 Now let's take a deep dive into the tests themselves to do that. I'm joined by Kim Lanar from Broadcom Kim. Welcome. Hey, thanks for having me Kim tell us about yourself. What do you do at Broadcom? So I am a performance architect here at Broadcom and I've been working with them for 15 years In my 22nd year of working at storage performance So what is the overall theme of what we're looking at in these tests? Why why are we running these tests to begin to begin with? Well when we design these products, we have an idea in mind on how well they're going to perform But it's really critical that we understand that it is actually able to perform at the expectations that we have for it so using an independent lab as we have allows us to do that and Convey that information to our customers. So let's be very specific about what's actually being tested here What is the specific hardware that's being tested in this case? What are we looking for? So we're trying to focus a little bit more on the storage Component of it, but you know the storage doesn't act in a vacuum So there's a lot of other components that are critical to making sure that the storage runs well for instance the PCIe slots The processing capabilities as well as the memory. Let's talk about test number one. Tell me what we did in test number one So the first test was actually focused on transactional database performance And what we were comparing here was the Dell PowerEdge 740 XD to the new PowerEdge 750 and in this in the previous generation our Storage controller was actually attached to SATA drives and in the R 750 instantiation We updated that to NVMe drives So it clearly shows the advantages of going from a SATA environment to an NVMe environment Let's take a look at some of those results because I want to I want to get your input on exactly what they mean So I'm seeing increases in new order per minute performance in one case with eight drives We see a 7x increase with 16 drives a 14x increase We go to log disk rights and we see a 5.6 x increase with eight drives going to 16 drives We see a 13.5 x increase I'm assuming that this would be with two controllers instead of just one going to 16 drives for log disk reads a 1.6 x increase with eight drives and a 9x increase with 16 drives Rebuild times which are obviously important going from the 740 XD to the 750 with the addition of the raid card 4.45 x faster in the 750 5.25 x faster Kim, what do these numbers mean and and why are they important to people who are Using server technology. Well, I think we can both agree that those are pretty impressive improvements and performance and What we're testing here is a TPCC like benchmark and it's an industry standard benchmark It's been around for well over 20 or 30 years even and in the form that it's in right now And what that does is it actually shows us what the transactional performance capabilities are of the server So this is a lot more holistic than just testing the storage because this is actually testing the memory It's really testing that the CPU and it's testing the storage too and one of the reasons why we focused in on the database performance and the log performance is because there are a lot of different components that go together for a sequel based Server environment in order to generate good performance and those are both very critical components Especially the log rights, especially today We got to make sure that we have very low latency log rights along with really high performance And of course you mentioned the reduction in the rebuild time So, you know one of the benefits of rate is high availability So your your storage can keep going even if you have a dry failure And what this shows is that we can get our customers databases back online even faster than we could before So Kim, what does this mean in the real world? What what is what does this performance translate into in terms of things that people care about? Well, I guess the reality is that that translates into more transactions per second from our customers databases so that our enterprise Businesses can actually get more work done. It's really important to make sure that they are able to Realize the benefits of the entire system. So by exercising the database We're actually showing what the entire Dell R750 is capable of doing especially compared to the previous generations And it provides an incentive and Reason for our customers to be able to move up into the newer technologies. Makes sense. So Kim walk us through test 2 What was it all about? So test 2 was a little bit more focused just on the storage unlike test 1 where we were testing Transactions test 2 is really just focusing on the IOPS and the bandwidth capability Again, we're testing the Dell PowerEdge 740xd and comparing that to the Dell PowerEdge 750 but what is different between these two is the gen 3 versus the gen 4 Storage infrastructure. So in particular, we're using a gen 3 PLX Switch in the 740xd as opposed to the gen 4 PLX switch in both cases We're testing 24 MVME drives. What we're doing is just really trying to see how far we can push it How many IOPS can this system handle? So not only did we scale up the number of drives that we were testing simultaneously We also scaled up the number of cores that we were testing to see how well that correlated to the end user performance Well, let's take a look at those results Test 2 shows an ability to process more outgoing storage requests up to 2.1X times the raw IOPS These are random read tests and we're looking at up to 12.3 million IOPS from 24 MVME drives In the sequential read testing, we're seeing similar gains up to 2.2X gigabytes per second in concurrent throughput That's a high number of 53.2 gigabytes per second with 24 MVME devices Kim, these are impressive numbers, but what do they actually mean? So what they mean is that the protocol for gen 4 PCIe is working. It's exactly what we were expecting and in fact, it's even a little bit more than what we were expecting We were anticipating a doubling of performance and where in the In the computer server world, do you see a doubling of performance so easily? So by using these gen 4 switches along with these gen 4 MVME drives, we're seeing a fantastic amazing improvement In the ability to move data within the server So what does that mean for customers? So for our customers, we all know that data is coming at us faster than it ever was before We're storing more and more data every single day So this shows that the ability to actually double the intake capabilities of these servers Kim, let's talk about test three. First walk us through the parameters of the test So it's similar to test two But what we've done is we've introduced not only MVME, but we've mixed that up with SAS and SATA drives Why is it important to test a mix of devices? Well today's servers, you actually have a large choice of drives Once upon a time, you really just kind of had one you had spinning hard drives Maybe there were SAS, maybe there were SATA But nowadays we've added into the mix we have SATA SSDs and we have SAS SSDs And we have MVMEs of all different types And these all have different kind of storage characteristics that fit well with different kinds of applications such as VDI or You know different Backup or cold storage and things like that So it allows our customers to kind of mix and mingle the drive types that fit their particular storage needs Understood. Let's look at the test results For test three where we tested a mix of media We saw a 2.4x increase in random read performance for 4k IOPS when moving from the 740xd to the 750 For random 4k writes, we see about a 1.3x increase in performance Sequential reads, we see a 2x increase in performance And finally for sequential writes, we see a 1.34x increase in performance So Kim, once again, help us understand why these numbers matter, especially in the real world context Well, because in the real world, we actually have a lot of different applications oftentimes simultaneously running on a single server So what this shows is the ability to You know layer on performance as the applications need it and to be able to do so in a very Articulate way where you can add just enough capacity or just enough performance to achieve your targets for your particular environment Kim, is there an economic factor that's involved when people are deciding whether to run sass or sata or nvme devices? Yeah, there absolutely is the different kinds of device have different costs associated with them And generally what we talk about is the Dollar or the sense per gigabyte depending on what makes sense and so being able to design it I mean, that's one of the most tough things right now for it Administrators is to try and balance the costs with the performance and this really allows them to fine tune that So Kim when we look at all three of these tests together What's the overall Message here the overall message is that we understand that it's very difficult for our customers to try and balance Their performance requirements with their costs, especially with today's budget So what this shows is that? You can double your performance just by going from gen three to gen four in the new del r750 architecture, so it's a big benefit for our customers So is it fair to say kim that hardware matters hardware absolutely matters It is absolutely critical in the decision making that goes into Trying to design your servers for your particular environment kim. Thanks for helping us understand these test results Well, thank you so much for having me