B2B Tech Talk with Ingram Micro
B2B Tech Talk with Ingram Micro

Episode · 3 weeks ago

Ethernet Switching for Enterprise Data Centers

ABOUT THIS EPISODE

When it comes to data center expansion and cloud migration, NVIDIA Spectrum Switches might be exactly what you need.

Shelby Skrhak speaks with Jeff Shao, Director of Ethernet Alliances, and Bill Webb, Director of Ethernet Switching - Americas, both with NVIDIA, about:

- The three pillars of data center expansion

- Why NVIDIA Ethernet Storage Fabric is important to modern data centers

- Joint solutions through NVIDIA’s ecosystem of partners

To join the discussion, follow us on Twitter @IngramTechSol #B2BTechTalk

Listen to this episode and more like it by subscribing to B2B Tech Talk on Spotify, Apple Podcasts, or Stitcher. Or, tune in on our website.

You're listening to B2B tech talk withingram micro the place to learn about new technology and technologicaladvances before they become mainstream. This podcast is sponsored by ingramMicro's. Imagine next. It's not about the destination, it's about goingsomeplace you never thought possible. Go to imagine next dot ingram micro dotcom. To find out more. Let's get into it. Welcome to B two B tech talk withingram Micro. I'm your host Shelby skirt Hawk and our guest today a Jeffxiao. Director of ethernet alliances and Bill Webb. Director of ethernetswitching for the Americas, both of NVIDIA. Jeff and Bill. Welcome back tothe show. Yes, um you know, great to be back. Yeah. Thanks for having me back.Excellent. Well, you know, today we are continuing the conversation aboutethernet switching for enterprise data center. So let's start with data centerexpansion and cloud migration, which you know, really does become inevitablefor most enterprises. So, Bill, how do in video spectrum switches facilitatethis expansion of the data center infrastructure? Yeah, thanks for asking.Shelby. And to answer that, you know, just as a reminder, you know, last timewe talked about the software defined and hardware accelerated data center.And this is because today's applications are now virtualized withmicroservices that are spread across the entire data center and then datacenters are built with special hardware accelerated platforms, leveraging invideo gpu s that's critical for accelerating ai and data science aswell as DPS which accelerate the movement of data in and out of servers.And really ultimately this means that the data center is now the new unit ofcompute and because of that, the network is really critical to themodern data center and this is where our invidious Spectrums which has comein um in fact, thousands of customers have chosen or solution for theenterprise cloud and our switches are...

...perfect for really any sized network,you know, anywhere from a half rack of um hyper converge servers to hundredsor even tens of thousands of servers. It all all can be built with our ourSpectrum switches as the building blocks. Um, and in that, you know, howdo we facilitate the expansion of these data centers? It's really around threemain pillars. Okay. The first one is operational efficiency. The software inour switches is cumulus Lennox and it's really great for automation. So ourcustomers can use the same Lennox based automation and monitoring that theyused today, both for their on prem deployment and also in the cloud. Wehave what we call our digital twins environment, which is a fully simulatedreplica of our customers production network. So, so our customers canactually develop their automation in this digital twin environment fullyvalidated Prior to being rolled out in the production network and then use thesame automation tools. Actually roll it out, this accelerates deployment of thenetwork and also prevents, you know, problems that occur sometimes when whenpeople fat finger configurations, that's our first pillar. The second oneis hardware efficiency. We've been building you know high performancenetworks for over 20 years. We have best in class you know hardware for forthe modern data center. And this is critical if you look at the trafficpatterns in the modern data center it's critical to keep up with the demandsand and really get the most out of your investment. And the last pillar is whatwe call, you know our visibility and telemetry. We're really focused onaccelerating mean time to resolution. This means you know when the networkengineer gets a call that there's something wrong, they can you know withour technology actually get instant you know root cause information aboutwhat's wrong. So so that means the network engineer gets their nights orthe weekends back right And I think that that's great for everyone. Yeah Ithink we had talked about this on a previous episode uh you know my husbandyou know works in I. T. And and you...

