00:00
Welcome and thanks for joining us for three steps to managing a growing radiology practice sponsored by pure storage and welcome to Raleigh's Research Triangle, the home of Wake Radiology, UN C Rex, a leading radiology group that operates 14 outpatient imaging centers and provides specialized reading services for five hospitals. I'm Mary Tierney, the Chief content Officer of Radiology Business and the moderator for
00:26
today's webinar and live Q and A. As we all know, growth can be both exciting and challenging, especially here in Wake County, the fourth fastest growing metropolitan area in the United States. But despite the challenges, this 70 year old practice has managed to grow and thrive, becoming twice the size of its closest competitor.
00:47
So how have they done it by focusing on three key priorities, high tech, high touch and high efficiency, wig radiology, UN C Rex invests in state of the art imaging and it technology including artificial intelligence and has a strong focus on providing a high touch experience for patients refers and hospitals. They also prioritize radiologist and practice efficiency with the goal of driving down costs
01:15
and increasing accuracy, productivity and value. However, even successful groups run into snags in this case, a need to upgrade their packs due to a lack of IOPS. So they needed more storage and that's where pure storage comes in. Wake radiology, un C Rex, Cio Nat Dewey did the research and knew that flash was the
01:38
answer with his vote of confidence. The group's leadership, including president and managing partner, Brent Townsend and radiologist, William W and Matt Hoyes and others made the decision to partner with pure A move that has helped them to boost performance scale up over time as well as reducing costs. So let's listen in to hear about wake radiology, Un C Rex's journey and the lessons they've
02:04
learned along the way. And then you will get a chance to ask the team some questions at the end of the webinar. So feel free to pose your questions any time using the Q and A feature on your screen. So, Doctor Way our attention is yours. So Chief Medical Officer, one of my jobs was to ensure that we could run seamless operations 24
02:23
73, 65 across a very broad geographic area that encompasses multiple counties at this 0.5 hospitals in about 12 imaging locations in a highly specialized subspecialized environment. Um uh for radiology interpretations um quality is paramount. And uh we want to be sure to be able to get the right images to the right people to read in a timely manner so that we can add value to the uh chain of patient care.
02:51
And um part of that is to ensure that images are available. Um The infrastructure that supports the dictations, the reports, the distribution of those reports is always functioning at a very high level across a very broad network. We support multiple modalities and the the amount of data that is running across our networks is is is great.
03:11
And so the infrastructure to support that has to be fast has to be reliable. And uh our radiologist need to count on it to uh provide them with the information they need at the point of service for us as radiologists, the number of studies that we are asked to interpret the number of images per study and the speed with which we have to turn around those studies is always making us push the boundaries of what we're able to do by having
03:36
the latest technology. This increases the efficiency of our radiologist improves the accuracy of our reads and makes sure that the patients are getting what they need when they need it. And we're getting the right answers for them. A uh candidate pool to hire new radiologists is diminishing.
03:53
So radiologists, not, not only reading more images per study, but we're also increasing the demands upon the radiology itself to be more productive because uh of the difficulties of nationwide in recruiting and hiring new radiologists. So it's become very important for us now to optimize our efficiency as best we can across all services that we provide.
