1 00:00:02,670 --> 00:00:04,620 If you're not sure which Azure virtual machine 2 00:00:04,620 --> 00:00:06,900 is the right fit for your app or workload, 3 00:00:06,900 --> 00:00:09,630 in the next few minutes, I'll break down your options. 4 00:00:09,630 --> 00:00:11,790 Azure, with its vast compute infrastructure 5 00:00:11,790 --> 00:00:14,280 spanning 70-plus global regions, 6 00:00:14,280 --> 00:00:15,390 gives you the flexibility 7 00:00:15,390 --> 00:00:18,450 to choose from one of the broadest selections of VM types 8 00:00:18,450 --> 00:00:21,600 among major cloud providers to run any workload. 9 00:00:21,600 --> 00:00:24,600 A portfolio of hundreds of VM options has been designed 10 00:00:24,600 --> 00:00:26,940 to scale to the performance, cost, 11 00:00:26,940 --> 00:00:30,360 and specific technical requirements your workload needs, 12 00:00:30,360 --> 00:00:31,350 while providing access 13 00:00:31,350 --> 00:00:33,990 to the latest CPU technologies from Intel, 14 00:00:33,990 --> 00:00:36,450 plus Microsoft's custom Cobalt CPUs, 15 00:00:36,450 --> 00:00:39,240 with their power efficient, Arm-based design, 16 00:00:39,240 --> 00:00:42,270 integrated systems from our close partnership with AMD, 17 00:00:42,270 --> 00:00:44,520 and, of course, NVIDIA for the very latest 18 00:00:44,520 --> 00:00:46,350 in GPU innovation. 19 00:00:46,350 --> 00:00:48,780 And you can deploy Windows Server-based applications 20 00:00:48,780 --> 00:00:51,510 or use your favorite Linux distro on Azure. 21 00:00:51,510 --> 00:00:54,420 Our Azure-tuned Linux kernels incorporate new features 22 00:00:54,420 --> 00:00:56,970 and performance improvements at a faster cadence 23 00:00:56,970 --> 00:00:59,340 compared to default or generic kernels, 24 00:00:59,340 --> 00:01:02,580 meaning there's no need to repackage your apps and services. 25 00:01:02,580 --> 00:01:04,290 And for SUSE and Red Hat Linux, 26 00:01:04,290 --> 00:01:07,170 our engineers are co-located for integrated support 27 00:01:07,170 --> 00:01:10,050 to streamline and accelerate the resolution of any issues 28 00:01:10,050 --> 00:01:12,540 you might encounter in the least amount of time. 29 00:01:12,540 --> 00:01:14,670 So let's break down your core options 30 00:01:14,670 --> 00:01:17,550 to select the right VMs for your workloads. 31 00:01:17,550 --> 00:01:19,140 Azure VM families are optimized 32 00:01:19,140 --> 00:01:20,730 to run any size of workload, 33 00:01:20,730 --> 00:01:23,100 from general purpose to memory, compute, 34 00:01:23,100 --> 00:01:24,960 and storage intensive workloads, 35 00:01:24,960 --> 00:01:28,140 in addition to high-performance computing and AI scenarios. 36 00:01:28,140 --> 00:01:30,540 Now, when you're looking at Azure VM sizes, 37 00:01:30,540 --> 00:01:31,890 there's a structured format 38 00:01:31,890 --> 00:01:35,010 that helps you understand the characteristics of each VM. 39 00:01:35,010 --> 00:01:36,210 It starts with the family, 40 00:01:36,210 --> 00:01:38,370 followed by an optional subfamily, 41 00:01:38,370 --> 00:01:39,990 the number of vCPUs, 42 00:01:39,990 --> 00:01:42,510 the processor type and additive features, 43 00:01:42,510 --> 00:01:43,830 and the version of the VM. 44 00:01:43,830 --> 00:01:47,490 Importantly, an a signifies AMD-based processes 45 00:01:47,490 --> 00:01:49,950 and p signifies that the VM is powered 46 00:01:49,950 --> 00:01:51,870 by Arm-based processes. 47 00:01:51,870 --> 00:01:54,120 If you don't see an a or a p, 48 00:01:54,120 --> 00:01:56,670 the underlying processor is Intel based. 49 00:01:56,670 --> 00:01:58,170 And then a set of lowercase letters 50 00:01:58,170 --> 00:01:59,610 represent additive features, 51 00:01:59,610 --> 00:02:01,860 for example, s for premium storage. 