WEBVTT 00:00:00.080 --> 00:00:07.759 just yank it ah come on you bastard 00:00:05.120 --> 00:00:12.639 oh wow now it's really mad whole thing's still running 00:00:09.280 --> 00:00:14.240 meet z16 the latest member of ibm's 00:00:12.639 --> 00:00:19.520 mainframe family that's right not only do mainframes still exist 00:00:17.520 --> 00:00:24.560 powering high frequency transactional industries like global banking and 00:00:21.680 --> 00:00:28.039 travel but as of today you can get a brand spanking new mainframe configured 00:00:26.800 --> 00:00:33.360 with up to 256 cores built on the latest seven 00:00:31.119 --> 00:00:38.480 nanometer node from samsung and with up to 40 terabytes of memory 00:00:36.719 --> 00:00:44.399 very few outsiders have ever been invited to ibm's 100 000 square foot z 00:00:41.360 --> 00:00:46.879 test lab so we're gonna pack as much as 00:00:44.399 --> 00:00:51.360 we can into the very short time that we get to spend on site today just like i 00:00:49.200 --> 00:00:55.920 pack our videos with sponsors ridge wallet time to ditch that bulky wallet 00:00:53.680 --> 00:00:59.760 ridge wallet can hold up to 12 cards comes with a cash strap or money clip 00:00:57.920 --> 00:01:04.799 and is available in tons of different colors use the code Linus to save 10 and 00:01:02.719 --> 00:01:07.799 get free worldwide shipping at the link below 00:01:21.040 --> 00:01:28.720 the core pillar z is right in the name it's short for zero like 00:01:25.520 --> 00:01:31.920 zero downtime and their target is seven 00:01:28.720 --> 00:01:34.320 ninths of reliability or about three 00:01:31.920 --> 00:01:36.960 seconds of down time a year on average and to achieve that takes an 00:01:35.680 --> 00:01:41.600 architecture that's built from the ground up for resiliency 00:01:39.119 --> 00:01:45.280 now out of the hundreds of z systems that they're beating the heck out of 00:01:42.960 --> 00:01:50.720 here in the lab there's just one next-gen z16 that they got all gussied 00:01:48.479 --> 00:01:54.320 up for us with the fancy covers and everything only for us to immediately 00:01:52.640 --> 00:01:57.600 ignore them and start digging around in its guts 00:01:55.680 --> 00:02:03.040 starting at the bottom i initially assumed that these heavy black chunguses 00:02:00.479 --> 00:02:07.840 here were cooling lines but as it turns out z16 does not support being plumbed 00:02:06.000 --> 00:02:12.879 into a building-wide cooling system instead they opted for a self-contained 00:02:10.479 --> 00:02:18.239 system where every rack that has a compute drawer gets this pump and 00:02:15.599 --> 00:02:23.280 radiator unit down here at the bottom if you look closely you can actually see 00:02:20.480 --> 00:02:27.840 the two redundant cooling pumps that's in case one fails there's actually also 00:02:25.760 --> 00:02:31.840 an empty spot where it looks like one of the pumps is missing that's because in 00:02:29.840 --> 00:02:36.239 the validation phase they weren't sure if these pumps were going to meet their 00:02:33.280 --> 00:02:41.280 reliability standards so the plan was to just tuck three of them in there just in 00:02:38.640 --> 00:02:46.080 case these are actually power lines then routed through the three foot crawl 00:02:43.280 --> 00:02:52.560 space that's all underneath us covering the entire test facility and each of 00:02:48.720 --> 00:02:55.360 these two is capable of carrying 60 amps 00:02:52.560 --> 00:02:58.879 of 208 volt three-phase power on this particular config 00:02:56.879 --> 00:03:04.000 power that needs to be distributed that's where these come in each of these 00:03:01.120 --> 00:03:08.640 power distribution units or pdus down the side here are meant to be fed from a 00:03:06.560 --> 