system revision 1.2
1 $NetBSD: system,v 1.2 2008/03/19 20:34:34 jmmv Exp $ 2 3 NetBSD System Roadmap 4 ===================== 5 6 This is a small roadmap document, and deals with the main system 7 aspects of the operating system. 8 9 NetBSD 5.0 will ship with the following storage elements in place: 10 11 1. Modularized scheduler 12 2. Real-time scheduling classes and priorities 13 3. Processor sets, processor affinity and processor control 14 4. Multiprocessor optimized scheduler 15 5. High-performance 1:1 threading implementation 16 6. Pushback of the global kernel lock 17 7. New kernel concurrency model 18 8. Multiprocessor optimized memory allocators 19 9. POSIX asynchronous I/O and message queues 20 10. In-kernel linker 21 11. SysV IPC tuneables 22 12. Improved observability: minidumps, lockstat and tprof 23 13. Power management framework 24 25 We are working on the following elements, and expect them to be in 5.0: 26 27 14. Full kernel preemption for real-time threads 28 15. 64-bit time values supported 29 16. POSIX shared memory 30 17. namei() tactical changes 31 32 The following projects are expected to be included in NetBSD 6.0 33 34 18. Better resource controls 35 19. Improved observability: online crashdumps, remote debugging 36 20. Processor and cache topology aware scheduler 37 38 We expect to branch for the 5.0 release in April or May 2008, which would 39 put a NetBSD 5.0 Release date in Q4 2008. An annual release cycle would 40 thus mean that 6.0 would be branched in April or May 2009. 41 42 Pictorially, this looks like: 43 44 2008 2009 45 Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug 46 |===|===|===|===|===|===|===|===|===|===|===|===|===|===|===|===|===|===|===| 47 48 | ------- 5.0 --------- | | ------- 6.0 49 branch release branch 50 51 1-8 11,15-17,18 9,10,12-14 19 20 52 53 We'll continue to update this roadmap as features and dates get firmed up. 54 55 56 Some explanations 57 ================= 58 59 1. Modularized scheduler 60 ------------------------ 61 62 Traditionally the only method of control on process scheduling was the 63 'nice' value assigned to each process. The scheduler interface has been 64 redesiged to allow for pluggable schedulers, selected at compile time. 65 At the current time, there are no plans to switch schedulers at run-time, 66 since there is little appreciable gain to be had from that, and the extra 67 performance hit to provide this functionality is thought not to be worth 68 it. 69 70 The in-kernel scheduler interface has been enhanced to provide a framework 71 for adding new schedulers, called the common scheduler framework - more 72 information can be found in the csf(9) manual page. 73 74 Responsible: ad, dsieger, rmind, yamt 75 76 2. Real-time scheduling classes and priorities 77 ---------------------------------------------- 78 79 The scheduler has been extended to allow provide multiple new priority 80 bands, including real-time. POSIX standard interfaces for controlling 81 thread priority and scheduling class have been implemented, along with 82 a command line tool to allow control by the system administrator. 83 84 3. Processor sets, processor affinity and processor control 85 ----------------------------------------------------------- 86 87 A Solaris and HP-UX compatible interface for defining and controlling 88 processor sets has been added. Processor sets allow applications and 89 the administrator complete flexibility in partitioning CPU resources 90 among applications, down to thread-level granularity. 91 92 Linux compatibile interface controlling processor affinity, similar 93 in spirit to processor sets, is provided. 94 95 A new utility to control CPU status (cpuctl) is provided. cpuctl 96 allows the administrator to enable and disable individual CPUs at 97 the software level, while the system is running. It is expected that 98 this will in time be extended to support full dynamic reconfiguration, 99 in concert with a hypervisor such as Xen. 100 101 4. Multiprocessor optimized scheduler 102 ------------------------------------- 103 104 An intelligent, pluggable scheduler named M2 that is optimized for 105 multiprocessor systems, supports POSIX real-time extensions, 106 time-sharing class, and implements thread affinity. 107 108 5. High-performance 1:1 threading implementation 109 ------------------------------------------------ 110 111 A new lightweight 1:1 threading implementation, replacing the M:N based 112 implementation found in NetBSD 4.0 and earlier. The new implementation is 113 more correct according to POSIX thread standards, and provides a massive 114 performance boost to threaded workloads in both uni- and multi-processor 115 configurations. 116 117 6. Pushback of the global kernel lock 118 ------------------------------------- 119 120 Previously, most access to the kernel was single threaded on multiprocessor 121 systems by the global kernel_lock. The kernel_lock has been pushed back to 122 to the device driver and wire-protocol layers, providing a significant 123 performance boost on heavily loaded multiprocessor systems. 124 125 7. New kernel concurrency model 126 ------------------------------- 127 128 The non-premptive spinlock and "interrupt priority level" synchronization 129 model has been replaced wholesale with a hybrid thread/interrupt model. A 130 full range of new, lightweight synchronization primitives are available to 131 the kernel programmer, including: adaptive mutexes, reader/writer locks, 132 memory barriers, atomic operations, threaded soft interrupts, generic cross 133 calls, workqueues, priority inheritance, and per-CPU storage. 