52 00:02:01,860 --> 00:02:06,860 Now, to learn more, you can refer to aka.ms/VMnames, 53 00:02:07,110 --> 00:02:09,840 and you can leverage free tools like Azure Migrate 54 00:02:09,840 --> 00:02:12,750 to assess the requirements of your on-prem workloads 55 00:02:12,750 --> 00:02:15,780 and right-size your infrastructure on Azure. 56 00:02:15,780 --> 00:02:18,090 Then your choice of VM will depend on the workload 57 00:02:18,090 --> 00:02:21,150 as each series has different characteristics. 58 00:02:21,150 --> 00:02:23,460 So let's start with our entry level B-series, 59 00:02:23,460 --> 00:02:25,170 or Burstable VMs, 60 00:02:25,170 --> 00:02:26,520 which are useful for workloads 61 00:02:26,520 --> 00:02:29,940 that typically run at a low to moderate CPU baseline, 62 00:02:29,940 --> 00:02:31,530 but sometimes need to burst 63 00:02:31,530 --> 00:02:34,350 to significantly higher CPU utilization 64 00:02:34,350 --> 00:02:36,030 when the demand rises. 65 00:02:36,030 --> 00:02:37,830 An example would be a web front end, 66 00:02:37,830 --> 00:02:40,710 like a check in/check out application at a hotel 67 00:02:40,710 --> 00:02:43,440 where you need to plan for sporadic compute capacity 68 00:02:43,440 --> 00:02:45,510 to handle traffic spikes. 69 00:02:45,510 --> 00:02:48,210 That said, most of your general purpose workloads, 70 00:02:48,210 --> 00:02:52,020 such as app servers or small to medium relational databases, 71 00:02:52,020 --> 00:02:55,740 are best run on the D family of Azure virtual machines. 72 00:02:55,740 --> 00:02:59,940 These VMs offer the vCPUs, memory, and temporary storage 73 00:02:59,940 --> 00:03:03,360 to meet the requirements of most production workloads. 74 00:03:03,360 --> 00:03:06,210 Or you can opt for no local temporary data disc 75 00:03:06,210 --> 00:03:07,830 to reduce your TCO. 76 00:03:07,830 --> 00:03:09,360 The latest D-series VMs 77 00:03:09,360 --> 00:03:13,350 include fast, larger local NVMe SSD storage 78 00:03:13,350 --> 00:03:14,880 and are designed for applications 79 00:03:14,880 --> 00:03:18,390 that benefit from low-latency, high-speed local storage, 80 00:03:18,390 --> 00:03:21,120 such as applications that require fast reads and writes 81 00:03:21,120 --> 00:03:23,023 to temporary storage. 82 00:03:23,023 --> 00:03:25,950 Conversely, memory-optimized VM sizes 83 00:03:25,950 --> 00:03:27,720 that are part of our E series 84 00:03:27,720 --> 00:03:30,510 offer high memory-to-CPU ratios. 85 00:03:30,510 --> 00:03:31,770 These VMs are ideal 86 00:03:31,770 --> 00:03:35,040 for large relational databases, data analytics, 87 00:03:35,040 --> 00:03:36,150 as well as other large 88 00:03:36,150 --> 00:03:38,160 in-memory business-critical workloads 89 00:03:38,160 --> 00:03:40,500 and applications like SAP NetWeaver. 90 00:03:40,500 --> 00:03:42,210 Then, depending on your requirements, 91 00:03:42,210 --> 00:03:44,430 you can select E-series VM sizes 92 00:03:44,430 --> 00:03:48,510 that include large and fast local NVMe SSD disk storage 93 00:03:48,510 --> 00:03:50,910 for applications that benefit from low latency 94 00:03:50,910 --> 00:03:52,980 and high-speed local storage, 95 00:03:52,980 --> 00:03:55,800 or, again, the no local temporary data disk option 96 00:03:55,800 --> 00:03:57,900 is available to reduce your TCO. 97 00:03:57,900 --> 00:03:58,950 Now, for workloads demanding 98 00:03:58,950 --> 00:04:00,930 the cutting edge of network performance, 99 00:04:00,930 --> 00:04:04,290 such as network virtual appliances, large-scale e-commerce, 100 00:04:04,290 --> 00:04:05,790 and media processing, 101 00:04:05,790 --> 00:04:09,480 network optimized variants, identified by the n, 102 00:04:09,480 --> 00:04:12,150 are available for both general purpose D-series 103 00:04:12,150 --> 00:04:14,670 and memory optimized E-series VMs. 104 00:04:14,670 --> 00:04:17,220 Next, for compute-intensive applications, 105 00:04:17,220 --> 00:04:21,180 F and FX-series VMs have high vCPU performance 106 00:04:21,180 --> 00:04:23,730 and are great for compute-intensive workloads, 107 00:04:23,730 --> 00:04:26,100 such as video encoding and rendering, 108 00:04:26,100 --> 00:04:27,900 electronic design automation, 109 00:04:27,900 --> 00:04:29,550 and gaming applications. 