00:03:13.599 separate breaker in the event of a power loss and basically you can think of 00:03:10.959 --> 00:03:17.599 these as the enterprise grade equivalent of a power strip 00:03:15.120 --> 00:03:23.360 so using the onboard Ethernet port a technician can monitor power usage turn 00:03:20.959 --> 00:03:28.560 off particular plugs for maintenance or even update the firmware does your power 00:03:25.519 --> 00:03:29.680 strip have upgradable firmware no 00:03:28.560 --> 00:03:36.640 gross next up things get really spicy each z16 00:03:33.760 --> 00:03:41.920 system can be configured with up to four of these compute drawers and each 00:03:39.280 --> 00:03:46.560 compute drawer contains up to four of their new telum chips and these things 00:03:44.400 --> 00:03:48.799 are really cool let's go take a closer look 00:03:49.040 --> 00:03:56.959 what a monster telum is a chiplet design and you can 00:03:53.680 --> 00:03:59.840 actually see the two separate dies here 00:03:56.959 --> 00:04:05.680 it's got support for ddr4 and a total of 16 cores and 64 lanes of pci express per 00:04:03.920 --> 00:04:11.280 socket which is neat but that tells only a fraction of the 00:04:08.319 --> 00:04:16.880 story while a consumer x86 CPU might contain dedicated hardware for 00:04:13.519 --> 00:04:19.040 let's say decoding popular video codecs 00:04:16.880 --> 00:04:24.080 telum brings some very different specializations to the table 00:04:21.199 --> 00:04:30.400 each one of these cores so there's eight per die has a coprocessor to accelerate 00:04:26.960 --> 00:04:33.680 storage crypto and ibm's own compression 00:04:30.400 --> 00:04:36.160 algorithm then each die gets its own 00:04:33.680 --> 00:04:41.040 gzip compression accelerator and this is one of their biggest announcements an ai 00:04:38.639 --> 00:04:46.800 accelerator the reason for bringing that on die was that customers like financial 00:04:43.680 --> 00:04:48.880 institutions have these complex machine 00:04:46.800 --> 00:04:54.479 learning models that they built for for example fraud detection 00:04:51.120 --> 00:04:56.240 but data scientists care about accuracy 00:04:54.479 --> 00:05:00.639 not necessarily the downstream performance impact so what they found 00:04:58.639 --> 00:05:05.440 out was that they had these great models but they didn't have the performance to 00:05:02.639 --> 00:05:09.840 actually apply them to every transaction or well rather they could apply them to 00:05:07.759 --> 00:05:15.120 every transaction if they had all the time in the world but they don't 00:05:12.400 --> 00:05:19.120 so if your bank has an sla or a performance guarantee of nine 00:05:17.199 --> 00:05:23.120 milliseconds on a transaction and there's not enough time to apply their 00:05:20.960 --> 00:05:29.280 fraud detection model they either have to just let it go fraudulent or not or 00:05:26.639 --> 00:05:32.560 decline it risking the customer just pulling out the next card in their 00:05:30.639 --> 00:05:36.800 wallet effectively giving that business to their competitor by putting the ai 00:05:34.560 --> 00:05:41.600 processor on die they're giving it direct access to the transaction data 00:05:38.960 --> 00:05:46.080 that is sitting in the CPU's cache rather than forcing it to go out to 00:05:43.120 --> 00:05:50.880 memory for it and compared to last gen z ibm is figuring on a 20 to 30 x 00:05:48.880 --> 00:05:56.479 improvement in performance with and this is critical better consistency and 00:05:53.759 --> 00:06:03.280 accuracy the cash on this chip is really special too but to properly talk about 00:05:59.039 --> 00:06:03.280 that we need to zoom out a little bit 00:06:04.400 --> 00:06:12.080 now we're looking at a full compute drawer here each of which