Goodbye PXE, Hello HTTP Boot

all right so this is a fourth session there's one more after my session so today's topic is about actually two areas that I'm going to be touching one is the the pixie situation and the other is the red fish so earlier today in the morning session marked or and did mention both of them in the data center management so I'm going to be talking in more detail in there's two areas so I'm done way I work for HP so my talk is going to cover two parts one is the standardization front and so I'm going to be wearing the Hat of a UEFI forum spokesperson you know talking about the spec content I'm a vice president in the UEFI Forum doing the chief executive role of signing checks for a lot of activities and inside of HP I'm the fellow and VP for the enterprise side so the the first topic that we're going to be talking about is the the challenges of the firmware in the data center so like Mark was talking earlier today the focus has been on the client side for the last few years so we're just moving to look at the issues that that we're facing in the in the data center portions of a world and here are some of the things that we we are usually facing right so we're going to have the the bare-metal provisioning because we are deploying you know a lot of servers in the data center you know tens of thousands of servers in certain cases and we have you know the deployment issue firmware updates you know we talked about some of the new possibilities and I'm going to be also talking about and you know other possibilities there as well firmware configuration you know the BIOS settings you know how you deploy those settings across the data center for in all those machines automation whether you want to have a way of scripting the deployment security I think we talked about a few enhancement in treating the secure boot scalability you know because we're deploying such a large quantity of systems in the data center we need to have solutions that scale and that actually falls into this pixie discussions as I'll be talking about and we have to have an ecosystem where you know we have OSBI trees or EMS and you know BIOS vendors all work together you know the community like this to make everything work so the UEFI solution actually touches almost all these areas here are just some of the examples right so in a bare-metal provisioning you know we're going to be talking about Pixies row or the HTTP row how the network build should happen and in the deployment situations we we have you know the the boot device selections how you do the arresting store and recovery which I think Jeff mentioned earlier as well scripting language the UEFI shell we're going to talk a little bit about that you know the firmware updates the capsules updates that we talked about there's also this firmware management protocol that people can use to deploy the the firmware images to the individual components of the device and for more configuration where you where you need to address you know the BIOS settings and make sure everything is set in a way that you want across-the-board and scalability you know you you can't just boot from C Drive right so that's how you efi has improved the life in the flexibility of how you want to have different choices buuut until you have the device past and driver model and so on and so forth and security you know secure boot is is certainly one of them and we have all kinds of other things related to the crypto related to the you know the security protocols that we need to have encrypted disks that Jeff mentioned about and you know UEFI is also capable of providing the support with the TCG standards where you can go to TCG and look for the UEFI support as well you know that you can support TPM in fact I think the grub 2 has a project that that that now can support TPM right and in the ecosystem side you you want to have you know the you know we talked about SCT we talked about the love hours and you know you want to have everybody playing in this field so UEFI offers solutions to you know it offers the you know the foundation of today's data center firmware challenges but there are certain pieces that we need to look at the the issue and try to provide a solution that addresses some of the problems so on the data center management interface side we are also facing some challenges you know we certainly want to use the security practices that that everybody is is moving forward to and we need to support more than architecture right UEFI being one of the new I guess new kid in the in the in the scene as we deploy UEFI servers in the data center and scaling is is a big issue right you know you know when you deploy so many systems in the data center you need to have a way to you know to support the scare out servers in a way that's different from the traditional enterprise deployment and I think you know if you have been to some other Linux conferences before a lot of complaints about these interoperability zuv extensions especially in the in IPMI arena right I remember massive Garrett was talking about some of the problems that that he has seen so I'm going to be talking about you know some possible improvements okay so before you know I decide on this presentation I was actually debating you know for this many summits should I be talking more about the ACPI front because we don't have a CPR session today right but this session you know a number of us did at the Intel Developer Forum in August and we received a lot of interest you know the people you