Infrastructure Acceleration: Introducing Pensando Systems

[Music] my name is uh prime jam i'm the ceo of prinzano systems my name is soniji dhani i'm the chief business officer at pensandra in the cloud there is a need for acceleration of the infrastructure because for virtualizations for containers as well as for bare metal and there was a second thing which we saw that we can provide that scale we can provide the low latency we can provide all the network services feature which don't exist in any device per se the the other trend that we did see in the service provider space was that while there was a usage of commodity servers and the whole notion of sd-wan nfv services running on the compute environment there was also a dire need for the acceleration of these services and the ability to bring multiple of these functions together in a distributed manner more closer to the edge and that the scale was a completely different number particularly when you look at ai or when you look at 5g types of environments scale becomes a very important aspect so solely relying on volume servers would not be the solution so this consistent trend which was more towards specialized silicon and the ability to have deep programmability and the ability to go meet the architectural shifts occurring in the enterprise in the cloud and also in the service provider space there were it was happening across the across the various infrastructure stacks so in our case we said okay data plane we want to do it as fast as possible for meeting the objectives which we told you that we need to get make sure highest performance possible lowest latency possible and jitter also on the latency very controlled jitter power wise also make sure that it is suitable to go into any server the separating out the control plane data plane and the control plane we added the arm processors okay in our design also so that control plane management plane can come through this yes and then the data path from the host to the network or from network to the host can be processed uh you know very fast with a programmability policy is the way we have defined it is the same as being done in the cloud and that api is basically is used to define these policies to turn on the order anywhere you want you can define the policy here and we become the gate for security we become the gate for all these services whether they are allowed to talk to the other other device or not and they're all driven by policies and that is done at a high level and that is what i was talking about management plan there's another aspect to this architecture where we didn't think about security as an afterthought but we embedded it as part of the design so whether it is the whole notion of a hard gap and so when you insert our technology within a server it provides a hard gap between your workloads and your operating environments and your network infrastructure so there's a lot of embedded security that was also built into the platform even before you turn on additional security services whether it's encryption decryption whether it's stateful firewalls or whether it's ipsec or any variety of these functions we also embedded a lot of security within the platform we call as the hot gap and it is needed both in the case of bare metal because you want to have the bare metal come in uh you know change the configurations of the you know of this particular device it's also needed because you want to have isolation between the vms other thing which you can do is also tls container you mentioned you know that between the containers one group of people saying okay the container to container communication since is outside i want to have it heavily protected right so means i want to use the tls between the container to container we can do tcp termination using these p4 engines we can do the capability of using the encryption acceleration which is we have built in combination of the two we can do tls for the uh security between the you know the communication between the devices or outside even the other thing which you can do is also for storage we can run the classic nvme and run over the tcp i just have the nvme driver and i need to connect to these storage in the back end okay or distribute the storage it doesn't really matter okay that is also very powerful to put the storage wherever they want to do it and then and and then share it and nothing has to run on the on the host there is a trouble which is going to happen how soon can i find it and what tools is available to me so we have put a lot of emphasis again right from the very beginning on telemetry troubleshooting and you know collecting all the data which is really needed for accountability as well as for compliance as well as for down the road machine learning and artificial intelligence the service provider is we know for a long period of time and the cloud is for last many years and the enterprise forever yes from that point of view i think there is a need in all three segments for technology like this okay if they need to grow their business if they need to provide these services for this new capabilities which is just coming on the horizon this 5g iot new applications artificial intelligence they want to collect all the data all that kind of thing and provide these services so i think architecturally this is a transition [Music] working very closely with our customers i mean that has been a core part of the dna of the company it wasn't innovation for the sake of innovation it was really applicability of the innovation to go solve real-world problems and that has been a core part of our strategy and will continue to [Music]

Channel: Next Gen Infra

Views: 1,086

Rating: 5 out of 5

Keywords: Infrastructure Acceleration, Pensando Systems, Storage, Security

Id: t7Cz-FcnGXs

Length: 6min 16sec (376 seconds)

Published: Sun Aug 30 2020