 

I'm trying, but I really fail to see how a traditional power outlet is more benefical....?

 

Avvid(anon) | ‎11-17-2011 04:53 PM

GreenIT:  You're abosutely right.   HP take on this is very short sighted, and not accurate (Doesn't saying anything about Cisco's EnergyWise that actually monitors and saves PoE energy).

 

What about Trade floor Turrets?  What about Kioks?  What about Monitors (Samsung is already making a uPoE monitor with an embedded zero client?   What about higher-powered access points (coming soon), what about powering laptops and PC's through PoE? (Again, coming soon).

 

uPoE is the next logical step.

 

HP's spinning FUD because they don't have it - simple.

 

(And no, I don't work for Cisco)

Sreeram Krishnamachari(anon) | ‎11-17-2011 11:48 PM

Hi Awid,

 

Thanks for your comments & perspective.

 

HP can monitor and save PoE energy using both IMC and PCM+ - the same functionality is available with HP.

 

Higher PoE+ capacity would be the next logical step for a set of niche applications – no denial there – but widespread market adoption will not be realized until it is standardized – It is really up to the customers who are procuring them if they want to risk their infrastructure on procuring non-standard solutions that may not be compatible with future solutions from other vendors.

 

-Sreeram

Leave a Comment

We encourage you to share your comments on this post. Comments are moderated and will be reviewed
and posted as promptly as possible during regular business hours

To ensure your comment is published, be sure to follow the community guidelines.

Be sure to enter a unique name. You can't reuse a name that's already in use.
Be sure to enter a unique email address. You can't reuse an email address that's already in use.
Type the characters you see in the picture above.Type the words you hear.
Search
Showing results for 
Search instead for 
Do you mean 
About the Author


Follow Us
The opinions expressed above are the personal opinions of the authors, not of HP. By using this site, you accept the Terms of Use and Rules of Participation