know in the audience was we're talking to us afterwards you know a lot of people were engaging with us this is actually a big problem you know the pixie has been there for a long time I think Marc probably can attach to the creation of such a framework in the what was called a wfm or you know infrastructure right so the pre-boot execution environment has a number of issues so the the critical pieces are the security issues so it has no support for encryption or authentication and it just has this DHCP server in there and it's very susceptible to middleman attacks and it's creating the the 1998 timeframe and where you know the scalability was not a you know big issue at the time so number of issues related to the TFTP timeouts and UDP packet losses it's not really reliable right and so so we're facing with issues like you know download time and deployment time which turns into dollars in the large-scale data center appointment so you know in this new era of you know you know big deployment and and you know cloudy kind of situation the problem is being aggregated so people are not standing still right just because we are not addressing this issue some people would so OMS and users have you know being working on workarounds and you know certainly one example of which is the chain loader of the third party boot loaders I pixie and mini OS or you know a couple of the examples and we have someone here that that has been leading the eye pixie effort so I'm going to be talking a little bit about it and Michael can help us so pixie is not keeping up with the modern data center requirements so that's the base that we're going to have to find issue find the solution for so I pick C so before the the time of UEFI 2.5 which is the latest back yeah I talked about we have this open source pixie client and boot loader that requires the chaining of change loading you know basically use you use pixie boots to load I pick C and then used I pick C to do the HTTP download so HTTP boot is a way to solve the problem that we have seen right the reliability the speed and the scalability and so so I pick C added the support for the HTTP boot and I'm so so I was talking about that before and after okay so before UEFI 2.5 when the I pick C solution was available it used to work only with the traditional BIOS right so so then we are faced with this problem you know we are you know we HP as an OEM we talked to a lot of the cost we were facing this issue where we were trying to convince the customers to you know move to the UEFI secure boot situation right and you know we advertised that you know the UEFI secure boot is a way to protect against the malware in the pre boot space and so on and so forth and you know the idea was excellent but you know for a lot of the large you know ice please they're like no I still have to use the HTTP boot and you know at that time it was still only supporting the legacy bars so we can't move on to UEFI boot right so it became the single largest issue in the UEFI adoption in data center and it's still the case today so we're trying to find a way to to address that problem and it you know I think there were initially some efforts on you know having the UEFI support for I pick C at the time there was no HTTP boot available and recently more work is being done and I touched on some more portions especially I actually met with Michael yesterday and he has done some excellent job even just yesterday so I had to make changes to this slide you know this morning actually I woke up at 5:00 and I had to do some changes so why not solve the pixi food challenges natively in the standard of in a standard way in UEFI you know we we have the the forum right we should be able to address this this problem more natively okay so let's look at what's available in UEFI 2.4 well we already have a networks networking stack right it's already there you know it's been there for quite a while over time you know we added more support initially it was ipv4 and then we added ipv6 and you know and then we had it is cozy support as well so this this network stack was you know created when UEFI was in it was in place like actually it was in place when EFI was was there and then we gradually you know added more support in capabilities of it so the foundation is already there so what's missing well so what's missing is really some of the key elements of supporting the HTTP boot and you know certainly you know also support DNS and TLS you know more security more you know wider network kind of capability so right so in 2005 with it add on top of the spec we we used to have and and now we have ways to support TLS we have ways to support HTTP and HTTPS so you know it's not that difficult right and when people are faced with this problem and when we already have a stack available it's just very natural force to add something on top to to address that and in the meantime we certainly also added the Bluetooth and Wi-Fi and you know all the rest of it so very attractive so in terms of the UEFI native HTTP boot you know we created this HTTP boot - wild protocol and it could boot from a URL and it you know the target could be a year file network boot program you know or it could be a shrink wrap ISO image you know whatever you like to see and you know you can certainly have the URL pre-configured or through DHCP discovery so it's not that different from what you know pixie used to be but it's more reliable it's more secure and it addresses the issues that I talked about you know associated with