we’re looking forward to training some information on how remanufacturing applies to IT and how this differs from mechanical remanufacturing so we found a lot of the public exception is the IT hardware degrades over time and so you see a performance decrease of remanufactured against original products the xfurbish of East London recently carried out research that proves remanufactured service perform identically to new in fact that the immediate previous generation can outperform.
The latest if it’s configured correctly so joining me on this this presentation this webinars is teal she’s the research associate from the University of East London hi everyone so before we start with the research I’m going to explain a little bit about our manufacturing process which is covered on this slide so tech buyers a global expert in buying refurbishment and selling of enterprise IT.
Servers storage and networking for the data enter environment or we’d classes Enterprise spares a lot of our business is focused on bringing in redundant IT from a variety of sources we execute a 25-point testing program on servers data sanitization where necessary and everything is restored to an as new condition to be sold as spares and upgrades we also have a remanufacturing portion of the online business which rebuilds fully configured service.
Which is that with a client according to individual client needs from the component parts and installs the relevant software everything gets issued a three year warranty on all our products anything we can’t use ghost recycling but everyone in the business has a genuine hatred of waste that’s company-wide if we can reuse or repair we do that as a preference every single time our technical staff repair components wherever possible and our expertise and this is growing so from sort of memory and tape drives and all sorts of things I would say our customers care less about that though than the economic benefits of remanufacture so we had their paper written by a post-grad one of our local business schools that said the customers buy from us because of cost and performance rather than the environmental considerations the standard argument against using remanufactured equipment in the data centre is that new equipment is more energy efficient since the IT load for the biggest proportion of energy use in the sector it’s important for customers consider that in the past new service significantly increase the amount of compute power per watt so the advice was janitor refresh often and to refresh with new because there was significant benefits on the latest generation however as times.
New generations are making less of an improvement on older models this is for a variety of reasons and this is what our study with the University of East London was designed to explore so I’ll hand over to Noor now to take you through it thanks rich so a server is in many ways a large computer if you’re open wound up you will see a motherboard CPUs which are behind the heat sinks in this picture memory storage transistors and fans to keep everything from overheating if we stirred this down two essential components we have the CPU which gives the processing power memory which provides short-term access to data and storage which you can think of as the server’s filing cabinet these are the components we concentrated on during our experiments when most people think of the compute power of a server they think the CPU is the workhorse it is generally front and centre of all discussion around compute performance however there are two helpers in this function storage and memory that can be very impactful if you add more storage or more memory you can upgrade the machine we are about to show you what that looks like a big part of the study was understanding server trends over time a study was conducted by uptime Institute which analyses performance data.
for hundreds of servers between 2007 and 2013 the graph displays how CPU technology evolved over time along with performance and idle power consumption the figure reveals interesting insights during the beginning of the decade the move from one CPU lithography to another for example 65 nanometres to 49 nanometres presented major performance per Watts gains represented in the orange line as well as substantial reduction idle power shown in the blue law however over the past few years while lithography was fixed at 40 nanometres the increase in performance per watt has been escorted by a study increased an idle power consumption perhaps due to the increase in core count to achieve performance gains this is one reason why the case for hardware refresh for more recent get has become weaker especially that servers and real-life deployments tend to spend 75 percent of their time on average an idle mode as such the increase in idle power may offset energy gains from performance it was of great interest for us to study the performance gain between latest server generation and an older generation released after 2015 we expected the gain to be minimal due to the server trend discussed above therefore remanufacturing machines from a previous generation could be a good answer for energy efficiency as well as saving on the vital precious metals and critical raw materials that IT hardware.
Contains the first stage was to verify that a new and refurbished server against the same make and model had the same level of performance this is done by measuring energy draw and performance measured as number of operations done per second of the CPU storage memory as well as the overall score we then tested different combinations of new and refurbished for example new machines with refurbished components and refurbished machines with new components our results showed that there was no real difference between the two first this was not a surprise given the last moment service is much longer than 3 years.
However it was good to have this verified by university to the best of our knowledge this is the first time this type of research had actually ever been done we really wanted to know what would happen when an older server was upgraded with additional capacity we would see if it could outperform the last generation that was covered by the next stage of testing the showcase comparison is between an HPE ProLiant dl380 gen9 which was released in quarter 4 in 2014 and an HPE ProLiant dl380 gen9 released in quarter for 2017 these were chosen since they are very popular servers but we plan to expand on this testing for more makes and models we run the same set of tests as we did for comparing new and refurbished energy efficiently but this time varied the components and the older generation 9 model we found that a generation 9 with a 53 and 54 duo CPUs outperformed a generation 10 model with base configuration using dual sky lake silver cpus just by adding more ram the results verified that the performance gain between latest generation and previous one was in fact very little as discussed earlier on we’d like to add though.
