Welcome back to Raritan Radio everyone. My name is Steve St. Clair. Today, we're privileged to be interviewing Jawahar Swaminathan who is just back from an interesting conference, the Green Data Center Conference, otherwise known as GDCon in Santa Clara, California. Welcome Jawahar!
Thank you Steve! I'm glad to be here.
Now, you just got back. I know the conference ended this week. You're back just a day or two, so you've probably got some very fresh thinking on what you experienced, who you met and what -- more importantly, what the hot topics and hot issues were that everyone was talking about at GDCon. Can you tell us a little bit about what you saw and heard?
Absolutely Steve. So there were a couple of things that were very interesting that I thought I could share with you and the audience. There were a lot of topics that were focused on getting more green and efficiency in the data center world and so there were a lot of talks about the recent New York Times' article as well. So let me start off with some of the new trends that we saw that were being discussed and we'll get into the New York Times' article so...
Yeah. So there used to be some kind of bragging rights on every new technology or every new certifications that comes up for these data centers and it used to be lead certification that used to be the bragging rights for most of the data center operators and developers. Now, it's becoming -- it's being replaced by EOE. Now, that is the new term that everybody wants to throw around and there's a lot of importance attached to it even within the respective corporations. So that is something that's catching on very well after about three or four years since Green Grid announced it originally as a __02:30__. We also had a lot of discussions on fuel cell technologies that are being deployed in data centers to act as the data center baseload rather than a secondary power source and with power being the supplemental source for our fuel cell technologies. So that was very interesting to learn about. Solar farms, we have heard more about solar farms in the past, but the reality is sinking in with the data center operator these days as to how to properly use solar farms in the data center world and some of the medium enterprises are using it to shave off peak loads so that they can avoid paying the utilities based on the peak loads because if you see how utilities charge data centers, say, even if you have a peak load for like 15 minutes in the entire month, that is what they're going to take as your consumption rate and the utility is going to start charging it based on that. So it's far more important for data centers to shave off those peak loads and solar farms are coming in very handy to offset that peak load.
Wait, so whatever your peak load is, they take as your rate and they make that a flat rate based on your peak load?
At least for that month. It depends on the utility and the municipality that they're located in. But that's the trend that the data centers are facing.
So, yeah. So, as we discussed earlier, there was the New York Times' article that came up recently that had a lot of attention within the industry, outside the industry and there were a lot of discussions in this conference as well to freshen their mind. What I realized is that there are always two sides to every story and the overall impression in the convention was that some of the facts that were represented in the article, they're kind of sensationalized for the broader audience which is understandable. The industry itself is collectively making many improvements and innovations to solve one the major factors of energy raised in data centers and servers that's highlighted in the article. That is to bring up servers only when they're needed, on demand, and as fast as possible. So that way, you don't have servers running all the time at idle load, but still consuming power. So there have been a lot of innovations done.
Hey Jawahar, can I just interrupt for one second just so -- those listening understand some people might have missed the article. Jawahar is referring to an article that ran in the New York Times and shows up on their website from September 22, 2012 and it was written by a gentleman named James Glanz and the headline is called "Power, Pollution and the Internet" and I think universally, the view of this article is that it was somewhat, how do I say it, not friendly in its view. It seemed to have set out with an intent to somewhat disparage the data center business. Go ahead Jawahar, sorry about that.
Sure. It's a great addition to that. Yeah, so the article focuses about all the waste that's happening in the data center world in terms of energy produced and energy being consumed in the data center, although the article is correct in terms of the awareness that it brings to the world, to the data center world, to see how they can solve this problem. The reality that's also highlighted in the same article is that in 2010, the energy consumption of data centers all together in the United States accounted for only 2% of the overall electricity consumption in the United States. So what they're really talking about here is optimizing the energy consumption of that 2% used in the data centers. So that's (crosstalk) things in perspective.
That's actually quite an interesting statistic. I hadn't heard that before the 2%. What was the buzz at the show about all these? Were people negative about it? Were they positive about it? Do they think it was a useful thing? I've read a lot on Twitter that some people think this was a good wake-up call.