04:15
And in turn, that requires that all the tools we use, be very reliable and work in a manner that is seamless. I've got so many different customers. Right. I've got the radiologists, I've got the referring physicians and I have the patients. And um, and so we have to think about the best way to make things work for all of those
04:34
constituents. Right. And then of course, I have, I have our internal customers uh for, you know, all of the employees. So um we have to think about all things uh it related from all of them. So my expectation for the workflow uh proper workflow is that, you know, when I open an exam that the data is immediately available to me um with the current
04:58
and the prior studies. Um and I'm able to, you know, go through the study and interpret it and make my comparisons and then issue my report so that I have uh you know, timely um in a production of, of a clinical report for the referring physician when we were looking at doing an upgrade to our packs. Uh The vendor said your IOPS are not where they
05:20
need to be. So uh we had to look at different systems um that we were using in a, a legacy system that was mainly spinning disk at the time. And uh we needed to do something that had uh many more IOPS than what that could do. Um, so, uh, we looked at a couple of different vendors and, um, pure rose to the top. And one of the nice things about that was that,
05:44
uh, they did a good job of managing the system for you, right. Uh, the software would take care of it. Um, and they said that it would have a really low, uh, total cost of ownership. And, um, besides the fact that it was really, really fast,
05:58
our previous vendor did have a system as well. But the the cost of it was much more pure was brand new to, to me and to wake radiology. In fact, nobody on my it team had worked with pure um prior to purchasing uh this, this flash, our job is to interpret studies and get the reports out to the referring clinician so that patients can get the care that they need if you turn on that
06:27
computer and the study isn't there, that's slowing you down. If you go to dictate and the system isn't working, that's slowing you down. So by having a robust infrastructure that constantly make sure that these systems are working and functional, constantly upgrading them to make sure that they are the latest technology that allows us as radiologists to do our job better.
06:50
We are able to quickly make changes that allow us to adapt to changing scenarios. For example, suddenly having 20 radiologists reading from home instead of a centralized location was an added challenge for our it staff. But because we had the infrastructure in place, we were able to support those radiologists, making sure their network connections were
07:14
strong, making sure their security was strong, making sure that they could read when they had to read without any hiccups. My radiologists are very expensive. Right. They are the most expensive part of my whole company even more than my modalities. The pack infrastructure is what drives everything that they do.
07:34
We can't have downtime. Our flash storage is supporting our 14 imaging centers and six hospitals with up to 60 radiologists um that are reading on our system including uh systems that are at their homes for on call 24 7 is how often this system is running is always going supporting these, these hospitals and our imaging centers all the time.
08:01
We also are supporting er S 24 7 and we have to be up all the time, even patches and upgrades are thought about when they need to be done and to minimize the amount of downtime that is going to be happening. Um and, and affect our hospital partners. We have all of our servers backed up in two separate locations such that if one goes down,
08:25
the other can immediately come on. So the radiologist can keep working flash is a game changer. I mean, it, it's been around for quite some time. I had always thought of it to be too expensive. In this case, it wasn't the the other thing that was really interesting to me is that many of the vendors had said you don't need to buy as much flash as you would traditional disk.
08:47
Um And one of the reasons for that is because flash is so fast, you can do the duplication and things like that that they talk about and you will see a difference. Flash for radiology should be a requirement because the radiologists are your most expensive thing. You have to get the images to them as fast as possible.
09:05
And it makes such a difference in speed that there there really is no other choice. An image is acquired at a particular location. It runs the C AD system in a separate uh database, but then rolls back into the infrastructure that supports, that underlies the network that they can then deliver that to me in the reading room for interpretation within a few minutes uh of acquisition of those images.
09:28
So it's, there's there's a lot of complicated processes that go into that particular modality as an example that uh require speed of multiple components. Um a large component which is delivering just the base data to us. The ability to, to grow and scale is important to us because we're constantly adding additional studies in hospitals and things to our workflow.
09:51
Um And radiologists, we have to be able to have the headroom to have all of the systems perform at the level that we need and, and we can't have them uh sitting on the borders of uh you know, the ability to be up or not. Um So we have, with our pure system, we have tons of headroom uh related to IOPS, we're, we're not even close to uh where we have to be
10:19
concerned. The configuration that we use to read is a very interesting and complicated process. In the Greater Raleigh area. There's a lot of movement of patients amongst multiple different facilities including the UN C facility which uh uh many years ago deployed a very sophisticated VN A that captures all of those images. But even the hospitals that are on the UN C
10:41
system up until very recently have been operating in different silos of packs for their um for, for their, for their patient database. So that's part of the challenge is they're not immediately available to us and our packs that we read from on the outpatient side. And so uh we have very sophisticated process in place to retrieve those and present those to
11:00
the radiologist at the time of interpretation. Uh And, and as a consequence, the number of agenda that I issue now to compare with prior studies has almost gone to zero because we make sure that our uh priors are available uh at the at the point of interpretation, which is shortly after acquisition, a recent uh deployment of a new P A system for the university uh does a lot of that in
11:25
automated fashion, but they still use a separate packs than what we use in our office practice. And so there is um uh cross talk, so to speak between their pack system that runs off the same VN A that our pack system runs off of so that we can gain access to those images and our individual packs. But that requires the infrastructure to be able to find them,
11:45
pull them, aggregate them and assign them to the appropriate patients. So that when I open up an exam to read all the studies are there uh from that patient, irrespective of where they have been seen in the UN C system. So it's a wonderful database. Uh And uh the infrastructure supported is, is, is unmatched when I'm wearing my hat as a clinical radiologist and I am reading those
12:07
studies. What I want to do is get through that list so that I can help my patients, that I can make the clinicians happy having a strong infrastructure in the background that allows us to read faster and more efficiently is what matters the most. I'm happy to know that I have that support behind me, but I don't think about it on a day to day basis and that's a benefit.
