1page.title=Low RAM Configuration 2@jd:body 3 4<!-- 5 Copyright 2013 The Android Open Source Project 6 7 Licensed under the Apache License, Version 2.0 (the "License"); 8 you may not use this file except in compliance with the License. 9 You may obtain a copy of the License at 10 11 http://www.apache.org/licenses/LICENSE-2.0 12 13 Unless required by applicable law or agreed to in writing, software 14 distributed under the License is distributed on an "AS IS" BASIS, 15 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 16 See the License for the specific language governing permissions and 17 limitations under the License. 18--> 19<div id="qv-wrapper"> 20 <div id="qv"> 21 <h2>In this document</h2> 22 <ol id="auto-toc"> 23 </ol> 24 </div> 25</div> 26 27<h2 id="intro">Introduction</h2> 28 29<p>Android now supports devices with 512MB of RAM. This documentation is intended 30to help OEMs optimize and configure Android 4.4 for low-memory devices. Several 31of these optimizations are generic enough that they can be applied to previous 32releases as well.</p> 33 34<h2 id="optimizations">Android 4.4 platform optimizations</h2> 35 36<h3 id="opt-mgmt">Improved memory management</h3> 37<ul> 38<li>Validated memory-saving kernel configurations: Kernel Same-page Merging 39(KSM), and Swap to ZRAM.</li> 40<li>Kill cached processes if about to be uncached and too large.</li> 41<li>Don't allow large services to put themselves back into A Services (so they 42can't cause the launcher to be killed).</li> 43<li>Kill processes (even ordinarily unkillable ones such as the current IME) 44that get too large in idle maintenance.</li> 45<li>Serialize the launch of background services.</li> 46<li>Tuned memory use of low-RAM devices: tighter out-of-memory (OOM) adjustment 47levels, smaller graphics caches, etc.</li> 48</ul> 49 50<h3 id="opt-mem">Reduced system memory</h3> 51<ul> 52<li>Trimmed system_server and SystemUI processes (saved several MBs).</li> 53<li>Preload dex caches in Dalvik (saved several MBs).</li> 54<li>Validated JIT-off option (saves up to 1.5MB per process).</li> 55<li>Reduced per-process font cache overhead.</li> 56<li>Introduced ArrayMap/ArraySet and used extensively in framework as a 57lighter-footprint replacement for HashMap/HashSet.</li> 58</ul> 59 60<h3 id="opt-proc">Procstats</h3> 61<p> 62Added a new Developer Option to show memory state and application memory usage 63ranked by how often they run and amount of memory consumed. 64</p> 65 66<h3 id="opt-api">API</h3> 67<p> 68Added a new ActivityManager.isLowRamDevice() to allow applications to detect 69when running on low memory devices and choose to disable large-RAM features. 70</p> 71 72<h3 id="opt-track">Memory tracking</h3> 73<p> 74New memtrack HAL to track graphics memory allocations, additional information 75in dumpsys meminfo, clarified summaries in meminfo (for example reported free 76RAM includes RAM of cached processes, so that OEMs don't try to optimize the 77wrong thing). 78</p> 79 80<h2 id="build-time">Build-time configuration</h2> 81<h3 id="flag">Enable Low Ram Device flag</h3> 82<p>We are introducing a new API called 83<code>ActivityManager.isLowRamDevice()</code> for applications to determine if 84they should turn off specific memory-intensive 85 features that work poorly on low-memory devices.</p> 86<p>For 512MB devices, this API is expected to return: "true" It can be enabled by 87 the following system property in the device makefile.<br/> 88<code>PRODUCT_PROPERTY_OVERRIDES += ro.config.low_ram=true</code></p> 89 90<h3 id="jit">Disable JIT</h3> 91 92 <p>System-wide JIT memory usage is dependent on the number of applications 93 running and the code footprint of those applications. The JIT establishes a 94 maximum translated code cache size and touches the pages within it as needed. 95 JIT costs somewhere between 3M and 6M across a typical running system.<br/> 96 <br/> 97 The large apps tend to max out the code cache fairly quickly (which by default 98 has been 1M). On average, JIT cache usage runs somewhere between 100K and 200K 99 bytes per app. Reducing the max size of the cache can help somewhat with 100 memory usage, but if set too low will send the JIT into a thrashing mode. For 101the really low-memory devices, we recommend the JIT be disabled entirely.