This would had not been the case for servers released prior to mid-2014 because server performance jumped significantly after which but we do expect this to be the case for the next few years until a breakthrough and CPU technology takes place now these tests were part of the longer-term projects where we building a data center energy efficiency modeling tool to provide you with scientific recommendations to improve your overall data center efficiency primarily by improving server hardware configuration however what is most interesting here is that we’ve been able to get the immediate previous generation two out for latest model that’s what’s demonstrated on this slide at the moment for a long time the industry believed you Apso.
We had to refresh with the new and latest in order to achieve maximum efficiency what this proves is that in certain environments remanufacture machines could be the better choice to cost performance and energy drop as I mentioned at the beginning the presentation cost and performance are obviously really important many organizations most people they care about the bigger picture of cutting waste and pollution as well as protecting our natural resources the exact size of the environmental impact is under debate at the moment different experts will tell you different things we’ve tried to keep this this slide in line with this we do know that both waste and energy usage is on a steep upward curve for the data center industry so resource and waste and reduce air emissions are all increasing so this is the latest information.
That we have on what it costs us to make a new server it’s from a Deloitte report in 2015 which informed the eco design directive so it’s kind of the best information available at the time tech buyer is currently involved with a multinational research project called sedate which will update and improve on this information the likelihood is that modern servers have a higher carbon footprint and water usage the nice artistic suggests but as you can see on the slide here this there’s quite a lot of knowable materials and energy gone into the actual production of servers the macro picture on this is that humanity is spending way over its limit in terms of its use of natural resources and also what the earth can recover from by 2050 they say we spend will be three times the amount the planet can provide that’s unless we see we change how we do things this kind of slide talks about the ways we can make best use of resources but the order is important too so if we maintain our assets for as long as possible we use a hundred per set of materials and use minimum additional energy refurbishment which is restoring goods to factory condition and reuse which is building them into new machines is our next best option in our sector this is what remanufacturing means there are limits to which components go with which so remanufacturing servers.
Sticking within the confines of each-generation but we can increase performance within these parameters the last and final option is to recycle especially with ICT this is because many of the materials exist in trace amounts and not all of them are recoverable so you’ll never get 100% return rate also with recovering the materials you have an additional energy cost.
Which you want to avoid if you can this kind of thinking comes in the heading circular economy now although the remanufacturing industry existed long before the term was invented those that support circle economy in the data centre industry point to the Google model to show that it can be economically as well as environmentally successful so you know looking at the slide some of the key points on this one is that they’ve kind of been refurbishing their instance 2015 which is a long time they don’t just make a distinction between refurbished and new inventory both are considered equivalent so they both sit on the shelf and they’re not separately labelled and you know it saved Google about billion dollars in operations to energy efficiency alone by kind of refurbishing its own computing infrastructure now Google is able to do this because it manufactured its own servers and control the outgoing and return supply chain so what companies we do is provide smaller organizations with the capacity to do the exact same thing by effectively working as one of their service arms it’s a little different financially because we buy equipment in and then resell but this works well for our customers because they don’t have the overhead they’re also probably not big enough to do on a rolling basis so they get the benefit of specialist remanufacture perfected over many years even if they only use it once so we know that the remanufactured service perform as well as new and that it’s possible to get the immediate past generation to outgun the latest generation with the correct component level upgrades so what does that mean for remanufacturing as a whole we believe it means taking a different approach to IT remanufacturing rather than mechanical remanufacturing for IT we can take customers to a decision tree where questions these are asked on the slide you know there is no wear and tear for IT in the same way that there is for mechanical components you know in fact previous models can outperform newer models with the right component upgrades.
What you’re buying, what it’s gonna be used for and where it’s going is this going to be production is it going to be test is it going to be dev we know is it a noose and you know look at ways you can prolong the life of existing equipment by using at great upgrades of components rather than complete refreshing and also look at the cost saved and what you could buy by saving that so this could be security software a critical infrastructure it could be off-site backup solutions there’s a lot stuff you could be spending the money on better than spending it on new kit that you don’t need so I guess you know what’s the point of all this well mostly that it makes sense and if you have a product that cost less performs just as well and is good.