Absolutely. The consensus in the farm both on-stage and off-stage was that, yes, they understand in the data center world that there are a lot of failings to be realized and we are trying our best to do that, but at the same time, the prospective is that this just amounts to just 2% of the overall energy that's being consumed in the United States. So the -- although there is the general agreement that there's a lot to be done, there is also the agreement that there are a lot of things that are being done even today and some them is to just make sure these comatose servers that they're talking about which basically all servers that are running and no one knows what they're doing or whom they belong to. So no one really wants to step up to it and unplug these servers because they assume that it's not doing any productive work. So that's where Raritan has provided some solutions like Power IQ, the management software, where you could do or you could identify these servers with outlet-level metered power strips from Raritan itself and potentially be commissioned based. So that's one of the ways to do to eliminate these comatose servers and there are other technologies that are being deployed and that are being designed as we speak that are trying to make these servers as efficient as possible so that they can go to sleep the moment they are not needed and they will wake up the moment they are needed back in servers.
And that's quite an issue because one of the things the article pointed out I recall quite vividly that really stood out to me was the fact that we consumers of the internet out here expected to be always on. We expect to be looking at whatever we typed in within seconds, milliseconds even, and the faster the internet gets, the greater the demand and now there's video and texting and all sorts of issues like this. So the demands for it are extremely high and the demand for data centers to have 99.5% uptime is extremely high. So they've got service-level agreements on there and it's just quite a mess for data centers, who really are, in my opinion, caught in the middle between the consumer and web services, are they not?
Absolutely right Steve. There were also thoughts about how to avoid these redundancies like some of the things the article points out like you mentioned Steve that the data center -- the power that's coming in to the data center is always overprovisioned just in case there are any kind of averages on one or the other and there was one mention about some of the customers that are considering what they called redundancy in the data center world to be just non-redundant and the entire flow erupted and the conversation was so focused on having like you said the SLAs being met and having disaster recovery plans in place at such or so. So even the mention of removing redundancy and having just the right amount of power and just the right amount of redundancy for data centers, this starts a whole lot of debate among data center operators. So you're right, this is not an easy problem to solve and there's much of an awareness around that and all the solutions and products that we put in place to realize these efficiencies are all in that context of how to make sure I run my data center as efficient as possible but at the same time maintain the expectations from whether it's the end-user community or the SLAs and liability.
Very important issues. How do you think this gets -- where does this get resolved and this might be a good segue into your presentation at GDCon as well. But where does this go? How does this eventually get resolved? I mean you mentioned some companies are looking for solutions to have on-demand power available. But if it's even a millisecond late, the game is up, right?
Right, right and especially on some of the platforms where people play online games, even more important on Wall Street where transactions are dealt in milliseconds and nanoseconds. So there are critical infrastructures that may never be impacted by any of these and there are things that could be changed in environments that are in as super critical in their demand response times. So as you said, Raritan has been in the energy management market solutions for a while and about six years back, we pioneered the outlet metered power strips called Dominion PX that will provide the most accurate and granular data consumption or consumption data for every single server that you plug into these power strips and at a short time from then, we introduce Power IQ management and reporting software and our customers have been using both these solutions to achieve things like __13:49__ to departments, cost-awareness reports to internal department or their end users identifying comatose servers that are running for no obvious reasons creating What-if reports so they can do a much better capacity planning as they move on and also...
Sorry Jawahar, what did you call those reports?
What-if reports, yeah. And also schedule some downtime for servers and equipment that aren't going to be used, let's say after 6 p.m. everyday until 6 a.m. in the morning so you can completely turn them all off and if you leave it up to the end users to do it, they may not do it every single day, but you can automate all these through our solutions so that you can realize a substantial savings automatically without any manual interruption.
Jawahar, what would you say -- what would you say the adoption of these kinds of solutions is right now within data centers around the world or at least in North America? There is obviously growing awareness that there's a solution available. What would you guess is the adoption?
The adoption rate has increased significantly since we entered this market. So just from our own experience what we just mean is, when we introduced these products in around 2006, there were a few enterprise companies that looked at this and go "this was exactly what we are looking for in order for us to achieve the energy savings we are liable or we have committed ourselves to do." Now, the rest of the industry was lagging and they were looking at it and go well, which all believe to have that much of data and granularity, but I'm not in any form or shape ready to realize those data or utilize those data to realize the savings. Now, fast forward about six years from 2006, now, we are at a completely different stage where we see a lot more acceptance in the market in terms of solutions that bring up energy savings and cost awareness to IT Department, to data center operators with which they can use those data to plan their capacity going forward. So we have come a long way as an industry to accept and adapt to the new requirements of not only just having all these servers up and running all the time, but also making sure the unnecessary servers and applications and devices can be either turned off or put to sleep as long as they are not needed.
