Microsoft's Sleep Proxy Lowers PC Energy Use

alphadogg writes "Microsoft researchers have slashed desktop energy use with a sleep proxy system that maintains a PC's network presence even when it is turned off or put into standby mode. Microsoft has deployed the sleep proxy system to more than 50 active users in the Building 99 research facility in Redmond, Wash., according to the Microsoft Research Web site and a paper that will be presented at the Usenix technical conference in Boston later this month. ... Sleep proxies allow machines to be turned off while keeping them connected to the network, waking the machines when a user or IT administrator attempts to access them remotely."

Comments

Wake on Lan?

[hedwards]

This is something new? Isn't this basically just wake on lan with an external box? Meaning that rather than having a part of the computer powered on in case the packet to wake up comes through, they're doing it with an external box. I'm a bit curious as to why this justifies any particular coverage.

Re:Wake on Lan?

[Stupendoussteve]

I guess they're claiming it's smart wake on lan. WoL requres sending a specific packet to the machine. Most people don't know to do this (an admin should, but otherwise...) and the network resources will be unavailable in the meantime. This system keeps the resources available and wakes the computer if they are actually needed. It does not rely on someone being smart enough to wake up the system themselves.

Macs have the option to Wake on Demand which requires the use of an Airport base station but seems to follow the same basic concept.

Re:Wake on Lan?

[jridley]

It probably is, but for instance I can't use WOL because it requires a packet that can't propagate through a router. I've thought in the past about setting up a machine on my subnet that I could poke via HTTP or whatever and make it send the wake packet to my PC, and anyone else could use it too. But since I'm probably the only person on my floor of the building that gives a crap about power consumption, it'd be silly to set up a 2nd machine running 24/7 so that I could turn mine off a few hours a day.

Re:Wake on Lan?

[Cley Faye]

It probably is, but for instance I can't use WOL because it requires a packet that can't propagate through a router.

There is the possibility of having a smart router that allow WOL packets; some of them have a "act as a WOL proxy" option built-in, for examples.

Re:Wake on Lan?

[cmacb]

You mean they copied Apple?

Huh. First time for everything.

Re:Wake on Lan?

[Rockoon]

So, basically it's a wake on lan, but that which works everytime some moron is doing a portscan or ssh-breakin attempt on your system? Why would such a system even have a off mode?

..so basically, no its not like wake on demand.

"SleepNotifier alerts SleepServer just before the client goes to sleep, and SleepServer ensures that all incoming traffic meant for the client comes to the proxy instead," Microsoft writes in another article titled "Trying to cure PC insomnia." "The proxy server's role is to monitor traffic and respond accordingly. For some requests, it responds on behalf of the client so the client can continue sleeping, and others it ignores. Some traffic, such as a user access request, causes the SleepServer proxy to awaken the client and present the user with apparently seamless remote access."

So basically we have a system that uses Wake On Lan to wake the remote machine automatically for user requests, but also avoids waking it for stupid shit like pings.

This is, in effect, what other researchers are trying to solve in a decent manner. Wake On Lan requires the waker to know a thing or two about the sleeping system (for example, that its sleeping) and simple frontend devices that have solved this in the past wake the system for everything and are also permanent proxies (proxying even when the system ISNT sleeping, for example)

Re:Wake on Lan?

[v1]

Macs have the option to Wake on Demand [apple.com] which requires the use of an Airport base station but seems to follow the same basic concept.

Macs actually implement the Wake On LAN standard feature. (sometimes referred to as WOMP, or Wake On Magic Packet) This relies on the computer's ethernet hardware remaining awake while the computer sleeps, and any computer on the LAN can send a special UDP packet containing its ethernet MAC (Machine Address Code, unrelated to MACintosh) to trigger the computer to wake up. The only Apple-specific part of this feature is that Apple extended it to wireless use, keeping wifi hardware also active and listening for the magic packet so computers could be woke up wirelessly. Come to think of it thought, WOMP over wireless does require an Airport base station and Mac ethernet adapters - Apple extended the WOMP specs (in an open way) but I think are the only ones presently implementing it?

Looks like Microsoft yet again attempts to take credit for "inventing" something that we've all been using for years. This time it wasn't even ripped off from Apple, it's been in use on all manner of PCs for some time now. This is just MS's first specific support in their OS?

I see a comment immediately below, "it'd be silly to set up a 2nd machine running 24/7 so that I could turn mine off a few hours a day"..... Actually, that's exactly how you wake up machines on different networks such as waking up a work machine from home. Unless your server is asleep too I don't see this being an issue? Remote into it, use it to WOMP your workstation, and then connect to it? Even if you don't have a server, surely keeping one machine awake to provide access to many other machines (easily tens to hundreds) is hardly a hardship.

This is news?

[bushing]

This sounds awfully familiar... http://en.wikipedia.org/wiki/Sleep_Proxy_Service

Re:Give them a break

[XMode]

Reinventing something that's been available for years is not 'coming up with good technologies'.

Now what they SHOULD have done is just cache the MAC of the PC in AD along with the rest of the object (It might already be there as part of the auth stuff) and then mod the remote access client to try and ping first, no reply? Send a Wake on Lan packet.

Misleading Headline

[John Hasler]

I was expecting a sleep proxy for me so that I could stay up all night while the proxy wasted time sleeping.

great for botnets

[e**(i pi)-1]

Swell, botnets can even operate with computers which had been turned off.

Re:MS invented here JUST LIKE THEY ALWAYS DO

[0100010001010011]

It's also OpenSource: Note: see mDNSResponder source code at www.macosforge.org, which includes a full implementation of the DNS-SD/mDNS Sleep Proxy Service, available under the Apache 2.0 Open Source license. AND written up as a specification http://tools.ietf.org/html/draft-cheshire-dnsext-multicastdns-11

Meaning if Linux or *BSD wanted to they too could also have it too. In fact, I'm really hoping that they do because I'd love to not have to send a WOL to my HTPC or Server when I want it to download something. I can just have my sheevaplug wget an address and have it wake itself.

Re:My Macs have been able to do this for some time

[camperdave]

Interesting. Could we have a link, please?

Yes, as soon as the machine wakes up.

Linux or *BSD

[200_success]

It's also OpenSource [...] Meaning if Linux or *BSD wanted to they too could also have it too.

They could, if it weren't patented.

Re:Wake on Lan wakes too often, leads to insomnia

[louarnkoz]

If you read the fine article, you will see that they acknowledge wake on lan and other similar work. They are addressing a practical problem in large networks. Classic implementations of Wake-on-Lan wake the computer when another computer sends it a packet. This looks fine in theory, "my computer wakes up when it has something to do," but it does not work well in practice, in a large network.

In any network of a certain size, there is a lot of noise, scans, keep alive traffic. That traffic causes packets to be received frequently, maybe a couple times per minutes. When a computer awakes, it takes some time to put it back to sleep, maybe a minute. Given enough background traffic, the computer never goes to sleep.

The solution is some form of filter, to only wake up the computer if the incoming data packet is "important." For that, you need a proxy. And the proxy needs a lot of tuning. If it does not wake up on "important" traffic, the users are pissed. If it wakes up too often for trivial pings, the energy bill increases. What they claim here is that after a year of trial, they have validated a particular tuning that works well. Seems interesting indeed.