12:26
We don't have to pay as much attention to the system as we did before. It notifies us. The engineers notify us if there's a uh a firmware upgrade that we have to do. Um And uh it's pretty much hands off whereas our previous system, we had to pay a lot of attention to, um, and spend a lot of time watching the storage and
12:46
seeing what's happening with it. I've got two system engineers um, that maintain my, my systems across the board. I've got, um, you know, hundreds of servers that they have to pay attention to. And prior to this, they had to spend a lot of time thinking about storage,
13:07
how it was being used. Um Now we have alerts to tell us when there's, there's going to be a problem or uh you know, early detection, but they still had to pay a lot of attention to that um with pure, the amount of time that they have to spend managing, that has gone down drastically. In fact, uh you know, I recently asked them, you know,
13:28
what kind of time are you spending? And they're like almost nothing. It just, it works so fundamental to the practice of radiology is having a, an infrastructure that supports current technology and all future anticipated technology that may be coming down the line that we may not be aware of. For example, we, we have uh we've taken advantage of um vendors that supply vetted A
13:53
I products so that we can then test them and then deploy them as uh uh as potential tools for us in the future that I I feel certain are gonna grow with time and the need is gonna grow with time. But in order to be able to do that, we have to have the infrastructure that can incorporate those new technologies uh without having to re invent the wheel.
14:12
So we do have an artificial intelligence package um that helps us by uh preprocessing the data and essentially um highlighting those nodules that are important that makes the review uh much more confident um for the body radio and provides an extra layer of security to us. And also the patients right now, our priorities with artificial intelligence are focused on the breast imaging program and the
14:42
lung cancer screening programs. Artificial intelligence increases the sensitivity and the speed with which we can read those studies in the future. Artificial intelligence will help with newer radiology. For example, looking for early onset Alzheimer's disease. Other applications in the future may include bone density screening,
15:00
assessing someone's risk for osteoporosis. One tool that we have found invaluable for A I is one that we implemented about three years ago, which was a spatial localization algorithm that allows us to synchronize current and prior studies and uh to to be able to spatially localize two or three or four examinations simultaneously on the desktop. Because of an algorithm that's been applied to
15:20
the data, the data set that you're looking at saves an enormous amount of time. What's very important from an infrastructure perspective is making sure that all of these new technologies, including artificial intelligence are well integrated into what the radiologist does on a day to day basis. So as a radiologist um in a, in a group that is growing and expanding, you know, it's important for us to know that,
15:46
you know, in the background. Um All these it instructure are being taken care of to meet the ever growing need of uh you know, big data and the increased volume in imaging studies. Um you know, so that we can get our job done every day and have a very reliable, high quality product to provide that high level care for patients. They have a good path forward,
16:14
right? They, they've thought about it, they've designed the whole system with that path forward in mind. Um So you don't have to do the forklift upgrades that have had to happen in the past. They, they just get it. Um I'm Mary Tierney and you have already met our expert panel. Um and all of them are live with us this
16:34
afternoon. So, so thank you guys for being here. We appreciate that. Um Nice to see you. Um So, so quick thing. So um everyone can feel free to continue to, to ask questions um in the box. Um And we will share them um as we go on for
16:51
about 15 minutes or so. So we have a bunch of questions that have um we get some pre questions that have come in as well too. Um So, so just keep go ahead and share as many as you want. Um, just as long as we're, we're able to chat, um, which will be for probably about 50 minutes or so. So, um, I do, I do see, I think, um that do,
17:10
I think you're going to be a very busy man because there really are a lot of questions uh specific, more technical and things like that. They're talking, talking about um speeding up packs in enterprise imaging and things like that. So, uh so let's jump in. So, um Matt, I hope you're gonna take our first question. Um What should radiologists know about flash