</p> 102 103<p>This can be achieved by adding the following line to the product makefile:<br/> 104<code>PRODUCT_PROPERTY_OVERRIDES += dalvik.vm.jit.codecachesize=0</code></p> 105<h3 id="launcher">Launcher Configs</h3> 106 107 108 <p>Ensure the default wallpaper setup on launcher is <strong>not</strong> 109using live-wallpaper. Low-memory devices should not pre-install any live wallpapers. </p> 110 111 112<h2 id="kernel">Kernel configuration</h2> 113<h3 id="kernel-tuning">Tuning kernel/ActivityManager to reduce direct reclaim </h3> 114 115 116 <p>Direct reclaim happens when a process or the kernel tries to allocate a page 117 of memory (either directly or due to faulting in a new page) and the kernel 118 has used all available free memory. This requires the kernel to block the 119 allocation while it frees up a page. This in turn often requires disk I/O to 120 flush out a dirty file-backed page or waiting for <code>lowmemorykiller</code> to kill a 121 process. This can result in extra I/O in any thread, including a UI thread.</p> 122 123 <p>To avoid direct reclaim, the kernel has watermarks that trigger <code>kswapd</code> or 124 background reclaim. This is a thread that tries to free up pages so the next 125 time a real thread allocates it can succeed quickly.</p> 126 127 <p>The default threshold to trigger background reclaim is fairly low, around 2MB 128 on a 2GB device and 636KB on a 512MB device. And the kernel reclaims only a 129 few MB of memory in background reclaim. This means any process that quickly 130 allocates more than a few megabytes is going to quickly hit direct reclaim.</p> 131 132<p>Support for a new kernel tunable is added in the android-3.4 kernel branch as 133 patch 92189d47f66c67e5fd92eafaa287e153197a454f ("add extra free kbytes 134 tunable"). Cherry-picking this patch to a device's kernel will allow 135 ActivityManager to tell the kernel to try to keep 3 full-screen 32 bpp buffers 136 of memory free.</p> 137 138<p>These thresholds can be configured via the framework config.xml</p> 139 140<pre> 141<!-- Device configuration setting the /proc/sys/vm/extra_free_kbytes tunable 142in the kernel (if it exists). A high value will increase the amount of memory 143that the kernel tries to keep free, reducing allocation time and causing the 144lowmemorykiller to kill earlier. A low value allows more memory to be used by 145processes but may cause more allocations to block waiting on disk I/O or 146lowmemorykiller. Overrides the default value chosen by ActivityManager based 147on screen size. 0 prevents keeping any extra memory over what the kernel keeps 148by default. -1 keeps the default. --> 149<integer name="config_extraFreeKbytesAbsolute">-1</integer> 150</pre> 151 152<pre> 153<!-- Device configuration adjusting the /proc/sys/vm/extra_free_kbytes 154tunable in the kernel (if it exists). 0 uses the default value chosen by 155ActivityManager. A positive value will increase the amount of memory that the 156kernel tries to keep free, reducing allocation time and causing the 157lowmemorykiller to kill earlier. A negative value allows more memory to be 158used by processes but may cause more allocations to block waiting on disk I/O 159or lowmemorykiller. Directly added to the default value chosen by 160ActivityManager based on screen size. --> 161<integer name="config_extraFreeKbytesAdjust">0</integer> 162</pre> 163 164<h3 id="lowmem">Tuning LowMemoryKiller</h3> 165 166<p>ActivityManager configures the thresholds of the LowMemoryKiller to match its 167expectation of the working set of file-backed pages (cached pages) required to 168run the processes in each priority level bucket. If a device has high 169requirements for the working set, for example if the vendor UI requires more 170memory or if more services have been added, the thresholds can be increased. </p> 171 172<p>The thresholds can be reduced if too much memory is being reserved for file 173backed pages, so that background processes are being killed long before disk 174thrashing would occur due to the cache getting too small.