110 00:04:29,550 --> 00:04:33,600 Alternatively, if you need to run Big Data, NoSQL databases, 111 00:04:33,600 --> 00:04:36,480 or large data warehousing solutions on Azure, 112 00:04:36,480 --> 00:04:40,200 L-series VMs are optimized for storage-intensive workloads 113 00:04:40,200 --> 00:04:43,320 and are built for the speed of data access 114 00:04:43,320 --> 00:04:47,370 And they feature ultra-fast, low-latency NVMe storage 115 00:04:47,370 --> 00:04:49,373 that's physically mapped to the host. 116 00:04:49,373 --> 00:04:51,570 This local storage layer is ideal 117 00:04:51,570 --> 00:04:54,570 for handling temporary data with high throughput, 118 00:04:54,570 --> 00:04:57,090 making it perfect for storage-intensive workloads. 119 00:04:57,090 --> 00:04:58,230 These VMs give you access 120 00:04:58,230 --> 00:05:01,650 to terabytes of storage and millions of IOPS. 121 00:05:01,650 --> 00:05:03,270 Then taking things to the next level, 122 00:05:03,270 --> 00:05:06,270 the M-series VMs are designed for applications 123 00:05:06,270 --> 00:05:09,330 that process large amounts of data in memory. 124 00:05:09,330 --> 00:05:12,720 They offer the largest memory of any VM series on Azure. 125 00:05:12,720 --> 00:05:15,990 M-series VMs are ideal for extremely large databases 126 00:05:15,990 --> 00:05:18,570 or other applications like SAP HANA 127 00:05:18,570 --> 00:05:20,700 that benefit from massive memory footprints 128 00:05:20,700 --> 00:05:23,610 and extremely high vCPU counts. 129 00:05:23,610 --> 00:05:27,570 Also, because some software is charged on a per core basis, 130 00:05:27,570 --> 00:05:29,430 to reduce the cost of software licensing 131 00:05:29,430 --> 00:05:31,680 for memory and storage-intensive workloads, 132 00:05:31,680 --> 00:05:35,850 we also provide the option of constrained vCPU-capable VMs 133 00:05:35,850 --> 00:05:38,340 across a number of VM series. 134 00:05:38,340 --> 00:05:40,170 Some database workloads, for example, 135 00:05:40,170 --> 00:05:42,180 may not need as many cores. 136 00:05:42,180 --> 00:05:45,000 So, with this option, we limit the vCPU count 137 00:05:45,000 --> 00:05:47,070 all while leaving memory, storage, 138 00:05:47,070 --> 00:05:49,380 and I/O bandwidth unchanged. 139 00:05:49,380 --> 00:05:52,470 Next, switching gears for more specialized workloads, 140 00:05:52,470 --> 00:05:54,210 the H-series VMs support 141 00:05:54,210 --> 00:05:56,160 high-performance computing workloads 142 00:05:56,160 --> 00:05:58,470 where the speed of memory is critical, 143 00:05:58,470 --> 00:06:01,500 like scientific workloads such as weather forecasting 144 00:06:01,500 --> 00:06:03,690 where tracking dangerous storms requires 145 00:06:03,690 --> 00:06:05,490 the highest precision forecasts 146 00:06:05,490 --> 00:06:07,980 with precise computations at speed. 147 00:06:07,980 --> 00:06:10,740 And there are many more critical everyday workload examples 148 00:06:10,740 --> 00:06:11,940 where these VMs can apply, 149 00:06:11,940 --> 00:06:15,120 from designing safer cars or planes and buildings 150 00:06:15,120 --> 00:06:16,380 to cheaper energy 151 00:06:16,380 --> 00:06:18,690 and even to studying the fluid dynamics 152 00:06:18,690 --> 00:06:20,340 of, for example, beer. 153 00:06:20,340 --> 00:06:22,860 Adding to this, if high-performance computing power 154 00:06:22,860 --> 00:06:25,800 for graphics-intensive and parallel processing tasks 155 00:06:25,800 --> 00:06:27,120 is more your thing, 156 00:06:27,120 --> 00:06:30,450 the N-series family