contains four 00:06:09.840 --> 00:06:15.520 telum chips actually this one's only got three in it so far i'm assuming i have 00:06:13.680 --> 00:06:20.400 this esd strap on so i can install the last one right is it this one wait so you guys handed 00:06:18.639 --> 00:06:24.160 me a working chip before oh well that was your first mistake 00:06:22.800 --> 00:06:29.120 okay uh where's the nifty install tool ah yes 00:06:26.160 --> 00:06:32.400 okay cool so i've seen this demo once which should be more than enough for me 00:06:30.880 --> 00:06:38.560 to perform it for you the telum chip goes in a little something like that 00:06:35.520 --> 00:06:40.080 then this doohickey majig lines up with 00:06:38.560 --> 00:06:43.680 the dots on the thing then you grab the chip 00:06:41.919 --> 00:06:49.919 all right you make sure it's not going to come off over something soft ish 00:06:47.840 --> 00:06:55.440 holds on to a little something like that we really hope the CPU doesn't come out 00:06:52.800 --> 00:06:57.680 if it comes out in the socket this boy's done 00:06:58.000 --> 00:07:03.759 won't right just have a bit of a reputation i don't 00:07:01.759 --> 00:07:06.560 need it right now i don't need the sass from my camera operator here 00:07:05.759 --> 00:07:12.479 then i poke this in here i think 00:07:09.919 --> 00:07:14.880 i poke in there and then i squeeze to lock it 00:07:13.440 --> 00:07:18.400 we'll unlock it and boom it's in 00:07:16.800 --> 00:07:22.479 we give it a little wiggle heck yeah 00:07:20.880 --> 00:07:27.280 now we're halfway the next thing we got to do is install 00:07:24.160 --> 00:07:29.120 our cooling and this is super cool one 00:07:27.280 --> 00:07:33.680 of the first questions i asked them when i walked up to this compute drawer was 00:07:31.919 --> 00:07:40.000 what the heck kind of thermal compound is that my initial gut feeling was that it 00:07:36.880 --> 00:07:41.520 looked like some kind of liquid metal 00:07:40.000 --> 00:07:47.680 like i mean it's not like it's not being used in commercial products the playstation 5 is using liquid metal but 00:07:45.360 --> 00:07:53.759 actually it's solid metal so rather than being like an indium gallium mix of some 00:07:50.240 --> 00:07:55.759 sort this is just an indium thermal pad 00:07:53.759 --> 00:08:00.000 and what's great is you can actually see how 00:07:56.960 --> 00:08:02.400 there you go they've got the size just 00:08:00.000 --> 00:08:07.599 exactly right to cover where the dyes are creating hot spots under the 00:08:04.400 --> 00:08:10.800 integrated heat spreader so these are 00:08:07.599 --> 00:08:12.879 going to be the interface 00:08:10.800 --> 00:08:16.800 between the integrated liquid cooling system in the chassis 00:08:15.440 --> 00:08:21.360 there we go and our processors 00:08:18.720 --> 00:08:25.919 and i guess uh i can't even tell is the latch on it look kind of looks like it 00:08:23.360 --> 00:08:29.039 where's my torque wrench or screwdriver oh is this it that's it look at that 00:08:27.840 --> 00:08:34.800 it's all configured for me and everything yes this feels like a lot 00:08:32.640 --> 00:08:40.000 i mean i guess that's a lot of pins but still 00:08:36.159 --> 00:08:42.320 oh this really feels like a lot is that 00:08:40.000 --> 00:08:47.600 okay putting that kind of pressure on that 00:08:44.880 --> 00:08:50.399 kind of spicy expensive stuff is stressful 00:08:48.640 --> 00:08:55.279 i'm sure you guys are used to it though this is adorable little 3d printed hose 00:08:53.519 --> 00:08:58.320 holders do you know how many times i've been working on a system and i wished i 00:08:56.800 --> 00:09:02.720 had one of these to hold a block out of the way while i'm doing stuff i love it 00:09:00.880 --> 