</p> 175 176<pre> 177<!-- Device configuration setting the minfree tunable in the lowmemorykiller 178in the kernel. A high value will cause the lowmemorykiller to fire earlier, 179keeping more memory in the file cache and preventing I/O thrashing, but 180allowing fewer processes to stay in memory. A low value will keep more 181processes in memory but may cause thrashing if set too low. Overrides the 182default value chosen by ActivityManager based on screen size and total memory 183for the largest lowmemorykiller bucket, and scaled proportionally to the 184smaller buckets. -1 keeps the default. --> 185<integer name="config_lowMemoryKillerMinFreeKbytesAbsolute">-1</integer> 186</pre> 187 188<pre> 189<!-- Device configuration adjusting the minfree tunable in the 190lowmemorykiller in the kernel. A high value will cause the lowmemorykiller to 191fire earlier, keeping more memory in the file cache and preventing I/O 192thrashing, but allowing fewer processes to stay in memory. A low value will 193keep more processes in memory but may cause thrashing if set too low. Directly 194added to the default value chosen by ActivityManager based on screen 195size and total memory for the largest lowmemorykiller bucket, and scaled 196proportionally to the smaller buckets. 0 keeps the default. --> 197<integer name="config_lowMemoryKillerMinFreeKbytesAdjust">0</integer> 198</pre> 199 200<h3 id="ksm">KSM (Kernel samepage merging)</h3> 201 202<p>KSM is a kernel thread that runs in the background and compares pages in 203memory that have been marked <code>MADV_MERGEABLE</code> by user-space. If two pages are 204found to be the same, the KSM thread merges them back as a single 205copy-on-write page of memory.</p> 206 207<p>KSM will save memory over time on a running system, gaining memory duplication 208at a cost of CPU power, which could have an impact on battery life. You should 209measure whether the power tradeoff is worth the memory savings you get by 210enabling KSM.</p> 211 212<p>To test KSM, we recommend looking at long running devices (several hours) and 213seeing whether KSM makes any noticeable improvement on launch times and 214rendering times.</p> 215 216<p>To enable KSM, enable <code>CONFIG_KSM</code> in the kernel and then add the 217following lines to your` <code>init.<device>.rc</code> file:<br> 218 219<pre> 220write /sys/kernel/mm/ksm/pages_to_scan 100 221write /sys/kernel/mm/ksm/sleep_millisecs 500 222write /sys/kernel/mm/ksm/run 1 223</pre> 224 225<p>Once enabled, there are few utilities that will help in the debugging namely : 226procrank, librank, & ksminfo. These utilities allow you to see which KSM 227memory is mapped to what process, which processes use the most KSM memory. 228Once you have found a chunk of memory that looks worth exploring you can use 229either the hat utility if it's a duplicate object on the dalvik heap. </p> 230 231<h3 id="zram">Swap to zRAM</h3> 232 233<p>zRAM swap can increase the amount of memory available in the system by 234compressing memory pages and putting them in a dynamically allocated swap area 235of memory.</p> 236 237<p>Again, since this is trading off CPU time for a small increase in memory, you 238should be careful about measuring the performance impact zRAM swap has on your 239system.</p> 240 241<p>Android handles swap to zRAM at several levels:</p> 242 243<ul> 244 <li>First, the following kernel options must be enabled to use zRAM swap 245 effectively: 246 <ul> 247 <li><code>CONFIG_SWAP</code></li> 248 <li><code>CONFIG_CGROUP_MEM_RES_CTLR</code></li> 249 <li><code>CONFIG_CGROUP_MEM_RES_CTLR_SWAP</code></li> 250 <li><code>CONFIG_ZRAM</code></li> 251 </ul> 252 </li> 253 <li>Then, you should add a line that looks like this to your fstab:<br /> 254 <code>/dev/block/zram0 none swap defaults zramsize=<size in bytes>,swapprio=<swap partition priority></code><br /> 255 <code><br /> 256 zramsize</code> is mandatory and indicates how much uncompressed memory you want 257 the zram area to hold. Compression ratios in the 30-50% range are usually 258 observed.