pixie you know certainly HTT all right so this is a fourth session there's one more after my session so today's topic is about actually two areas that I'm going to be touching one is the the pixie situation and the other is the red fish so earlier today in the morning session marked or and did mention both of them in the data center management so I'm going to be talking in more detail in those two areas so I'm done way I work for HP so my talk is going to cover two parts one is the standardization front and so I'm going to be wearing the Hat of a UEFI forum spokesperson you know talking about the spec content I'm a vice president in the UEFI Forum doing the chief executive role of sunny checks for a lot of activities and inside of HP I'm the fellow and VP for the enterprise side so the the first topic that we're going to be talking about is the the challenges of the firmware in the data center so like Mark was talking earlier today the focus has been on the client side for the last few years so we're just moving to look at the issues that that we're facing in the in the data center portions of a world and here are some of the things that we we are usually facing right so we're going to have the the bare metal provisioning because we are deploying you know a lot of servers in the data center you know tens of thousands of servers in certain cases and we have you know the deployment issue firmware updates you know we talked about some of the new possibilities and I'm going to be also talking about and you know other possibilities there as well firmware configuration you know the BIOS settings you know how you deploy those settings across the data center for in all those machines automation rather you want to have a way of scripting the deployment security I think we talked about a few enhancement in treating the secure boot scalability you know because we're deploying such a large quantity of systems in the data center we need to have solutions that scale and that actually falls into this pixie discussions as I'll be talking about and we have to have an ecosystem where you know we have OS vsi trees Williams and you know BIOS vendors all work together you know the community like this to make everything work so the UEFI solution actually touches almost all these areas here are just some of the examples right so in a bare metal provisioning you know we're going to be talking about Pixies row or the HTTP row how the network boot should happen and in the deployments is situations we we have you know the the boot device selections how you do the arresting store and and recovery which I think Jeff mentioned earlier as well scripting language the UEFI shell we're going to talk a little bit about that you know the firmware updates the capsules updates that we talked about there's also this firmware management protocol that people can use to deploy the the firmware images to the individual components of the device and for more configuration where you where you need to address you know the BIOS settings and make sure everything is set in a way that you want across the board and scalability you know you you can't just put from C Drive right so that's how you efi has improved the life in the flexibility of how you want to have different choices of boots and so you have the device past and driver model and so on so forth and security you know secure boot is is certainly one of them and we have all kinds of other things related to the crypto related to the you know the security protocols that we need to have encrypted disks that Jeff mentioned about and you know UEFI is also capable of providing the support with the TCG standards where you can go to TCG and look for the UEFI support as well you know that you can support TPM in fact I think the grub 2 has a project that that that now can support TPM right and in the ecosystem side you you want to have you know the you know we talked about SCT we talked about the love hours and you know you want to have everybody playing in this field so UEFI offers solutions to you know it offers the you know the foundation of today's data center firmware challenges but there are certain pieces that we need to look at the the issue and try to provide a solution that addresses some of the problems so on the data center management interface side we are also facing some challenges you know we certainly want to use the security practices that that everybody is is moving forward to and we need to support more than architecture right UEFI being one of the new I guess new kid in the in the in the seeing as we deploy UEFI servers in the data center and scaling is is a big issue right you know you know when you deploy so many systems in the data center you need to have a way to you know to support the scale-out servers in a way that's different from the traditional enterprise deployment and I think you know if you have been to some other Linux conferences before a lot of complaints about these interoperability zuv extensions especially in the in the ipmi arena right I remember Matthew Garrett was talking about some of the problems that that he has seen so we I'm going to be talking about you know some possible improvements okay so before you know I decided on this presentation I was actually debating you know for this many submit should I be talking more about the ACP iframe because I we don't have a CPR session today right but this session you know a number of us did at the Intel Developer Forum in August and we received a lot of interest you know