00:09:05.839 i don't like this screwdriver we're going to have to make a better torx 00:09:03.920 --> 00:09:09.279 screwdriver for you guys we're working on our own screwdriver 00:09:07.519 --> 00:09:12.000 lttstore.com yeah he knew i was going to say it 00:09:10.640 --> 00:09:16.240 bloody hell that looks freaking awesome 00:09:14.320 --> 00:09:20.399 like that is hardware porn right there if i've ever seen it 00:09:18.080 --> 00:09:27.200 this four CPU config gives us a total of 64 cores in this drawer and a whopping 00:09:23.800 --> 00:09:28.800 256 pci express gen 4 lanes but you 00:09:27.200 --> 00:09:32.160 might have noticed the motherboard that they're installed on is really unusual 00:09:31.200 --> 00:09:36.880 looking where are the memory slots 00:09:34.240 --> 00:09:41.839 where are the vrms let's answer the second one first 00:09:38.399 --> 00:09:43.839 this is called a poll card or a point of 00:09:41.839 --> 00:09:48.399 load card they haven't showed me how to take it out but i'm sure i can figure it 00:09:46.320 --> 00:09:53.200 out ah yes there we go these are super cool these take the bulk 00:09:51.279 --> 00:09:58.560 12 volt power that comes from the power supplies over on the end here and step 00:09:55.920 --> 00:10:04.160 it down to the approximately one volt that we need for the cpus and the dram 00:10:01.680 --> 00:10:09.519 modules and the craziest thing about these is that they've got 14 power 00:10:06.480 --> 00:10:11.920 phases each and they can be dynamically 00:10:09.519 --> 00:10:17.040 assigned depending on where they are needed with two of them 00:10:14.800 --> 00:10:21.519 actually being completely redundant so these three do all of the work and these 00:10:19.920 --> 00:10:24.959 two including one of the ones that i took out 00:10:23.040 --> 00:10:30.839 are doing absolutely nothing unless something fails i love seeing that old 00:10:27.360 --> 00:10:34.519 school ibm ass logo on like cutting 00:10:30.839 --> 00:10:39.200 edge total combined output not to exceed 00:10:34.519 --> 00:10:41.600 660 amps okay i'll be sure not to 00:10:39.200 --> 00:10:44.640 now they get some additional help from ah yes here they are come on out little 00:10:43.519 --> 00:10:50.480 buddy there we go this is called a voltage 00:10:47.120 --> 00:10:53.040 regulator stick or vrs it's smaller with 00:10:50.480 --> 00:10:58.399 fewer phases but it contributes to step down for io cards and things like that 00:10:56.480 --> 00:11:02.560 and then oh one other really cool thing i want to show you guys is the 00:10:59.839 --> 00:11:05.519 oscillator card in the front of the oh i think this thing is locked there we go 00:11:04.079 --> 00:11:09.200 i'm moving it in the front of the compute drawer here 00:11:07.600 --> 00:11:12.800 we talked in more detail about what cards like this do in our time card 00:11:10.800 --> 00:11:16.800 video but essentially they take an external signal and ensure that all the 00:11:15.279 --> 00:11:20.959 machines within a data center are running at exactly the same time to 00:11:18.959 --> 00:11:25.040 avoid wasting clock cycles and operate more efficiently now let's move on to 00:11:23.040 --> 00:11:28.720 memory and this is where things get really funky because i ain't never seen 00:11:26.959 --> 00:11:33.600 a memory module it looks anything like this thing there's at least three crazy 00:11:31.519 --> 00:11:38.079 things going on here number one is that even though this is ddr4 00:11:35.760 --> 00:11:42.480 some of the power delivery is actually on the module itself like we've seen 00:11:40.560 --> 00:11:49.440 with ddr5 second up is the ic configuration we've 00:11:46.000 --> 00:11:52.880 got 10 of these dram chips per side 00:11:49.440 --> 00:11:55.760 per