17:29
and how does it work and why does it matter to us? So, from a radiologist point of view, uh it's a major difference from uh the way that spinning disks work that used to be in normal computers. Um It's using uh a completely different technologies much closer to say how a USB drive works, right? Those are, those were called flash drives in the beginning,
17:51
right? Um And so it writes to chips instead of writing to magnetic disks. Um The difference in speed between the two is about 100 fold in a minimum. Um When you, when you start to talk about big systems, when you start to talk, uh you know, at your personal computer level, uh i it's much more than that uh you know,
18:12
a thousandfold Um but when you have larger systems, uh they spread the data across uh multiple disks. And so in the old days on spinning disk, you could get them to be much faster than one individual spinning disk because of that. Um the the cost is more than the cost has come down a lot. Um And it, it makes a huge difference in how fast you can get the data off.
18:35
Um And, and available into other parts that might be slower such as the network, but it still gets, it gets the data faster to the radiologist, the name of the game. Absolutely. Um So, and maybe you could, you could continue on with this question. Um which talks about um how much total pack storage do you have?
18:52
Is it all flash or are there also some spinning disk? So we still do have some spinning disk though, we use uh caching mechanisms to bring it off of the spinning disk onto uh into the flash for the radiologist. Um So total spinning disk. Uh we have about uh half a pet petabyte um from a flash point of view, we don't need as much for that particular reason.
19:18
Um We've got about uh 80 terabytes that we use. Um that's, that's available um to the radiologists and we just, we only process um about uh two terabytes of data a week. Um So you, you can see that we'd have quite a bit available even with the, the 80 terabytes we have, right? Ok.
19:39
Um And maybe Doctor Wei and, and Matt, um maybe you can, you can both chime in on this one. when you knew your pack was out of. Iops. So I guess what we're talking about in input output operations per second, right, of talking about speed. Um Where and to whom did you seek advice on the available options?
19:55
So, we talked to a couple different um vendors that we have bars. Um And uh they actually pointed us in the same directions, uh which was kind of interesting um what the vendor that we previously had. Um So we, we looked at them, uh and we actually talked to them about all their different systems that they had, they had multiples that they sold. Um And then we talked to pure.
20:19
Um and then there was a few others that we talked to up down the, the value chain. So there, there were some that were uh you know, not top tier players that we looked at as well. Um And they were up and coming. Um And uh so, so there were, there were quite a few different people that we looked at. Um Did you? And then I guess the other piece of that is,
20:40
is, is, you know, were there other people that you were other, other um organizations like your own or anything like that people that you were reaching out to other cio s and things like that to get their opinion. So I had talked to some, um, many of them had used the same system. I mean, there's, there were only so many, uh, to choose from.
20:57
And so, um, so I did talk to different ones that had used different systems. Um, and there were a few that were on pure and a few on, on, uh, the ones we were on. Um, and so I, I did get some feedback on that as as well. And let, let me add, you know, our, uh uh we're radiologists and we don't really understand the
21:16
infrastructure, the, the ins and outs of the infrastructure. And so we rely very, very heavily upon our, it staff to um to uh to, to make those judgments and meet on a regular basis with uh the it staff and the management practice to understand the costs related to the various options that, that are available to us and how they're gonna able us to do our business.
21:36
But uh and Matt's got a great team of people who um uh really understands the technology in a way that we could never master and uh just makes it, it's, it's, it's imperative to an organization to have that kind of uh uh knowledge and support uh that they can bring to the table to enable us to do our job. He understands what we need to do and we rely on him to take that information and work with.