<br /> 259 <br /> 260 <code>swapprio</code> is optional and not needed if you don't have more than one swap 261 area.<br /> 262 <br /> 263 You should also be sure to label the associated block device as a swap_block_device 264 in the device-specific <a href="{@docRoot}security/selinux/implement.html"> 265 sepolicy/file_contexts</a> so that it is treated properly by SELinux. <br /> 266 <code>/dev/block/zram0 u:object_r:swap_block_device:s0</code><br /> 267 <br /> 268 </li> 269 <li>By default, the Linux kernel swaps in 8 pages of memory at a time. When 270 using ZRAM, the incremental cost of reading 1 page at a time is negligible 271 and may help in case the device is under extreme memory pressure. To read 272 only 1 page at a time, add the following to your <code>init.rc</code>:<br /> 273 <code>write /proc/sys/vm/page-cluster 0</code></li> 274 <li>In your <code>init.rc</code> after the <code>mount_all /fstab.X</code> line, add:<br /> 275 <code>swapon_all /fstab.X</code></li> 276 <li>The memory cgroups are automatically configured at boot time if the 277 feature is enabled in kernel.</li> 278 <li>If memory cgroups are available, the ActivityManager will mark lower 279 priority threads as being more swappable than other threads. If memory is 280 needed, the Android kernel will start migrating memory pages to zRAM swap, 281 giving a higher priority to those memory pages that have been marked by 282 ActivityManager. </li> 283</ul> 284 285<h3 id="carveouts">Carveouts, Ion and Contiguous Memory Allocation (CMA)</h3> 286 287<p>It is especially important on low memory devices to be mindful about 288carveouts, especially those that will not always be fully utilized -- for 289example a carveout for secure video playback. There are several solutions to 290minimizing the impact of your carveout regions that depend on the exact 291requirements of your hardware.</p> 292 293<p>If hardware permits discontiguous memory allocations, the ion system heap 294allows memory allocations from system memory, 295eliminating the need for a carveout. It also attempts to make large 296allocations to eliminate TLB pressure on peripherals. If memory regions must 297be contiguous or confined to a specific address range, the contiguous memory 298allocator (CMA) can be used.</p> 299 300<p>This creates a carveout that the system can also use of for movable pages. 301When the region is needed, movable pages will be migrated out of it, allowing 302the system to use a large carveout for other purposes when it is free. CMA can 303be used directly or more simply via ion by using the ion cma heap.</p> 304 305<h2 id="app-opts">Application optimization tips</h2> 306<ul> 307 <li>Review <a 308href="http://developer.android.com/training/articles/memory.html">Managing your 309App's Memory</a> and these past blog posts on the same topic: 310 <ul> 311 <li><a 312href="http://android-developers.blogspot.com/2009/01/avoiding-memory-leaks.html">http://android-developers.blogspot.com/2009/01/avoiding-memory-leaks.html</a></li> 313 <li><a 314href="http://android-developers.blogspot.com/2011/03/memory-analysis-for-android.html">http://android-developers.blogspot.com/2011/03/memory-analysis-for-android.html</a></li> 315 <li><a 316href="http://android-developers.blogspot.com/2009/02/track-memory-allocations.html">http://android-developers.blogspot.com/2009/02/track-memory-allocations.html</a></li> 317 <li> <a 318href="http://tools.android.com/recent/lintperformancechecks">http://tools.android.com/recent/lintperformancechecks</a></li> 319 </ul> 320</li> 321 <li>Check/remove any unused assets from preinstalled apps - 322development/tools/findunused (should help make the app smaller).</li> 323<li>Use PNG format for assets, especially when they have transparent areas</li> 324<li>If writing native code, use calloc() rather than malloc/memset</li> 325<li>Don't enable code that is writing Parcel data to disk and reading it later.