side for a total of 40 which doesn't 00:11:52.880 --> 00:11:59.839 really correspond with any kind of ecc error correction that i've ever seen in 00:11:58.880 --> 00:12:06.160 my life and it turns out that that's because it's not done on the module 00:12:03.839 --> 00:12:12.480 each memory controller in the system actually addresses eight of these memory 00:12:09.040 --> 00:12:16.720 modules and the ecc is handled more like 00:12:12.480 --> 00:12:19.519 raid with the parity data striped across 00:12:16.720 --> 00:12:22.079 the eight different modules it's all managed at the memory 00:12:20.720 --> 00:12:25.680 controller level finally we've got this bad boy right 00:12:24.160 --> 00:12:30.560 here what's this thing called again with the copper chipset explorer it's the 00:12:28.320 --> 00:12:35.920 explorer this is a proprietary buffer chip that basically adds a latency hit 00:12:33.920 --> 00:12:42.800 but allows the memory controller to address vast vast amounts of memory with each of 00:12:40.160 --> 00:12:49.920 these systems able to handle 10 terabytes of ddr4 memory absolutely 00:12:46.639 --> 00:12:49.920 flipping insane 00:12:51.120 --> 00:12:57.920 now let's talk about why i needed all four of these chips to explain telum's 00:12:55.519 --> 00:13:01.839 cache configuration where a normal CPU would have a tiny lightning fast level 00:13:00.399 --> 00:13:07.279 one cache sitting right next to the processing course followed by a larger 00:13:04.480 --> 00:13:10.000 slower level two and then a larger slower level three and so on and so 00:13:09.120 --> 00:13:16.480 forth telum has an already aggressive by 00:13:12.880 --> 00:13:19.360 consumer standards 256 kilobyte private 00:13:16.480 --> 00:13:26.399 level one cash per core so they've got eight of those per die 16 per socket 00:13:22.320 --> 00:13:30.399 that is then backed up by a whopping 32 00:13:26.399 --> 00:13:32.240 megabytes of level 2 cash per core to 00:13:30.399 --> 00:13:38.800 give you some idea of what this means in practical terms look at this die shot 00:13:35.440 --> 00:13:39.760 this thing has more cash than compute by 00:13:38.800 --> 00:13:44.880 area and it does away with level three cash 00:13:42.480 --> 00:13:50.079 all together out the window now ibm's engineers probably could have 00:13:47.120 --> 00:13:52.240 said well the level two cache is so big we probably won't need level three 00:13:51.360 --> 00:13:57.440 anyway fair enough but they didn't 00:13:54.720 --> 00:14:00.639 instead when a line needs to be evicted from one of the core's private level two 00:13:59.600 --> 00:14:06.160 caches it will actually look for empty space in 00:14:03.519 --> 00:14:11.839 another course level two and then mark that as virtual level three cache 00:14:09.360 --> 00:14:16.000 now for that core to then fetch it from another course cache does add latency 00:14:14.480 --> 00:14:20.399 compared to if it had been in its own level two right next to the core 00:14:18.160 --> 00:14:24.639 but that would be true of a shared level three cache anyway and this approach 00:14:22.639 --> 00:14:29.360 affords them an incredible degree of flexibility particularly for cloud 00:14:26.959 --> 00:14:34.160 customers who might have very different hardware needs from one to the next 00:14:31.920 --> 00:14:41.519 and it gets even crazier evicted lines can also be pushed off of 00:14:37.360 --> 00:14:45.440 the chip to another CPU all together in 00:14:41.519 --> 00:14:48.560 the z16 system allowing a single core to 00:14:45.440 --> 00:14:51.760 conceivably have access to up to two 00:14:48.560 --> 00:14:54.399 gigabytes of virtual level four cache 00:14:51.760 --> 00:14:59.839 from the other cpus in the drawer and up to eight gigabytes if we