...know was part of a small kind of he wasthe swiss army knife. So he did a little bit of everything. But he wasthat one that got that call in the middle of the night when you know whenhe had to go out to the data center and reset something and you're right, youknow being able to have that peace of mind is a huge selling point andsomething that I think should definitely be emphasized there.Absolutely well so another migration and data centers is the convergence ofstorage and compute fabrics. So we've heard a lot about ethernet storagefabric but Jeff, can you tell us about, I mean I guess specifically what iseither net storage fabric and why is it important for modern data centers inthe cloud? Yeah, great. Thanks for asking. I think we talked abouttechnology advances and market trends in enterprise data center and cloudinfrastructure. Like you mentioned with the rapid expansion of data and thechanges in the computing models that consume the data. The datainfrastructures are moved to the scale out architecture and further to theconverged architecture. Just as you mentioned, where storage and computefabrics come into one. So that leads to the adoption of ethernet storage fabric.Before we talk about ethernet story fabric, let's first look at what'shappening to storage in data centers. So in today's data centers, bigmonolithic story systems are being replaced by desegregated scale outnetwork storage where storage in a cluster of servers is put to serviceapplications in a much more agile scalable and cost effective way sincedata is now distributed, there are a lot of data movements between storageservers and between storage and compute servers. So for better understanding,let's look a little closer about these data movements And first because of thesheer volume of storage and now data is stored in a tiered storage, currentfrequently accessed data is stored in primary storage and less frequentlyaccessed data in secondary storage. So...

...data is moved between primary andsecondary storage for access ike Iving and recovery. And secondly applicationsor compute jobs need to be met with data in the ideal scenario. A computejob accesses data locally on the same server for best performance. However,because of large data set and virtualized or containerizedapplications and this is not often the case. So data need to be moved to thecompute job over the network or the compute job moved to the data.Therefore, to improve efficiency, we converge computing storage in the samerack or even on the same server. So that's the convergence we, you know, wetalk about and that leads to the converged and hyper convergedinfrastructure. Still in such a converged or hyper convergedinfrastructure, we often have applications serviced by a group ofservice and resulting data movements between this service. So these datamovements between storage, service or storage and compute service in theconverged infrastructure is called East west Traffic, for which networkperformance is very critical. So at the same time, we all know that storyservers become much faster. Thanks to new storage media such as only a mereflash applications also become much more data and performance intensive,for example, large databases, big data analytics and machine learning. So allthese require a high bandwidth and low latency storage network for datatransfer in order to avoid starved computer servers idle story servers anddegraded application performance. So to summarize what we need now is a singleunified storage network that can support all data types and applicationsand provide high performance to maximize the capability of storage andcomputer service. And also this storage network needs to be highly scalable,reliable and ultimately for data...

...expansion and open security softwaredefined to support data center and cloud services. So these requirementsfit precisely the profile of ethernet networking. So even that supportsstorage of all data types and most applications run on internet. Theinternet also has high performance, it supports speeds from one gig to 400 gettoday and soon will also support 800 gig with advanced protocols such as ourDNA over converged ethernet further accelerating data transfers. Also,Internet is an open standard supported by a lot of vendors and usedubiquitously in data centers. So it really provide customer choices andcost efficiency. So this ethernet storage network in a leaf spyingarchitecture is what we call ethernet story fabric or Yes, that's insured. Soin the leaf spine architecture there are two layers of switch is a layer ofleaf switches in the racks providing silver connectivity and the upper layerof spine switches connecting those racks and to the external network. Soonly to this simple scalable architecture. So yes, it is the fastestand most efficient way to network storage. So we have talked quite a bit,just remember that sF is the data center storage network, that is simple,high performing, open and easy to scale integrated and ultimate. Just toclarify, you use the term converged infrastructure or hyper convergedinfrastructure, is there a difference between those? Yes. So typically forconverged architecture you have computing story servers in the samewreck, but they still have a separate, you have dedicated service just tocompute jobs and you have dedicated service to service storage and hyperconverge is that, you know, you have the common server and they can purposeto compute all storage. So basically you have the same survey in Iraq andbut they just gonna be repurposed for...