Right. Now, on a website and I just want to make sure the listeners know where to go. That's raritan.com and under Products, they'll see Power Management Solutions. If they click that link, they can learn a great deal more about this. You've got some bullet points that talk about the combination of intelligent PDUs with Power IQ. It can give energy management solutions in which the features and functionalities will allow data centers to manage capacity planning decisions, use power resources more efficiently, improve uptime, reduce capital expenditures and save operating expenses and measure PEU. I get all of those with the possible exception of improve uptime. How does getting a hand along your power improve your uptime?
Improve my uptime? That's a great question, Steve. So there are a couple of ways we can improve the uptime for the individual devices that are being powered by our power strip. Number one, the meantime between failures of all our devices, the power strip, are anywhere in the 20 to 25 years' time period which is a lifetime in terms of -- a career time in terms of data center operators. So the person whom might have put these devices in the rack may even retire before any of these would ever fail and so, in that way, we can guarantee that the devices that are being powered by these power strips are running and the uptime is as good as the industry wants it to be. The other factor that as we said in conjunction with Power IQ, when these servers and devices are utilized only when they're needed and you'll be surprised that most of these devices may not be needed or even -- I don't know, maybe even like 60% of the time. So that improves the reliability and the uptime of those devices as well.
So it improves the lifespan of it because you're just simply using it less and putting less wear and tear on it, keeping the temperature cooler and things like that, interesting. Okay, so you did a very interesting presentation at GDCon, didn't you, called Vendor Showcase.
And my understanding, that is -- when you were doing that, they gave you seven minutes to discuss energy challenges the data center managers are facing and how Raritan Solutions can solve them. Did they really only give you seven minutes? Did you push it over that just a little bit?
They were a little more generous so they -- I think I went over about a minute or so, but overall, it was a good time to present and I had I think enough slides to just go over it. And if you like, I can quickly talk about what we went over in the conference.
Well, actually on this format, you don't have to do it as quickly. We're going to give you like 15 minutes or 20 if you want so that's a good time, enjoy yourself.
That's plenty, thank you Steve.
Go for it.
So what I really wanted to present in the conference was some of the unique challenges that our customers have come to us with and how we have created or customized our products and solutions to solve exactly those challenges. So number one, there have been request from customers for higher densities in the rack. So what they mean by that is they want to add multiple, many more servers on the same existing racks or improving density of newer racks where they used to have plenty of servers or 24 servers. Now, they want to load all the way up to 40 servers and a few more switches on the top of it. So, these are higher densities that are coming into the data center. But at the same time, they do not want to increase the capacity to provide power for these racks, for these servers. So, what we did was we created a rack PDU, the Dominion PX series which can have all the way up to 54 outlets on a very slim 2.1 inches wide __21:47__. So this pack involves a number of outlets anywhere in the industry and at the same time keeping the capacity the same as what the customer had before. This gives them the best mix of increasing the density of the rack while keeping the capacity the same as what they had before.
Okay. One of -- each one of those outlets is intelligent and has a chip behind it and feeds you quite a bit of data, right?
That's correct, so you would be able to -- even though you have up to 40 servers and three switches on this rack, you would be able to draw information about every single server and how much power is consuming at this particular moment or like you said, that dedicated energy metering that they have for outlet. You can also transfer those information to see how much power and energy this particular server has done for a period of time whether it's a week or a month or even a year. The other interesting channels that we have come across quite a few times recently is the right sizing as the industry calls it. Some of our colocation customers wanted to utilize most of the capacity they pay for and do not want to leave any of the capacity on the table. What that means is if -- as our audience may already know, most of the colocation facilities charge customers for energy by the capacity they provide rather than the real-time consumption per month or per year. So that leaves the customers to say "Well, if I am paying for a certain capacity, then it's up to me to utilize the most as much as possible so that I don't leave any money on the table." So what we did for a customer like that is to create power strips that will be rated for the maximum available rating allowable for that particular __24:03__. While the most of the industry has chosen to stay under a certain rating to just minimize the PDU cost. What it leaves the customer with is a less than optimal solution. So, our solution of having the maximum rating regardless of the capacity needs that's allowable within NEC and ANSI guidelines while providing at a reasonable price and made sure they do not need to create far more cabinets in their data center or in the colocation facility and save a considerable amount of money and effort in bringing all these up.