go beyond 00:15:00.320 --> 00:15:06.880 just the beyond i suppose i believe i died and went to 00:15:03.839 --> 00:15:09.839 hardware heaven to my right is a fully 00:15:06.880 --> 00:15:14.639 kitted out z16 and now we can both continue and actually expand on the tour 00:15:12.240 --> 00:15:18.720 that we started before with that single rack unit 00:15:15.760 --> 00:15:22.800 this one has all four possible compute drawers populated you can see there's 00:15:20.240 --> 00:15:28.720 one here and there's actually three in the second rack here and in order to 00:15:24.880 --> 00:15:32.000 connect them ibm uses these super cool 00:15:28.720 --> 00:15:34.639 smp9 active cables these things are 00:15:32.000 --> 00:15:39.920 super badass they've actually got heat sinks on them and inside the sheath 00:15:37.120 --> 00:15:45.560 you'll find 10 plus one fiber for redundancy with each fiber carrying 25 00:15:42.959 --> 00:15:52.480 gigabit per second for a total of 250 gigabit per second oh and of course 00:15:50.079 --> 00:15:58.160 all of the links are redundant two of them between each one of the computers 00:15:55.440 --> 00:16:02.399 now over these links the system allows the cpus and the separate drawers to 00:16:00.720 --> 00:16:06.639 share that level two cache i talked about before as a virtual level four 00:16:05.600 --> 00:16:11.360 which can apparently in some cases 00:16:08.720 --> 00:16:18.160 actually be faster to draw off of this CPU's cache over smp9 versus pulling off 00:16:15.279 --> 00:16:22.079 of your own dram in your own compute drawer 00:16:19.759 --> 00:16:25.920 it's insane i'm just getting permission to pull a coupling link card out but i 00:16:24.240 --> 00:16:30.639 realized we didn't even talk about the reservoir how sick is this 00:16:28.720 --> 00:16:36.480 just want to fill up your mainframe no big deal got your like giant looks like 00:16:33.279 --> 00:16:38.560 stainless steel res going on here 00:16:36.480 --> 00:16:41.440 absolutely gorgeous i want 04 00:16:42.720 --> 00:16:48.839 just data center things how's it going 00:16:51.120 --> 00:16:56.240 look at these chunky quick connects man 00:16:54.959 --> 00:17:00.800 that is sick dare me to pull the collar back it'll be it'll be fine it'll be fine you'd have 00:16:59.120 --> 00:17:04.559 to actually link them together to let the liquid flow hey the lights on look 00:17:02.720 --> 00:17:09.120 at this and a one 00:17:05.839 --> 00:17:11.760 and uh i don't know sure like that hey 00:17:09.120 --> 00:17:18.000 there we go suckas this is a coupling link card it's using multi-mode fiber 12 00:17:14.880 --> 00:17:20.400 lanes per four redundant for short range 00:17:18.000 --> 00:17:26.079 high bandwidth linking of multiple z systems within the data center and it 00:17:22.559 --> 00:17:28.160 uses pci express the idea here is that 00:17:26.079 --> 00:17:33.200 depending on the customer's resiliency strategy which is often government 00:17:29.840 --> 00:17:34.960 mandated they could have a whole web of 00:17:33.200 --> 00:17:39.440 z systems configured for high availability which could allow an entire 00:17:37.360 --> 00:17:44.000 drawer to go down with immediate failover and no data loss that's why 00:17:42.000 --> 00:17:49.760 these plug directly into the compute drawers for less bandwidth intensive i o 00:17:46.400 --> 00:17:53.039 devices we rely on these you can fit so 00:17:49.760 --> 00:17:54.799 much PCIe in this bad boy 00:17:53.039 --> 00:17:59.200 codename hemlock makes them sound a little sexier than they are but they're 00:17:56.799 --> 00:18:03.919 still pretty cool using these direct attached copper cables actually ah cool 00:18:02.240 --> 00:18:08.240 i've got one of these right here they have a 16x pci