22:00
Um let's see. Ok. Doctor Doctor Townsend, doctor hos doctor, we, I'd love to get you guys part of this conversation. Um So Doctor Townsend, maybe you can start us out on this one. Can you talk about how your group is using A I I know you did mention breast and chest um and how you see that evolving over time.
22:19
Um Happy to start and then I'm gonna let doctors homes in a way kind of run with it. I'm a pediatric radiologist. And, and so when I'm reading in my subspecialty, I'm less, I use less A I than say the body radiologist. But when we're doing, say lung cancer screening, I know that's been very helpful in terms of uh increasing confidence of reeds and increasing accuracy of reads.
22:39
I know that our breast imaging section has been very pleased with the artificial intelligence we're using to assist in that field. And I know that our neuroradiologist are currently vetting several A I products to help them with uh getting more information to the neurologists and neurosurgeons that they need. But again, let me, let the uh ones who use it on a more day to day basis speak to that.
23:01
I'll defer to that. Hi. Um So in terms of A I, uh one of the products that we're using in the body division is lung nodule detection. And uh we um have it so that the data gets sent and preprocess um by the vendor and then comes back to us with the nodules highlighted. Um We um worked with them to streamline the process so that it was seamless for the
23:31
radiologists. It, it interrupted us. Um you know, the least uh so that we didn't have any interruptions in our workflow. The nodules came to us like automatic um highlighted and um we can generate the report, um you know, much faster. Um We did some analysis and we've, we've actually reduced like turnaround times on those reports.
23:54
And um the, the A I is uh you know, just really there to um help uh boost our confidence and um uh sort of as a um you know, safety nets for, for the radiologist to highlight those nodules that are in tricky spots of the anatomy, like the central regions of the chest, um or uh the uh you know, inferior portions of the chest.
24:21
Um So it's, it's been, it's been uh a resounding success um in, in the body division. Um And uh we are looking at other products as well. Um There's a lot of products that are coming on board um in terms of FDA approval. Um but a lot of them are not necessarily, you know, helpful for every single application that
24:43
we need. Um So we, we look them at a, at a case by case basis and, and make that judgment, um you know, as needed. Um There are some new products coming out that uh help with um looking at triage. So if, if there's uh emergent um pathologies like a, a bleed in the brain or a clot in the um vessels that supply the
25:11
lungs, uh called a pulmonary embolism. Um The A I can actually search for those pathologies ahead of the radiologist. Um and then populate those studies at the top of the list. Um So that the radiologist look, looks at those first. Um So those are like list triage A I algorithms.
25:31
Um But right now we're really looking at screening and detection of, of cancer, right, two other barriers that we're, we're using, we're using in breast imaging which, which uh it, it, it's very valuable in, in two different settings. And one I think that that applies here to this discussion is that we,
25:49
we do, we use, we apply it to diagnostic mammography in real time. And so the, the the steps of the images have to go through to get from us through the algorithm back to us and then present them to the radiologist is pretty important to have that in, in, in a, in a very rapid sequence so that we can um uh keep our diagnostic mammography volume, uh patient flow uh uh uh moving.
26:12
And uh we're also using it for 3d localization across modalities. And, and we do a lot of oncologic imaging where patients have got numerous prior studies and to compare them uh with um the current examination with one or two or three prior studies or with a different modality. For example, CTS and MRI S, it can localize the anatomy three dimensionally and allow you to link examinations in three
26:32
planes. But it, it, it finds uh fiducial markers effectively in, in an atomic markers in, in the patients uh to link up those examinations. And that's, that's been an enormous help when you're looking at two or three or four priors to determine growth or recession or regression of, of tumors um across time and very vague. Great, great.