And, we also -- I also talked about large-scale deployment which we have a tremendous amount of experience that when customers deploy thousands or even hundreds of these power strips in one time, that is a lot of devices to be set up with their credentials and they also -- they likely and also verify that they have been set up in the right place. Apart from the credentials, they may also want to set them up with certain policies and alerts. So we came up with a very simple solution of having it done through a USB stick. So what the customer needs to do is provide the USB stick with a list of PDUs and power strips and their corresponding credentials and policies that they need to have and put this in the thumb drive of the USB stick, all the operator has to do is go across every single PDU, plug this USB stick on them and after a while, the PDU will read what's in the stick and configure itself with the credentials that has to have and it will also put a smiley face on the screen to just make sure you know it went well and it took the light credentials...
I'm not kidding when I say it does put a -- it literally puts a smiley face on the screen.
I love it.
So that is the -- that is hundreds of man-hours along with thousands of dollars saved from just setting up the PDU. So that has been very invaluable for many of our large enterprise customers who constantly do exactly that just bring up hundreds of that at the same time in different places. We also talked about how we can eliminate or reduce the number of these comatose servers that the New York Times' article is __27:03__. Customers have used the Dominion PX outlet-level metered PDUs in conjunction with Power IQ to go ahead and identify those servers that are drying power, but has not been going up or down in terms of power consumption. That kind of gives a better idea for the customers, say, well I have been tracking this particular server for over for about a month or so and I do not see any sites or any depths and the power consumption of the facility. The back data if they can combine it with the CPU authorization data from other sources, they would be able to identify that as a comatose server and they can be commissioned at when it'll be appropriate time to do so. And also...
And also, some of our lab customers have used Power IQ to go ahead and just straightforward shut the servers down when they are not in use. So, if you look at the typical lab, any of these IT organizations or the right manufacturers, they have people coming in at maybe let's say 8 in the morning the engineers stop their work and they leave home 6 o'clock or so in the evening, but the devices they are testing, they all are running all through the night, some 6 p.m. all the way until 8 a.m. in the morning when they come back. And the labs will realize "Well, what if I turn those things off? Even if I can turn every single one of them off, still I can go ahead and identify the things that the devices that can be turned off for that particular time period when nobody is using it. And thereby, I can realize the considerable savings." So...
We have customers that have realized all the way up to 40% of savings on energy use just by implementing this policy that will automatically go ahead and turn things off in the evening and turn them back on in the morning.
Somebody like that are just no-brainers right? It's just that they haven't thought of it or didn't know it was possible.
That is for sure and also, there have been some push backs that they have to overcome before they could implement these policies and that's very, very much understandable where they want to make sure these servers when they are turned off and when they are turning them back on, they do come back on. So, what we did in Power IQ software is provide a feature called "graceful shutdown" where the servers are not going to be just turned off by just cutting out the power supply to them, instead Power IQ will communicate with these servers first and shut them down before turning the power off. So in that way, __30:18__ and lab managers or engineers are very comfortable that. When these servers are turned off in the evening, when they come back in the morning, they will come back off and they will...
And they will be in progress.
I like that phrase "graceful shutdown". That's very nice.
Yeah, so those are some of the things that we have been told in the conference and they had some very interesting questions following up the presentations. So, overall I would say it was a very good presentation and very well-received.
Now you have a slide, I happen to have the slideshow in front of me that you have presented. You have one slide that really stood out to me and that is slide seven which is what Raritan thinks you should get from a rack PDU. Can you talk about that a bit? You have got energy metering, power control, asset management. It's quite a few demands you're putting on a rack PDU, environmental monitoring, video security?
Yup! Absolutely, I would be glad to do that and by the way, thank you for bringing that up. I almost forgot to mention that. So this is one of the unique challenges that we were in a very good position to solve where customers have realized that every time you provide a network port for any device for that matter whether it's a rack PDU or a server or a switch, there is a cost associated with it and that cost happens to be anywhere between $350 to $400 depending on the network architecture or depending on the individual customers. So that's the range $350 to $400 for every single network connection provided in a data center. Now...
Now, we multiply that by 100 or 1000 or just for the power strips alone. That's quite a lot of money. So what we realize well, okay so if that's the cost associated with the network port, how can we minimize the paying for the customer by doing a couple of things? One, we wanted to cut that cost by a quarter, so that it becomes somewhere around the range of $100 for the network connection. So, what we did was to take out one network port that's coming in to one PDU and we would be able to share it across all the way up to four PDUs. So that minimizes the cost for its connection already, while in the same time not losing the identity of the individual PDU. So the PDU can still be addressed individually, but they do not necessarily need to have a separate network port for each one of them and we achieve that through USB connections between the PDUs.