...computer storage depend on the demand.So, bill when we talk about these E S F, can you expand on in videos, ethernetstorage fabric solution? Absolutely. With ethernet storage fabrics, youreally need sF capable switch and not every data center switch meets therequirements of E S F. A lot of people go in and they might try to use theexisting switching they have today, but it's just not ready. You may not havethe speeds may not have the performance, it may not have the ability to automate.Right. So if you look at our our switching solutions, you know, ourswitches are really purpose built for E S F. You know, we meet all theperformance demands of modern storage. In addition, if you look at our switchportfolio, you know, some of our switches have really unique formfactors including our half rack unit switches, what's really important, youknow, a lot of people will go build out a storage cluster which are notinexpensive but at the same time they'll try to use traditionalnetworking with it, which itself is not inexpensive. Our unique form factor isactually provides switches that are the right poor count for for a new storagecluster. So this provides some cost optimization to the storage cluster butalso get all the performance and meet all the other requirements of SF. Andwhat's really nice about RSF solution is, you know, our switches and thetechnology behind it is really the same that we use for our, you know, massivescale out data centers. It's the same building blocks to build you no es ffabric for for storage scaling to an entire data center, some other aspectsof the RSF solution, our integration, the ability to automate provisioninggoing from, you know, the hyper converged storage system into thenetwork. This means that the network becomes kind of invisible to the theoperator of the storage cluster, which makes their life a lot easier. It alsomeans that, you know, we work closely with the storage partners for thisintegration and also we build reference...

...architectures with with a lot of ourstorage partners, which means our customers are getting, you know, aproven solution from us and at the same time, I talk about invisible networkingbut we also provide a high level of visibility which I was talking aboutearlier. And in fact we include some you know, troubleshooting capabilitieswhich are help very much in the storage networking in particular, Jeff justmentioned, you know, already made or converged ethernet, this is a criticaltechnology to help accelerate, you know, next gen storage like envy me overfabrics and GPU direct storage are switches have great performance for forthese environments. But you know, our troubleshooting actually has alreadymay over converged ethernet specific features. So, you know, as ourcustomers deploy Rocky or DNA over converged ethernet, we make it veryeasy to both, you know, configure it and also troubleshooting so customerscan go with us for sF both for today's storage technology and be future proof,both in performance capabilities and speeds for for you know tomorrow'sstorage technology as well. Well definitely, you know, we we seeand we have more and more customers using a video networking with data withvarious data center solutions. But Jeff, can you share with us kind of yourecosystem partners and maybe highlights some joint solutions that that you'veseen in place. Yeah, so just like Bill talked about the media provides purposebuild networking solution for ethernet storage fabric. In fact our entry andsoftware defined how accelerated network solutions are widely used inall industries. So we have built a very broad ecosystem with leading datacenter and cloud solution providers from several Williams story systems andsoftware vendors, software defined networking and security solutionproviders and solution providers, you know, focused on specific verticalssuch as financial health care or in the...

...gas meaning payments just to name a few.So of course our broad ecosystem also includes our customers in variousindustries and also our channel partners such as in Grand Micro. Sohere are some examples. We are working with all leading storage partners inblock file object story systems, leveraging the high performance andmedia sf. Our story partners have delivered record breaking performancein various data center applications such as orders of magnitude, highertransactions per second in databases, much faster parallel files processingin file systems, accelerated ingestion, E T L in data warehousing and RTOappeal reduction in disaster recovery. So we're also working with all leadinghyper converged infrastructure solution providers such as VM ware, NutanixMicrosoft zero H Ci partners to provide a simple unified network for allworkloads on these H Ci platforms from end user computing such as virtualdesktop and unified communications. Business critical databases, large dataor big data analytics and each cloud computing. So our network solutions arealso purpose built for high performance computing and Ai machine learning. Sowe have published reference architectures with Netapp pure storage,vast data and other partners. So these reference architectures enablecustomers to fast build the ai machine learning platforms With the media TJX100 systems. And lastly, video spectrum switches are also used as underlyinghardware platforms for software defined networking and security and networkmanagement solutions. The special switches provide unique advantages inultra low latency program ability, advanced network dynamics and more. Byleveraging. These unique capabilities...