26:56
Um OK, so let's, let's jump into a question, I think um I think, you know, Doctor Way, you've been involved in a lot of these decisions. Um Perhaps Doctor Townsend um and, and Matt, maybe you can listen into this one. You talked about your team's administrative time savings, like most hospitals, our budgets are very thin. We were able to gain savings by this conversion
27:14
in other areas. Have you been able to do any ro i reviews or I guess, can we, can we talk about um you know, the practicality, I guess of this as well as far as the, the financial numbers and things like that, of whatever you can share with that? I, I will have to preface is saying that, that, that our, our practice operates independent of the hospitals.
27:32
We're in a, in a partnership with the hospital. But, but uh we're independent of the hospital budgets and um and procurement process. So we're a lot more nimble on our feet as far as the uh the ro I I'd have to turn to Matt to um to address that more specifically. But I think one of the things that he said in the in the video was that the amount of time that he has to spend paying attention to uh the infrastructure is reduced.
27:58
So we capture it. It, it may be expensive technology, but we capture a lot of that in, in time and cycles for other projects. So I turned over to Matt to address that. Yeah, it's a little harder to talk about the RO I related to the time we're not spending any time on dis compared to what we used to spend, right?
28:16
Um But I haven't and our volume has increased considerably. Um So we're doing more work as a group um with no, with no time spent towards that, you mentioned no longer worrying about downtime or disruptions. Do you mean there is no downtime or do you mean that it is shorter? We've had no downtime with this disc. So um none,
28:41
none at all. Um The patches and things that we've done have re not required us to do uh to take anything offline. Um So that doesn't quite mean that we had downtime related to the previous disc though. I mean, I, I I'm not inferring that uh the, but we did on firmware upgrades uh on the previous system, we did have to take it offline.
29:06
So that, that was the only time that we would ever do it and, and it was major, uh major firmware changes uh because uh enterprise disc for a long time has been set so that you really shouldn't have any downtime with the, with the disk subsystem. Right. Right. Ok. Um, let's move into another ma of course, you're gonna continue.
29:26
So, and everyone else chime in, please. Um How much improvement if any has been seen with the introduction of flash um in the use of voice recognition software, do you find any of your dictations? Reports are gene are being generated or stored any faster compared with spinning disk? Haven't you noticed any changes? No, I mean, voice wreck has been around for so long and it was,
29:46
it, I mean, it was written way before uh flash was even thought of. Uh So at least from, from a consumer point of view. Um and it performed there. It's, I mean, nuance is no known to perform it, you know, 100 milliseconds and flashes so faster than that you spinning disk is fast enough for that. So I really didn't see anything there.
30:08
A lot of that is done on the local machine and also not on uh on their service stack. So, um no, we, we didn't see any difference there. OK. OK. Um There are a couple of questions um coming from um the, the establishment of this, of this new system, right? So how long did it take to convert to your new
30:28
S A um planning stages? Poc stage implementation. Can you reference some of the lessons that you learned in the conversion process? So, Poc was really fast, they, they put it in place. Um We were able to transfer we, everything we had was uh and still is uh is virtualized or ma majority of a few machines are.
30:47
Um And so we could easily move them over just using the motion. Um We were able to test it very quickly and uh so it, it, it, it did not take us much time to put it in place to make it go. Um And the, the, the thing that took the most time, I think there's probably a 2 to 3 day install time, which was mainly their time and then training us on it.
31:07
Um So, and checking it and making sure that everything was ok. Um And the, in fact, I've got a new uh admin who's been with me six months now. And, uh, you know, I asked a question related to statistics on the pure and uh the other guy said, oh, I haven't even shown you that dashboard yet and that tells you that you don't really need to look at it. It really just works right.
31:30
Uh When you and he has deployed some disk from it, he just didn't know because he was doing it from the VM Ware side and, and it just sucked it in and used it. Um, so we didn't have to go to any other dashboard to make it work. Ok. I don't know if that answers the question, but I think it is. Um, so let's see it next question.
31:50
Um, for Matt and all really, um, are there any technical issues where ST, so this is pertaining to A I? Um Are there any technical issues where studies are being read prior to the A I data being presented? Essentially is the imaging, imaging and data available? When needed?