134 135 8. Multiprocessor optimized memory allocators 136 --------------------------------------------- 137 138 The memory allocators in both the kernel and user space are now fully 139 optimized for multiprocessor systems and eliminate the performance 140 degradation typically associated with memory allocators in an MP setting. 141 142 9. POSIX asynchronous I/O and message queues 143 --------------------------------------------- 144 145 A full implementation of the POSIX asynchronous I/O and message 146 queue facilities is now available. 147 148 10. In-kernel linker 149 -------------------- 150 151 A in-kernel ELF object linker has been added, and a revamped kernel module 152 infrastructure developed to accompany it. It is expected that the kernel 153 will become completely modular over time, while continuing to retain the 154 ability to link to a single binary image for embedded and hobby systems. 155 156 11. SysV IPC tuneables 157 ---------------------- 158 159 Parameters for the SVR3-compatible IPC mechanisms can now be tuned 160 completely at runtime. 161 162 12. Improved observability: minidumps, lockstat and tprof 163 --------------------------------------------------------- 164 165 The x86 architecture now supports mini crash-dumps as a support aid for 166 kernel debugging. Only memory contents actively in use by the kernel at 167 the time of crash are dumped to and recovered from disk, an improvement 168 over the traditional scheme where the complete contents of memory is 169 dumped to disk. 170 171 The lockstat and tprof commands have been addded to the system. lockstat 172 provides a high-resolution description of lock activity in a running system. 173 174 tprof uses sample based profiling in conjuction with the available 175 performance counters in order to better profile system activity. 176 177 13. Power managment framework 178 ----------------------------- 179 180 A new power management framework has been introduced that improves 181 handling of device power state transitions. As power management support 182 is now integrated with the auto-configuration subsystem, the kernel can 183 ensure that a parent device is powered on before attempting to access 184 the device. 185 186 With these changes comes an updated release of the Intel ACPI 187 Component Architecture and an x86 emulator which assists in restoring 188 uninitialized display adapters. 189 190 Leveraging this work, the i386 and amd64 kernels now support suspend 191 to RAM in uni- and multi-processor configurations on ACPI-capable 192 machines. This support has been successfully tested on a wide variety of 193 laptops, including (but not limited to) recent systems from Dell, IBM/Lenovo, 194 Fujitsu, Toshiba, and Sony. 195 196 Responsible: jmcneill, joerg 197 198 14. Full kernel preemption for real-time threads 199 ------------------------------------------------ 200 201 With the revamp of the kernel concurrency model, much of the kernel is fully 202 multi-threaded and can therefore be preempted at any time. In support of 203 lower context switch and dispatch times for real-time threads, full kernel 204 preemption is being implemented. 205 206 15. 64-bit time_t support 207 ------------------------- 208 209 The Unix 32-bit time_t value will overflow in 2037 - any mortgage calculations 210 which use a time_t value are in danger of overflowing at the present time - 211 and to address this, 64-bit time_t values will be used to contain the number 212 of seconds since 1970. 213 214 Responsible: christos 215 216 16. POSIX shared memory 217 ----------------------- 218 219 Implement POSIX shared memory facilities, which can be used to create the 220 shared memory objects and add the memory locations to the address space of 221 a process. 222 223 Responsible: rmind 224 225 17. Incremental namei improvements, Phase 1 226 ------------------------------------------- 227 228 Implement the rest of the changes to namei outlined in Message-ID: 229 <20080319053709.GB3951 (a] netbsd.org>. Simplify the locking and behavior 230 of namei() calls within the kernel to resolve path names within file 231 systems. This phase simplifies the majority of calls to namei(). 232 233 Responsible: dholland 234 235 18. Better resource controls 236 ---------------------------- 237 238 A resource provisioning and control framework that extends beyond the 239 traditional Unix process limits. 240 241 19. Improved observability: online crashdumps, remote debugging 242 --------------------------------------------------------------- 243 244 XXX crashdumps while the system is running 245 XXX firewire support in libkvm 246 247 20. Processor and cache topology aware scheduler 248 ------------------------------------------------ 249 250 Implement the detection of the topology of the processors and caches. 251 Improve the scheduler to make decisions about thread migration 252 according to the topology, to get better thread affinity and less 253 cache thrashing, and thus improve overall performance in modern SMP 254 systems. 255 256 Responsible: rmind 257 258 29. Incremental namei improvements, Phase 2 259 ------------------------------------------- 260 261 Implement the rest of the changes to namei outlined in Message-ID: 262 <20080319053709.GB3951 (a] netbsd.org>. Simplify the locking and behavior 263 of namei() calls within the kernel to resolve path names within file 264 systems. 265 266 Responsible: dholland 267 268 269 270 Andrew Doran 271 Wed Mar 19 18:26:49 GMT 2008 272
Indexes created Wed Sep 24 05:09:52 GMT 2025