link to the compute 00:18:07.280 --> 00:18:14.559 drawer and then you can stack these puppies up 00:18:11.120 --> 00:18:17.440 with a ton of i o cards nbne storage 00:18:14.559 --> 00:18:20.000 crypto accelerators network cards i mean you can actually run Linux on these 00:18:18.960 --> 00:18:26.960 things so pretty much whatever add-in board the customer desires can go in one of these 00:18:24.480 --> 00:18:32.320 and the pci express link between the hemlock right here 00:18:29.760 --> 00:18:36.960 and the compute drawer goes a little something like that 00:18:34.400 --> 00:18:42.960 pretty cool now you guys might have noticed that only the center two racks 00:18:39.840 --> 00:18:45.200 contain compute while the outer ones are 00:18:42.960 --> 00:18:51.360 all i o in fact this one over here on the right is just an entire tower of io 00:18:48.880 --> 00:18:56.960 that's normal and most of the cards that you're looking at in here are either 00:18:53.360 --> 00:18:58.960 ficon or sysplex cards and these work 00:18:56.960 --> 00:19:03.039 together to handle everything from connecting to 00:19:00.240 --> 00:19:07.600 storage racks within the data center to off-site synchronization and backup 00:19:05.679 --> 00:19:11.280 they can go unboosted up to 10 kilometers and hundred kilometers with 00:19:10.240 --> 00:19:16.720 boosters but in order to test that we're gonna 00:19:13.600 --> 00:19:16.720 have to take a little field trip 00:19:17.360 --> 00:19:25.799 server hairstyle just gotta 00:19:21.280 --> 00:19:25.799 mainframe up my view here 00:19:27.600 --> 00:19:32.480 welcome to the patch room in here you 00:19:30.640 --> 00:19:37.280 will find 50 000 Ethernet connections alongside 00:19:35.559 --> 00:19:43.200 200 000 i o connections the blue ones are 00:19:40.480 --> 00:19:46.799 om3 multi-mode fiber for carrying pci express links across the data center 00:19:45.120 --> 00:19:52.080 those are the ones we saw before and often these would end up at an ibm ds8k 00:19:50.080 --> 00:19:57.919 which is their enterprise class storage while these ones are almost all ficon 00:19:55.360 --> 00:20:01.280 and this is super cool if normal fiber channel drops a packet the origin is 00:20:00.240 --> 00:20:07.280 going to retransmit it to the destination no big deal says i no 00:20:04.960 --> 00:20:13.919 says ibm that is a big deal so ficon actually contains enough 00:20:10.000 --> 00:20:15.600 storage on each adapter in the link that 00:20:13.919 --> 00:20:20.720 only the last hop of the journey needs to be 00:20:17.919 --> 00:20:25.679 retransmitted this apparently has both performance and security implications 00:20:23.840 --> 00:20:33.120 sounds expensive then over oh crap i lost it um there it 00:20:29.280 --> 00:20:36.159 is check this out inside this box is 50 00:20:33.120 --> 00:20:38.720 kilometers of real actual fiber that 00:20:36.159 --> 00:20:41.840 they can combine with another 50 kilometer roll you can actually see it 00:20:40.799 --> 00:20:47.280 right there to test out their off-site capabilities 00:20:44.720 --> 00:20:52.159 under real-world conditions and while we're over here this is super cool 00:20:49.919 --> 00:20:57.280 each of these racks is set up kind of like a a site a 00:20:54.080 --> 00:21:00.880 site b and what these do is they can 00:20:57.280 --> 00:21:03.440 take hundreds of channels of 1310 fiber 00:21:00.880 --> 00:21:08.400 convert them to different wavelengths transmit them hundreds of kilometers 00:21:05.360 --> 00:21:10.799 away or whatever then split them all 00:21:08.400 --> 00:21:15.600 back out on the other side like with a prism into your separate channels 00:21:13.679 --> 00:21:20.080 freaking bananas right then there's the brains of the operation 00:21:17.520 --> 