32:05
Are the re read records accurate given A I intervention at the time of reading? You guys want me to answer. Well, I know one of the, yeah me, we, we set up a timer and one of these things so that the images wouldn't be available until the cycle had been completed for it to go through the algorithm to come back.
32:23
So there was a delay built in intentional for that. There may be some other things that Mat Yeah. In, in the beginning, we needed to do that. We haven't needed to do that really in, in quite some time. Um The and, and as doctor we said, the, you know, we're is for
32:41
mamo, uh we're fast enough to route it, get it to the algorithm and get it back to the rad before the tech walks down the hall. Um Matt was any consideration given to moving everything to cloud. Yes. So we did look at cloud. Um, and there were a few different things we looked at related to cloud. Um, the price point for cloud was still really high.
33:04
Um, there's if, if you didn't need to retrieve it quickly, then it wasn't bad or if you move the whole packs into the cloud, uh, then it wasn't as bad. It was more expensive but it wasn't, it wasn't uh something that would offset completely. Um, we already have two data centers um, that are inside of our buildings that we can't get
33:23
rid of. They're, they're surrounded by offices and such and I can't even switching them to cubicles or other offices doesn't work because of the way air conditioners and other things. Uh The, the cost would be more than the space I would gain. Um So that's one reason we didn't go with cloud. It's just there, there was too much expense that uh we would have in order to do it for no
33:43
savings. Um So, so we, we did do that evaluation and, and I continue to evaluate cloud um, for packs. It did, there's reasons that it would make sense for some places and, and reasons that it doesn't make sense for people like me right now today. Um That's not to say that I, I don't keep looking.
34:03
Ok. Um, so Matt, um maybe doctor way I don't want to chime in here too. Um How did you determine how much flash you needed? So, uh, the vendors came to us with different workbooks and said this is what we expect that you need. And I went, oh, I can't believe that's right. Uh, that's,
34:23
that's too little, is, is what I said. Uh, because I didn't expect we'd get the compression rates that they said we would get, um, images don't compress very well. Um So I really didn't expect to get much there. Um And I, I got a um uh a screenshot just a second ago that said we actually have 4.4 to 1 compression. Now, this is on uh pack stuff and uh window
34:49
stuff, everything, all my infrastructure um that's running on it. So, you know, I, they, they estimated I get 2.5 to 1. Um And they said most people do better than that, but because they knew I was going to be putting packs on images on it. They, they didn't want to overestimate. Um And so that's, that's how we estimate how much we need.
35:09
It runs the majority of my systems. So my whole VM ware farm runs on it. Um And then, uh which includes my pack servers and it includes my DNA. Um Let's see. Uh Maybe Doctor Townsend and Matt, um Maybe you can take this question. Uh My p vendor didn't call out the need for all flash.
35:28
But I'm now hearing from more organizations like yours that have replaced with flash. How did you justify the cost? Do you want to take it or you want? Uh Please go ahead now. So the, the, the way we did it is first, we didn't have to buy as much. That's, that's one second. The I could get there with a hybrid approach,
35:51
but it wasn't gonna be, it was still gonna be very expensive. The, the cost to go with the flash because I didn't need to buy as much because of the data reduction. Um actually made it uh the better and more attractive choice um because I, I was not gonna get that data reduction uh with the spinning disk um at the performance level that I needed to have.
36:12
So, um so that, that was really it. Um And uh you know, the, the vendors were, were pretty aggressive with the pricing much more than I expected when I went to the table too. So, um it's, it, it isn't as expensive as, as at least I expected it to be. Um So I think we're gonna, we're gonna end here. Um And I do want to um,
36:32
thank you so much, Doctor Townsend, uh Doctor Hos, Doctor Wei, um and Matt for all of your um great answers today. Um You really sharing some really practical knowledge on the things that you guys have learned um to pass along to other people. Just as we were talking about this, we need, we all need ways to learn about new things and this is another way to learn about
36:50
those things too. So, um so everyone that, that listened in today and shared some questions, we really appreciate you too. Um And I hope you have a good rest of your day. Thanks for including us. Yeah, thanks so much for being part. Appreciate it. Thank you.
37:03
Take care.
