Comparing Traditional, Converged and Hyperconverged Infrastructure

hello and thank you for joining us in this presentation we are going to be talking about the various physical infrastructure technologies available to you now you may be wondering why should I watch this video well here are a couple of questions to help you get to that answer one you want to better understand the differences between traditional converged and hyper-converged offerings two are you looking to evaluate these technologies in a normalized way three are you looking to deploy a more open infrastructure that is cloud ready so that you are not locked into one particular vendor if you answered yes to these questions it probably makes sense to watch this video so let's get started before we talk about the various options we should probably find a few terms different vendors use the same words to mean different things so for the purposes of this video when we use the word traditional we mean that the compute storage and network components of the solution are physically distinct and managed separately from each other when we use the word virtual we mean that one or more of the components has been abstracted by software from its physical components the most common virtual deployments are in the compute stack but storage and network virtualization have also gained traction in recent years when we use the word converged we mean that the compute network and storage are still physically discrete components but they're all managed from a single point as you will see these technologies may or may not be from a single vendor when we use the word hyper-converged we mean that the compute and storage have been collapsed into a single offering some vendors also include networking components whereas others leave that to the end-user software-defined storage is a phrase you'll usually hear when vendors discuss their hyper-converged offerings one other word you sometimes hear when discussing your infrastructure is containers although it's not technically a word I would use when talking about physical infrastructure it bares defining anyway a container takes the concept of a virtual server like shrinks it down into just the software you need to run a discrete servers instead of deploying an OS such as Windows or Linux into a virtual machine and then installing your application into the p.m. instead you package your workload into a container with just two prerequisites it needs to run this is discrete packaging and sometimes referred to as micro services it's a great architectural and discussing workload mobility between your data center and the cloud but will we not be spending too much time on it here in this video first let's take a look at traditional infrastructure in this topology your compute storage and networking environments are distinct entities sometimes the storage resides within the servers themselves but it's still considered traditional because it discs within each server are not pooled and are not managed centrally one of the benefits of this kind of environment is that you get to pick the vendor of your choice for each tier in this example we see HP servers IBM storage and Cisco Networking if this customer chooses to change the server vendor they could do so without any other disruption to the environment another benefit is that each tier is independently scalable I can add capacity to my storage array or add servers to my compute stack without having to impact the rest of the environment or take any systems offline a third benefit is that since I'm using an open set of technologies rather than a validated configuration there are few constraints as to the solutions I can deploy on this hardware some downsides to a traditional deployment is that it can take up more physical space it can be more costly and repurposing components for other workloads may be cumbersome you can mitigate these disadvantages in two primary ways form factor and virtualization form factor changes have seen data centers collapse standard rack servers into blades internal storage collapsed into larger more robust disk arrays and networks we architected for data locality these technologies help solve some of the management and capacity planning concerns virtualization is widely utilized to mitigate the amount of hardware you need to deploy it has you added benefit of increasing your operational efficiency by utilizing centralized management and allows for utilizing more of the capacity that you do have deployed in this example we could virtualize the service by adding VMware the network with Cisco's NSX offering and the storage using IBM SVC the downside to this approach the ad cost an overhead of the virtualization technology which may be one of the reasons we're starting to see a renaissance and traditional purpose-built architectures in a converged infrastructure the compute storage and networking is all controlled via a single management interface just because the peers have been brought together this does not mean that all components need to be from a single vendor a good example of this is Cisco's approach to converged a Cisco does not have its own storage practice they partner with others to deliver pre validated and centrally managed systems offerings like Vblock using the emc versus stack using IBM or FlexPod using that app storage or good examples of this in this picture we see a versus stack it contains dedicated Cisco switching and blade servers along with IBM flash storage remember the key to converges if these are all managed and maintained from a single interface where a support ticket is opened the vendors will work with each other for resolution rather than point the end-user to the storage vendor or the server vendor a downside to this type of offering is there are limits as to what components you can utilize they need to be validated to work within the solution in the last few years hyper-converged has become a huge buzzword in the industry some vendors would position hyper-converged as the natural next step in the evolution of the data center they'll say the traditional was the old way of thinking converged is more advanced and hyper-converged is more advanced still let's take a look at the architecture to see if that claim holds up the primary offerings in the space are nuke annex a software company that partners with Lenovo Dell or Super Micro to deliver the solution as an appliance simply MIDI who was purchased by an HP Enterprise and use a specialty hardware and software to achieve its goals and hyperflex which is an offering by Cisco VMware would also say that by using its software stack with your existing hardware you are creating a hyperconvergence environment all these vendors rely on software-defined storage they use inexpensive local discs within the individual servers and create a common pool over the network this disk is usually a combination of high speed solid-state use for disk pullian caching and then a much larger set of disks that are rotational multiple copies of all data is made across the nodes but technologies such as compression and deduplication help mitigate the space taken up by these extra copies this entire approach can help drive down the cost of your environment the pros for hyperconvergence environment are that it's centrally managed it's highly scalable its high-performing within its caching envelope and it can in certain circumstances save on overall cost of ownership the downsides are that you can only run the software the vendors are certified if some solutions you need to expand your storage footprint even if you just need more servers or vice versa if you over run the disk caching mechanism your performance will be unacceptable and there's a cost and overhead associated with hyper-converged software itself trying to compare solutions can be difficult especially when you are looking at the different topologies that we've been discussing in this video here are a few tips to help you make your decision networking many network vendors leave the network components out of their solution entirely make sure you have a clear understanding of the bandwidth latency and cost requirements and get an understanding from the vendor how they will help you support the environment in case of issues this capacity traditional technologies would use disk striping for data protection for instance raid 5 or raid 6 some newer solutions rely on making several copies of the data in multiple locations never compare the amount of raw disk space when evaluating solutions a 90 terabyte traditional array might net 70 terabytes of usable space a 90 terabyte hyper-converged array might only net 30 terabytes of usable space vendors mitigate this by adding compression deduplication and other technologies when you hear expected feasible discs be wary and always ask the vendor provides guarantees and what actual usable space you'll have after Dagupan compression computing memory with some solutions especially hyper-converged a piece of each server is used to run the hyper-converged software in most cases simplicity being the notable exception this software also runs the compression and deduplication we just discussed this can cause significant overhead to the individual nodes never take the face value of the number of cores and total ram when evaluating different solutions always ask how much CPU and RAM you can actually allocate to your virtual machines management studies have shown that the operational cost of any system is many times the money spent in acquisition make sure you understand how the new environment will be managed will you be able to use a single tool or the empty is multiple how all the vendors interact with the parts of the solution that don't belong to them scalability you can evaluate the initial cost of different systems against each other but always understand what happens when you grow there's a 10% increase in capacity six months from now equate to spending ten or fifteen percent of the initial price or is it more like thirty or forty percent is there a limit to the number of servers number of drives etc that you can deploy ask if you can attach other solutions such as external disk arrays to your new infrastructure always make sure there's a path of growth in any solution you deploy so what did we see we saw the differences between traditional converged and hyper-converged offerings we saw the way in which these offerings can be normalized that they could be evaluated side-by-side and we saw how making smart infrastructure choices you could deploy more open infrastructure so you're not locked into one particular vendor so I hope you enjoyed this presentation and feel free to reach out to us so we can help you like we help our other clients

Channel: Micro Strategies

Views: 102,400

Rating: 4.8082557 out of 5

Keywords: Micro Strategies, Infrastructure

Id: N46PFNZE9zM

Length: 10min 27sec (627 seconds)

Published: Mon Jan 30 2017