00:21:24.080 the support element or as i call it the my wife unit although because it's z 00:21:22.559 --> 00:21:28.240 there's two redundant ones and i don't know how much you'd like that 00:21:25.679 --> 00:21:32.000 um the support element acts as a management interface for the system 00:21:30.240 --> 00:21:37.919 making sure that everything is running smoothly by monitoring this intra-rack 00:21:35.520 --> 00:21:42.640 Ethernet network that is checking temperatures and functionality and 00:21:40.400 --> 00:21:46.960 reporting on failure events bringing us to the big question what does it 00:21:44.480 --> 00:21:53.039 actually do well the biggest draw for ibm's z customers is reliability and 00:21:50.559 --> 00:21:59.520 lightning fast accurate processing of transactional data that's what it does 00:21:55.679 --> 00:22:01.280 way better than x86 and z16 aims to 00:21:59.520 --> 00:22:05.440 improve over last gen by introducing tools that will help it do all of that 00:22:03.200 --> 00:22:09.520 and faster thanks to a machine learning boost as well as adding new tools that 00:22:07.520 --> 00:22:13.520 will help customers identify and streamline security and compliance 00:22:11.280 --> 00:22:16.960 obstacles in the organization sounds fancy 00:22:15.120 --> 00:22:20.960 so i'm sure you're wondering how much one of these costs well the truth is 00:22:19.200 --> 00:22:26.080 that the number in the title there assuming i haven't changed it by now is 00:22:23.360 --> 00:22:30.400 more of a guesstimate a bear cage actually starts in the neighborhood of 00:22:28.240 --> 00:22:37.039 250 000 but every single system is custom built 00:22:33.360 --> 00:22:39.600 to the spec of the purchaser and the ibm 00:22:37.039 --> 00:22:44.000 folks here laughed and but by the time you've got all the 00:22:41.360 --> 00:22:47.120 hardware and software to actually run it i'd say we're pretty comfortable with 00:22:45.520 --> 00:22:51.520 that million dollar number as an approximate so you won't be having any 00:22:49.280 --> 00:22:55.679 of these in your basement lab anytime soon but uh what you might have is 00:22:53.760 --> 00:22:58.960 freshbooks thanks freshbooks for sponsoring this video now i'm going to 00:22:57.600 --> 00:23:04.320 go ahead and guess that you're not an accountant which is why you're going to love this software freshbooks is built 00:23:02.320 --> 00:23:08.080 for freelancers and small business owners who don't have time to waste on 00:23:06.000 --> 00:23:11.520 invoicing accounting and payment processing in fact freshbooks users can 00:23:10.000 --> 00:23:16.159 save up to 11 hours a week by streamlining and automating pesky admin 00:23:13.360 --> 00:23:19.600 tasks like time tracking following up on invoices and expense tracking with 00:23:18.159 --> 00:23:24.720 features like their new digital bills and receipt scanner over 24 million people have used 00:23:22.720 --> 00:23:28.640 freshbooks and love it for its intuitive dashboard and reports it's easy to see 00:23:27.039 --> 00:23:32.159 at a glance exactly where your business stands and it's even easier to turn 00:23:30.720 --> 00:23:36.480 everything over to your accountant come tax season 94 of freshbooks users say 00:23:34.880 --> 00:23:40.000 it's super easy to get up and running and with award-winning support you're 00:23:38.000 --> 00:23:45.280 never alone so try freshbooks for free for 30 days no credit card required by 00:23:42.240 --> 00:23:47.120 going to freshbooks.com Linus 00:23:45.280 --> 00:23:51.280 my voice is shot so if you enjoyed this video maybe go 00:23:49.440 --> 00:23:57.799 check out our tour of sfu's super computer it was super cool i feel like 00:23:53.520 --> 00:23:57.799 this is like a scene at a gremlins