the people you know in the audience was we're talking to us afterwards you know a lot of people were engaging with us this is actually a big problem you know the pixi has been there for a long time I think mark probably can attest to the creation of such a framework in what was called a wfm or you know infrastructure right so the pre-boot execution environment has a number of issues so the the critical pieces are the security issues so it has no support for encryption or authentication and it just has this DHCP server in there and it's very susceptible to middleman attacks and it's creating the the 1998 timeframe and where you know the scalability was not a you know big issue at the time so number of issues related to the TFTP timeouts and UDP packet losses it's not really reliable right and so so we're facing with issues like you know download time and deployment time which turns into dollars in the large-scale data center appointment so you know in this new era of you know you know big deployment and and you know cloudy kind of situation the problem is being aggregated so people are not standing still right just because we are not addressing this issue some people would so OMS and users have you know being working on workarounds and you know certainly one example of which is the chain load of the third party boot loaders I pick C and many OS or you know a cup of the examples and we have someone here that that has been leading the I pick C effort so I'm going to be talking a little bit about it and Michael can help us so pixie is not keeping up with the modern data center requirement so that's the the base that we're going to have to find issue find a solution for so I pick C so before the the time of UEFI 2.5 which is the latest back yeah I talked about we have this open source pixie client and boot loader that requires the chaining of change loading you know basically use you use pixie boots to load a pixie and then use the I pick C to do the HTTP download so HTTP boot is a way to solve the problem that we have seen right the reliability the speed and the scalability and so so I pixie added the support for the HTTP boot and I'm so so I was talking about the before and after okay so before UEFI 2.5 when the I pixie solution was available it used to work only with the traditional BIOS right so so then we are faced with this problem you know we are you know we HP as an OEM we talked to a lot of the summers we were facing this issue where we were trying to convince the customers to you know move to the UEFI secure boot situation right and you know we advertised that you know the UEFI secure boot is a way to protect against the malware in the pre boot space and so on and so forth and you know the idea was excellent but you know for a lot of the large you know ice please they're like no I still have to use the HTTP boot and you know at that time it was still only supporting the legacy bars so we can't move on to UEFI boot right so it became the single largest issue in the UEFI adoption in data center and it's still the case today so we're trying to find a way to to address that problem and it you know I think there were initially some effort on you know having the UEFI support for a pixie at the time there was no HTTP boot available and recently more work is being done and I'll touched on some more portions especially I actually met with Michael yesterday and he has done some excellent job even just yesterday so I had to make changes to this slide you know this morning actually I woke up at 5:00 and I had to do some changes so why not solve the pixie boots Alan jizz natively in the standard of in a standard way in UEFI you know we we have the the forum right we should be able to address this this problem more natively okay so let's look at what's available in UEFI 2.4 well we already have a networks networking stack right it's already there you know it's been there for quite a while over time you know we added more support initially it was ipv4 and then we added ipv6 and you know and then we had it I scuzzy support as well so this this network stack was you know created when UEFI was in who was in place like actually it was in place when EFI was was there and then we gradually you know added more support in capabilities of it so the foundation is already there so what's missing well so what's missing is really some of the key elements of supporting the HTTP boot and you know certainly you know also support DNS and TLS you know more security more you know wider network kind of capability so right so in 2005 with it add on top of the spec we we used to have and and now we have ways to support TLS we have ways to support HTTP and HTTPS so you know it's not that difficult right and when people are faced with this problem and when we already have a stack available it's just very natural for us to add something on top to to address that and in the meantime we certainly also added the Bluetooth and Wi-Fi and you know all the rest of it so very attractive so in terms of the UEFI native HTTP boots you know we created this HTTP boot wild protocol and it could boot from a URL and it you know the target could be a fi network boot program you know or it could be a shrink wrap ISO image you know whatever you like to say and you know you can certainly have the URL pre-configured or through the HTTP discovery so it's not that different from what you know pixie used to be but it's more reliable it's more secure and it addresses the issues that I talked about you know associated