</li> 326<li>Don't subscribe to every package installed, instead use ssp filtering. Add 327filtering like below: 328<br /> 329 <code><data android:scheme="package" android:ssp="com.android.pkg1" /><br /> 330 <data android:scheme="package" android:ssp="com.myapp.act1" /></code></li> 331</ul> 332 333<h3 id="process-states">Understand the various process states in Android</h3> 334 335 <ul> 336 <li><p>SERVICE - SERVICE_RESTARTING<br/> 337 Applications that are making themselves run in the background for their own 338 reason. Most common problem apps have when they run in the background too 339 much. %duration * pss is probably a good "badness" metric, although this set 340 is so focused that just doing %duration is probably better to focus on the 341 fact that we just don't want them running at all.</p></li> 342 <li><p>IMPORTANT_FOREGROUND - RECEIVER<br/> 343 Applications running in the background (not directly interacting with the 344 user) for any reason. These all add memory load to the system. In this case 345 the (%duration * pss) badness value is probably the best ordering of such 346 processes, because many of these will be always running for good reason, and 347 their pss size then is very important as part of their memory load.</p></li> 348 <li><p>PERSISTENT<br/> 349 Persistent system processes. Track pss to watch for these processes getting 350 too large.</p></li> 351 <li><p>TOP<br/> 352 Process the user is currently interacting with. Again, pss is the important 353 metric here, showing how much memory load the app is creating while in use.</p></li> 354 <li><p>HOME - CACHED_EMPTY<br/> 355 All of these processes at the bottom are ones that the system is keeping 356 around in case they are needed again; but they can be freely killed at any 357 time and re-created if needed. These are the basis for how we compute the 358 memory state -- normal, moderate, low, critical is based on how many of these 359 processes the system can keep around. Again the key thing for these processes 360 is the pss; these processes should try to get their memory footprint down as 361 much as possible when they are in this state, to allow for the maximum total 362 number of processes to be kept around. Generally a well behaved app will have 363 a pss footprint that is significantly smaller when in this state than when 364 TOP.</p></li> 365 <li> 366 <p>TOP vs. CACHED_ACTIVITY-CACHED_ACTIVITY_CLIENT<em><br/> 367 </em>The difference in pss between when a process is TOP vs. when it is in either 368 of these specific cached states is the best data for seeing how well it is 369 releasing memory when going into the background. Excluding CACHED_EMPTY state 370 makes this data better, since it removes situations when the process has 371 started for some reasons besides doing UI and so will not have to deal with 372 all of the UI overhead it gets when interacting with the user.</p></li> 373 </ul> 374 375<h2 id="analysis">Analysis</h2> 376 377<h3 id="app-startup">Analyzing app startup time</h3> 378 379<p>Use <code>$ adb shell am start</code> with the <code>-P</code> or 380<code>--start-profiler</code> option to run the profiler when your app starts. 381This will start the profiler almost immediately after your process is forked 382from zygote, before any of your code is loaded into it.</p> 383 384<h3 id="bug-reports">Analyze using bugreports </h3> 385 386<p>Now contains various information that can be used for debugging. The 387services include <code>batterystats</code>, <code>netstats</code>, 388<code>procstats</code>, and <code>usagestats</code>. You can find them with 389lines like this:</p> 390 391<pre> 392------ CHECKIN BATTERYSTATS (dumpsys batterystats --checkin) ------ 3937,0,h,-2558644,97,1946288161,3,2,0,340,4183 3947,0,h,-2553041,97,1946288161,3,2,0,340,4183 395</pre> 396 397<h3 id="persistent">Check for any persistent processes</h3> 398 399<p>Reboot the device and check the processes.<br/> 400Run for a few hours and check the processes again. There should not be any 401long running processes.</p> 402 403<h3 id="longevity">Run longevity tests</h3> 404 405<p>Run for longer durations and track the memory of the process. Does it 406increase? Does it stay constant? Create Canonical use cases and run longevity 407tests on these scenarios</p> 408