The other thing we did was -- okay, what if we take this and resolve this cost for network connection even further. So we looked into Wi-Fi. So now with can plug in a Wi-Fi connector to the PDU. So that will eliminate a need for a network connection, a hard-wired network Ethernet connection to every single policy. So that's another thing they implemented which can resolve the cost even further. On top of it, you realize -- okay, Wi-Fi may not be for every customer and the cast stating that I was talking about through USB, to share the network port may not be for every customer either. So, what can we do for customers that would still need one network connection per power strip, but still would like to maximize the benefits of providing that network connections. So typically, if you all at the power system in the market, you may be able to get power information, power consumption information and some level of outlet control and also some substantive negative data. So we thought, we could do a lot more than that for the customer. We need to take our PDU and make -- if you know the brain of the rack itself. So, here comes the profit of a small track, there is a lot of intelligence now that send to the PDU that surrounds the rack. For instance that can give you consumption data, energy usage data and on top of that, we can gather information about the temperature, humidity and also various other sensors that we can implement within the PDU. So the PDU can have up to 16 different centers connective tool. Some of them could be air flow, temperature, humidity, differential air pressure, and contact pressure. So those are things that can give you a very realistic view of the rack in terms of all these different data book. So, you can either play hopscotch, the movement they occur. You can identify leakage/error, the moment they occur. On top of it, we realize -- okay, why do not we take this one step further and have some security build-in as well. The best, the other video security exhibits next in the slide, so you can plug in a standard off the shelf web camera into the PDU's USB port and right away you have a good view of that pretty __36:29__ rack of the role that direct to them. And lastly...
And lastly, you do not need to implement a separate system for the video surveillance. If you needed just for a certain ride or certain role or even across the internet data center at different levels.
I think that's great.
And then they started working with the EBay who asked us to design something for their asset management and that's why we came out with the asset management strips which are non-office based and they need -- they are based on proximity and connection where you can plug these little tiny magnets, that's how the __37:18__ ID including the asset management strip that can communicate the presence or absence of an asset to the PDU and the PDU can take that information and propagate that through FNNP which could be listed by any number of change management systems or a systems like this in trust, for instance that's why I listen to __37:41__.
If you don't mind, I want to point out to the listeners again, if you go to Raritan.com under products, they are going to the power management and if they look over on the right-hand side, you can actually see a case study, that Jawahar's referencing about the EBay solution and how much -- and basically, you help them cut their power cost in half, I understand, is that right?
That's pretty amazing for a company of that size with that much power consumption. Their power consumption must be tremendous and to cut that in half is quite an achievement.
Yeah, so that's the one of the -- like you mentioned Steve, that's one of the flags that we went over in the presentation which gave array for the customer to realize a lot of savings on the capital expenditure and the date put in place just by maximizing the amount of data and in the household that you can obtain from having this intelligence and the network connection to every single rack PDU of every single rack.
Well, that's great! That's great, I'm sure that reviews are tremendous too that you hearing a feed on Twitter quite bit and this is also quite a move just to start to wrap up it too. This is a quite a move you guys started to make from six years ago, where Raritan was known as a KVM company and now here you are coming out with tremendous software solutions and power solution and PDUs. You really going to -- you're becoming known as much as a software solutions provider for the data center. Aren't you?
That's right. But our decent solution of the DC Track and the Power IQ for power management and energy management. It become a long way in terms of the capabilities that we can add for the customers tool set to realize and visualize the internet data center rather is from an asset standpoint as to figure out where any particular asset is or for a capacity standpoint as to how the individual rack deployments are doing or it could be from a physical __40:10__ standpoint where we have more space to put more service instead of building your racks and so on. So, yeah these software capabilities are in -- works very well with our power strips and __40:25__ solution and also independently in the some of case within the history.
Well, Jawahar thank you very much for you time today. I think this has been more and more informative shows we have done in a while and it's been very interesting and I'm really glad you can join us and take us through what you showed that GDCon.
Well, thanks a lot and I remind the listeners that they go to raritan.com, they look onto the products, navigation bar at the top, scroll down to power management and you will find what they need in terms of managing and monitoring power usage in your data center. Thanks again Jawahar!
Steve, it was a pleasure to be on your show and thank you very much for having me on and as usual, you do a fantastic job.
Thank you Steve, have a good day.
Sorry we couldn't complete your registration. Please try again.
Please enter your email to finish creating your account.
Receive a personalized list of podcasts based on your preferences.