with pixie you know certainly HTTP just is the security TCP reliability and also you can have load balancing using HTTP so you know we have done this definition in the UEFI 2.5 and we already also have some code available in Tiano code or org the HTTP site is not yet available by the HTTP is I think Jeff was talking earlier about the kind of a scheduling of things that are available okay so here's a picture of how you you know how you do the HTTP HTTP boot through the DHCP discovery so you know I'm sure a few you are familiar with pixie boots and this picture doesn't look that different from that right it's just that now you have a DNS in the middle that can help and then you have a HTTP server rather than a pixie server serving you the image okay so so I mentioned that you know you could have the URL to point to an ISO image and all that it turns out that almost all the OS is you know like the the Linux side or the windows side they all have this issue where as you try to install the OS it's going to try to find some other files and you know through historical reasons they either went to the floppy which disappeared these days or they went to look for a DVD or CD ROM right and so that's very painful because you know a lot of machines don't have the DVD roms anymore so so what we did is we also created this Ram disk standard so UEFI 2.5 defined a ram disk device path node so it could support virtual disk or the virtual CD whatever format you you want to have and you know so UEFI is only available in pre boot right so so you know you you still have to have a way to carry forward that information to the OS so that's how the ACP i6o you know made into the seeing as marco was mentioning earlier you know we had to do this trials back simultaneous launch in April because of this mb ram you know touching all areas of these specifications and we took advantage of that so so the end fit table is is created to describe the MVD em you know memory regions and so on and you know so because we want to also review the you know the RAM disk place so we use that and we can describe the RAM disks to the OS so one one clarification and fit is created for the MV deems but it could describe normal Ram as well so it's not that you know the you know the DVD or the RAM disk at it has to be non-volatile now you know you can say you know here's the the RAM disk but you can still indicate that it is volatile you know in the memory space and fit allows the description for both certainly if you create if your system does support a medium and you want to place the RAM disk in that region for some reason that's fine too so mfortable is capable of doing that so that allows the runtime access of the ISO boot images in memory basically solving that problem where the OS would have to resort to finding a floppy or a DVD ROM so HTTP boot is is really the answer to to that those problems that we have seen in the in the pixie case and so so what's forcing me to change the slide this morning was this so Michael Brown actually has done some work after he became aware yesterday that we we are supporting HTTP both natively in in UEFI so he scrambled to work yesterday afternoon - to have the eye pixie changed - to be the Chiang loading from HTTP boot the UEFI native HTT people to exceed our pixie and then you can use the hypoxic - to download you know what what what's needed afterwards and he actually has written an application note on how to use that UEFI HTTP people to change load into I pick C so that's the URL that that is available so now we have two options really you know the options to address the the pixie challenges it's so funny right you have you have a problem and you know initially you don't really have a very good solution to it now we have two solutions so one is the pure native UEFI HTTP boot and if you have the OS that can support that you know that that would be great but if if that's not the case then I pixie can certainly beats there will be the chain loader solution for for you but it's not going to base on the pixie any more pixie is where all the problems were right okay so so changing topic so UEFI shell is a command-line interface that allows you to do either interactive communications or do a script so it has building commands that are standard you know it's it's like you know the file manipulations you know driver management device access and so on so forth and you can also extend that to to include the OEM commands as well to provide some additional value add and it can certainly be embedded as a boot option you know in the device selection or it could be you know bootable from a storage you can you know put it onto a USB stick and and you know go to it and it's fully documented in a specification so so it serves a few good purposes right it can be you know provided to do the scripting which allows you to automate your deployment and you know certainly the commands are allowing to do fire manipulations and and debug and testing and so so I this is this another topic that I want to touch on so so here was the slide that I mentioned that we were having some issues with the data center management interfaces and and we we concluded that the you know the current management interface such as ipmi are not really satisfying the needs for the data center so here comes the red fish red fish is an architecture successor to the previous management interfaces ipmi being one of them and there are many others so this is an industry standard effort now happening in DMT F it is in