tag:blogger.com,1999:blog-17500930.post6330084479383540802..comments2024-03-20T22:57:03.923+00:00Comments on Dean Bubley's Disruptive Wireless: MEC and network-edge computing is overhyped and underpoweredDean Bubleyhttp://www.blogger.com/profile/05719150957239368264noreply@blogger.comBlogger7125tag:blogger.com,1999:blog-17500930.post-44631563776196742072018-03-19T10:30:09.687+00:002018-03-19T10:30:09.687+00:00Dean, several points from my side (even if I agree...Dean, several points from my side (even if I agree that Edge is a bit overhyped, I think it will be somewhat necessary for services to scale)<br /><br />- I guess if we were to look at the continued cycle between mainframe / cloud and distributed computing just on the basis of power consumption then surely we would still be using dumb terminals / vCPE as it would always be more efficient to keep everything compute-wise central from an energy perspective and power down what we're not using. I believe that the goal with Edge compute is scale and autonomy.<br />- The more data collected, the more power is needed to process it. So while you're showing the power consumed by each device it doesn't take into account the data quantity that each device generates and the fundamental need to process the data centrally (ok, you say that solutions such as greengrass could catch on to some extent for scale)<br />- Data / privacy regulations need to be taken into account (e.g. national sovereignty, GDPR etc) and these can skew the numbers a bit in your model.<br /><br />Anyway, it's an interesting way to look at things but I have many doubts that we could get to the number of edge compute locations listed in your rough calculation in the 2023 timeframe. Physical and logical security of those locations also need to be taken into account significantly once data is processed and stored in them.<br />Paul Bnoreply@blogger.comtag:blogger.com,1999:blog-17500930.post-89153225535968082892018-03-18T12:42:22.985+00:002018-03-18T12:42:22.985+00:00I broadly agree with your reasoning but you haven&...I broadly agree with your reasoning but you haven't taken account of green energy. Generating 2kW from a solar array with batteries is trending to commodity according to BNEF. Tesla uses solar panels to supplement its charge stations today. <br /><br />Connection to the Grid may be optional in ten years as power storage tech improves. In fact, this will be driven by the fact that power grid in most countries is in poor state and local generation and storage will be a requirement not a choice. Greg Ferrohttp://etherealmind.comnoreply@blogger.comtag:blogger.com,1999:blog-17500930.post-37918168826468500242018-03-18T11:07:47.046+00:002018-03-18T11:07:47.046+00:00Thanks - and good question, I'm not sure.
I ...Thanks - and good question, I'm not sure. <br /><br />I was just looking at the whole computing "universe" through a lens of power "draw" from supplies. There are certainly numerous approaches to resource management, especially in constrained devices.<br /><br />There are certainly many deeper angles here, but I hadn't seen anyone else try to compare device/edge/cloud in this way, so thought it was an interesting avenue for discussion.Dean Bubleyhttps://www.blogger.com/profile/05719150957239368264noreply@blogger.comtag:blogger.com,1999:blog-17500930.post-86182073548136160642018-03-18T08:07:29.765+00:002018-03-18T08:07:29.765+00:00I mean similar in concept, not using the actual Gr...<b> I mean similar in concept, not using the actual Grid Engine or HT Condor </b>the memories of a product managerhttps://www.blogger.com/profile/09766645827205671111noreply@blogger.comtag:blogger.com,1999:blog-17500930.post-57944970447601485362018-03-18T08:03:38.838+00:002018-03-18T08:03:38.838+00:00This paper is brilliant. I appears the hurdles wil...This paper is brilliant. I appears the hurdles will delay the implementation of edge computing, but it not eliminate the need for it. using power consumption as a criteria, " (using)a rough, order-of-magnitude level, we can see that the total realistic <i> "network edge", with optimistic assumptions, will account for less than 1% of total aggregate compute capability. And with more pessimistic assumptions, it might easily be just 0.1% </i><br /><br />In Grid computing, not all CPU are used at one time,.Utilization factor can be max 18%, but it can reach 90% using a resource manager. Can we use a power consumption resource manager (like Grid Engine an HTCondor in grid computing) in edge, alias fog, computing?the memories of a product managerhttps://www.blogger.com/profile/09766645827205671111noreply@blogger.comtag:blogger.com,1999:blog-17500930.post-63571959828824536542018-03-18T04:50:33.868+00:002018-03-18T04:50:33.868+00:00This comment has been removed by the author.the memories of a product managerhttps://www.blogger.com/profile/09766645827205671111noreply@blogger.comtag:blogger.com,1999:blog-17500930.post-52938719774451804282018-03-18T04:48:52.569+00:002018-03-18T04:48:52.569+00:00This comment has been removed by the author.the memories of a product managerhttps://www.blogger.com/profile/09766645827205671111noreply@blogger.com