...are partners and media are deliveringintegrated and cost effective solutions to our data center customers. Remembera video and our ecosystem partners continue bringing innovation to cloudscale efficiency and cloud scale efficiency to our customers and channelpartners. So all of these enable our Channel partners to better sell supportand service their customers with the market leading highly differentiatedsolutions. So for those listeners that are are looking at all these optionsand and might need first off some guidance, I mean to to understand kindof what what situation, what their infrastructure looks like and maybewhat would best fit, you know, their industry, their latency needs. How canour partners work with you bill? Great question. I think a lot of thelisteners um, you know, our channel partners likely have storage partnersas well who we work with. Right? So probably the easiest way to get goingis, you know, let's potentially look at providing intense solutions to ourcustomers together. That's probably the first thing that we can look at. Youknow, the other part of it is we have a full channel team that's ready to workwith you, you know, for full, you know, enablement. So if you get in contactwith your ingram rap, you know, they will contact, they will get you incontact with an NVIDIA channel rep in your territory and we can do a lot ofdifferent things. We can provide, you know, training both technical and salestraining, we can get you plugged into our portal and all of our partnerprograms and you know, provide full sales enablement for you. And that goes,you know, both working with storage partners and in general, just you know,standard little an ethernet switching for for the data center, thank you Billand that helps to be able to have that resource there for listeners that wantto find out more, maybe they want to dive in a little bit deeper and do somemore research Jeff, can you share how...

...our listeners can find out more aboutwhat we talked about today. Right? So first of all, you know, become a mediapartner so that you can have access to not only the latest and greatest, youknow, solutions, but also the first class professional services and supportlike Bill said, so we have a partner proto so once you become signed up tobecome a partner, you have access to a lot of resources, you know, productasset sales tools, training material and also plus a various level ofprofessional services, so that will truly make both pre sales and postsales easy for you and also, you know, like we mentioned, we have a broadequal system so once you become part of the media ecosystem and then you haveaccess to you know, a comprehensive go to market kids, you know with you know,a media and ecosystem partners are jointly developed so that you canleverage these assets and use that for your own marketing campaign andpromotional events and better get the best ally for your marketing activities.I think this is just a better together you know, with the media and ourpartners. So we look forward to have you as part of our ecosystem andworking with you for the success and lastly, you know, we are three weeksaway from a Media Gtc. So yeah, come join a Media Gtc and meet with us andwe have a track on data center infrastructure where you can learn moreabout the media solution and our solution with our partners and here wecan discuss a lot more about how we can help you and work with you on to forour joint successes. When does GTC take place? It's from November or 5th to 8th.Fantastic. Well Jeff and Bill, thank you so much for joining us back on theshow and sharing all of your insight. Thank you Yeah, thanks a lot and she'llbe for having us back and the full...

...discussion and thank you listeners fortuning in and subscribing to B two B tech talk with ingram micro if you likethis episode or have a question, please join the discussion on twitter with thehashtag B two B tech talk. Until next time I'm Shelby scare hawk. You've beenlistening to B two B Tech Talk with Anger Micro. This episode was sponsoredby ingram Micro's. Imagine Next B two B Tech Talk is a joint production withSweet Fish Media and Anger Micro, ingram Micro production handled byLaura Burton and Christine fan. To not miss an episode. Subscribe today onyour favorite podcast platform. Mhm.

In-Stream Audio Search

NEW

Search across all episodes within this podcast

Episodes (331)