this scalable platform management forum it you know so so it's it's a sub forum inside of a DMT F and it has done the 1.0 spec with the specification schema markup white paper and browser so it has everything that you need to start with the the implementation and so it uses the the restful interface over HTTP so you know everything is kind of connected to HTTP now and it's it's using the JSON format to do the the schema and everything and it is skew secure it's it uses HTTP so it's scalable multi node and and rack levels servers very readable you know human readable output so very nice if you if you can see some of the demos you know that the MTF side is providing it's actually very very nice very easy to use so the data model you know you have you know a route of service and then each resource has a type and collectively you can describe all kinds of server hardware architecture you know either standalone multi no direct level you know whatever you have so it's very flexible you can make changes to how how these definitions are done for different kinds of servers and so to provide the connection between UEFI and redfish right we we also defined in UEFI 2.0 a new rest UEFI rest protocol so it's it is a standard pre built in band access allowing you to get get access to the restful api and and it also abstracts the the BMC specific access message so you know if you're familiar with BMC there has been all kinds of different interfaces that are that are available right so but UEFI you know when you are accessing those information from the the in line aspect you don't really care what what what that BMC access mechanism is so you want to be abstracted away from it so UEFI rest protocol is provides you just that so so whether you you you know which access max messily you use for BMC or whether you even have a BMC is up to you right as long as you can have you know these kind of API is presented to the applications that use this protocol so now I'm kind of switching the gear a little bit so previously it's all about standard effort and so on and I want to just switch to talking a little bit about how the HP proliant servers have been putting all these standards in in play so you know it's not just something on the paper right we actually have servers that allow allows you to do this kind of feature set so so certainly we we have been working on the UEFI network stack extensions HTTP FTP DNS so we we currently on the shipping servers we are already supporting booting from URL to efi file or the iso image and we also support the I scuzzy software initiator on the on the restful side we supported the HP restful api this is the pre redfish version of it and it's accessible in banned from the OS or outer band through the isle of islay is HP's BMC version so we have the HTTP support there and we also have currently the HP OEM extensions to support the UEFI BIOS configurations you know that that's what what I talked about you know you want to deploy the same configurations across a lot of different servers and on the embedded UEFI shell side we do embed our shell the UEFI shell into our boot selection menus and that show is carefully done it removed some of the standard commands those are you know kind of dangerous to have you know memory accesses and and so on you know write writing through some memories and whatever and and we also have HP value-add commands for some of the bare metal deployment so so you can use the startup scripts to automatically deploy you know lots of servers in a data center and here's a diagram that you can take a look afterwards I don't think I need to go into that it basically shows you where the users are you know how you interact with the system how you support the HTTP boot how you you know how you interact with the machines through the management network through the HTTP through the restful api getting access to the machine configurations right we also support the RAM disks and and so on so forth so this is how all these technology that I mentioned earlier you know how they are putting together on our servers and in terms of the embedded UEFI shell commands so people may wonder also you mentioned something about you know HP value-add commands so what are they so so here is kind of a list of things that you know may be of interest to the deployment where we have a ram disk so you can configure a ram disk at the shell level you can do web client and FTP to you know download things and the rest client is in band client that uses the the restful api the sis config is doing the configuration client command-line interface we also have a secure boot command that can alter the you know the the setup mode and the user mode and you know maybe in the future we will be doing that if I'm a deployed mode in the you know what was the other one the audit mode yeah so so we're going to be looking into that and then and then we certainly have you know a bhoot command and system pho is kind of giving you the the inventory of all the information that you are interested in firmware update you can do that from the shell we're not quite enthusiastic about supporting the OS driven firmware update we think that's actually could potentially cause a problem you know in this in the security sense and so on because OS is you know not necessarily always going to run right and when it crashes or when it you know does something funny then we may have a problem in that situation so we're looking at some other ways of you know maybe doing it more out-of-band you know