of VMs are GPU-enabled 157 00:06:30,450 --> 00:06:34,560 and are ideal for AI model training, inference workloads, 158 00:06:34,560 --> 00:06:37,380 along with model fine-tuning and distillation, 159 00:06:37,380 --> 00:06:38,988 as well as digital twins, 160 00:06:38,988 --> 00:06:42,617 video rendering, predictive analytics, and more. 161 00:06:42,617 --> 00:06:46,470 Combined with NVIDIA NVLink and InfiniBand technologies 162 00:06:46,470 --> 00:06:49,500 to connect multiple GPUs and CPUs, 163 00:06:49,500 --> 00:06:52,620 you can build your own multi-VM supercomputer 164 00:06:52,620 --> 00:06:55,170 for the most demanding tasks. 165 00:06:55,170 --> 00:06:58,230 Importantly, powering all the latest generation Azure VMs 166 00:06:58,230 --> 00:07:01,590 is a key infrastructure technology called Azure Boost, 167 00:07:01,590 --> 00:07:03,960 which enhances Azure virtual machine performance 168 00:07:03,960 --> 00:07:08,340 by offloading storage, networking, and host management tasks 169 00:07:08,340 --> 00:07:10,740 to purpose-built hardware and software, 170 00:07:10,740 --> 00:07:14,160 freeing up computing resources for workloads. 171 00:07:14,160 --> 00:07:17,460 At the same time, it enforces code integrity 172 00:07:17,460 --> 00:07:20,430 so that only verified code can run on the system 173 00:07:20,430 --> 00:07:22,530 as well as strict isolation of workloads 174 00:07:22,530 --> 00:07:24,390 for additional security. 175 00:07:24,390 --> 00:07:27,000 And, across a number of these VM options, 176 00:07:27,000 --> 00:07:28,320 if you need more security, 177 00:07:28,320 --> 00:07:32,610 Azure Confidential VMs and GPUs are backed by Intel TDX 178 00:07:32,610 --> 00:07:36,300 and AMD SEV-SNP technologies. 179 00:07:36,300 --> 00:07:37,770 These use hardware-based 180 00:07:37,770 --> 00:07:40,620 Trusted Execution Environments, TEEs, 181 00:07:40,620 --> 00:07:43,350 across both CPUs and GPUs 182 00:07:43,350 --> 00:07:46,110 to protect data while it's being processed. 183 00:07:46,110 --> 00:07:48,930 These ensure only verified and authorized code 184 00:07:48,930 --> 00:07:50,430 can access sensitive data, 185 00:07:50,430 --> 00:07:52,440 which is ideal for secure collaboration 186 00:07:52,440 --> 00:07:53,640 in regulated industries 187 00:07:53,640 --> 00:07:56,510 like healthcare, finance, and government. 188 00:07:56,510 --> 00:07:58,620 These confidential VMs are available 189 00:07:58,620 --> 00:08:01,140 for the D, E, and N series. 190 00:08:01,140 --> 00:08:02,640 So now you know your options. 191 00:08:02,640 --> 00:08:04,380 As you go to deploy your VMs, 192 00:08:04,380 --> 00:08:07,290 you also have options to do so at scale. 193 00:08:07,290 --> 00:08:09,360 Azure Virtual Machine Scale Sets let you create 194 00:08:09,360 --> 00:08:11,910 thousands of virtual machines in one go 195 00:08:11,910 --> 00:08:13,410 and increase or decrease 196 00:08:13,410 --> 00:08:15,720 the number of VMs needed automatically 197 00:08:15,720 --> 00:08:17,790 based on load or schedule. 198 00:08:17,790 --> 00:08:20,280 And with another option, Azure Compute Fleet, 199 00:08:20,280 --> 00:08:22,620 you can launch even larger groups of VMs 200 00:08:22,620 --> 00:08:25,770 based directly on workload or cost requirements 201 00:08:25,770 --> 00:08:28,950 with options to select many VM types in the same fleet. 202 00:08:28,950 --> 00:08:30,600 So that was a quick tour of your options 203 00:08:30,600 --> 00:08:32,130 for Azure Core Compute. 204 00:08:32,130 --> 00:08:34,650 Whether you have basic or advanced compute needs, 205 00:08:34,650 --> 00:08:37,680 we give you a huge range of VMs to choose from. 206 00:08:37,680 --> 00:08:39,420 You can find more resources on the topic 207 00:08:39,420 --> 00:08:41,700 at aka.ms/VMAzure. 208 00:08:41,700 --> 00:08:44,617 And as always, thanks for watching.