standard way and then we also have compress in and the IP ifconfig for supporting the network stack so so a lot of the extensions that are very usable to to the deployment and I'm also happy to say that you know I think the the the redfish 1.0 was was announced you know in in the summer time frame and we are already having the I love the BMC supporting the 1.0 in the in the last few days so I think it was toward the end of September we already have a release that conforms to the to the redfish 1.0 so where I said it was HP restful api now you can use the redfish api to support the rest interface and it's just a firmware update for the ILO device and it has been integrated with UEFI so you can get the UEFI bios settings the boot order secure boot all these things can be configured through the REST API so it's very easy to use the redfish interface to do the remote configuration and deployment and here are some well I have to show some code I guess to make it more real so here are some of the examples of like the UEFI BIOS settings and if you do the get interface through the restful api you can you can get all the UEFI bios settings that we support so here's a kind of a snapshot of it and here's a snapshot of how you sorry how you support the secure boot so you can do enable/disable you can reset all the secure boot variables you can clear or keys so on our servers it's certainly general purpose we're allowing everyone to you know change the keys it's not locked down you know you can do whatever you want okay so and you can also do scripting using the restful interfaces we have a tool called HP rest and you know here's a kind of a script language that you can use to to deploy the system's very easy very straightforward and this is an example of how you want to deploy or startup a Linux operating system so like like here you know the RAM disk command is what I show there's a command that you can use to pre config a memory region to act as a ram disk and and then you can you know put your download your files there and then you launch it so it's it's actually very straightforward okay so think I'm doing fine on time but I am I want to have Michael to show you show you this this work that he has done for the eye pixie to boot directly from HTTP to HTTPS what we've done in the last 24 hours is taking advantage of the fact that there's now a native UFE HDTV boot functionality in version 245 which allows you to change from your bare metal out of the cardboard box so that over HTTP into I fix T and then use a pixie from there to do whatever you were doing before with your existing deployment infrastructure that's why I'm hoping to show here but nothing is coming out on this display hey Mikey is there a light dimming device and it's changed in one way which was to enable HTTP boot we have annoyingly represents it as those two network devices these are actually the same Nick but the top one is the default among these the UFT HTTP boot flow yeah you can see but is everything use a visible change there apparently I am allowed to contribute changes to Tiano course although they do not respect changes so I was thinking of fixing that so actually displays something meaningful anyway one can do is pick this one so this is now kicking off the UFE HTTP brew flows can go by you very quickly you can just about see it picked up a URI and now we've dropped enjoy pixie and we've got all the normal things we can do so it can show you that we get on this this isn't a game but on a physical hard way we will get approximately one decline rate and compared to you Fe native HDTV which is 10 think of it and you can you can use your existing and their windows 8 deployments are here we are going to go into the cable and grab a win file off our server and this will drop into Windows 8 yeah and Windows 8 is kicking off now so basically any boot infrastructure you have which are really worked with I pick see whether that was on legacy files or on ufp using pixie and TFTP to get hold of a pixie and then from there you can down do on a UFP version 2.5 system with no TFTP so people just take it straight out the box you can also if you want to depending on to kill be consideration of classifying pixie onto the hardware so onto the next roll into the BIOS that lobs it require tools from the vendor B you have the option to have on the firmware or to changeover either way you don't have to use TFTP server which was as I said the main motivation of that work in okay finish off with some summary so we talked about three things today one is the UEFI 2.4 http boot bridges the gap of the network boot in the data center right we address the problem that pixie has and then redfish is emerging restful management API for for the modern data center management and then I also showed how we have utilized these 2 things in the HP servers and so the next steps for call to action is we want everybody to adopt the unifying permutations with HTTP build and adopt redfish implementation in servers and our management software and then you know we want to transition the data centers to use HTTP boot and redfish restful api so these are all good things for the data center side and here it is I think that's my talk and any questions I was

Channel: UEFIForum

Views: 31,926

Keywords: Preboot Execution Environment, Booting, Software (Industry), HTTP Boot, iPXE, PXE, BIOS, UEFI, Firmware, UEFI Forum, LinuxCon Europe, LinuxCon EU, UEFI Mini-Summit

Id: EjJUBxM7pNM

Length: 56min 27sec (3387 seconds)

Published: Mon Oct 19 2015