Based on kernel version 4.16.1. Page generated on 2018-04-09 11:52 EST.
1 Introduction 2 ============ 3 4 dm-cache is a device mapper target written by Joe Thornber, Heinz 5 Mauelshagen, and Mike Snitzer. 6 7 It aims to improve performance of a block device (eg, a spindle) by 8 dynamically migrating some of its data to a faster, smaller device 9 (eg, an SSD). 10 11 This device-mapper solution allows us to insert this caching at 12 different levels of the dm stack, for instance above the data device for 13 a thin-provisioning pool. Caching solutions that are integrated more 14 closely with the virtual memory system should give better performance. 15 16 The target reuses the metadata library used in the thin-provisioning 17 library. 18 19 The decision as to what data to migrate and when is left to a plug-in 20 policy module. Several of these have been written as we experiment, 21 and we hope other people will contribute others for specific io 22 scenarios (eg. a vm image server). 23 24 Glossary 25 ======== 26 27 Migration - Movement of the primary copy of a logical block from one 28 device to the other. 29 Promotion - Migration from slow device to fast device. 30 Demotion - Migration from fast device to slow device. 31 32 The origin device always contains a copy of the logical block, which 33 may be out of date or kept in sync with the copy on the cache device 34 (depending on policy). 35 36 Design 37 ====== 38 39 Sub-devices 40 ----------- 41 42 The target is constructed by passing three devices to it (along with 43 other parameters detailed later): 44 45 1. An origin device - the big, slow one. 46 47 2. A cache device - the small, fast one. 48 49 3. A small metadata device - records which blocks are in the cache, 50 which are dirty, and extra hints for use by the policy object. 51 This information could be put on the cache device, but having it 52 separate allows the volume manager to configure it differently, 53 e.g. as a mirror for extra robustness. This metadata device may only 54 be used by a single cache device. 55 56 Fixed block size 57 ---------------- 58 59 The origin is divided up into blocks of a fixed size. This block size 60 is configurable when you first create the cache. Typically we've been 61 using block sizes of 256KB - 1024KB. The block size must be between 64 62 sectors (32KB) and 2097152 sectors (1GB) and a multiple of 64 sectors (32KB). 63 64 Having a fixed block size simplifies the target a lot. But it is 65 something of a compromise. For instance, a small part of a block may be 66 getting hit a lot, yet the whole block will be promoted to the cache. 67 So large block sizes are bad because they waste cache space. And small 68 block sizes are bad because they increase the amount of metadata (both 69 in core and on disk). 70 71 Cache operating modes 72 --------------------- 73 74 The cache has three operating modes: writeback, writethrough and 75 passthrough. 76 77 If writeback, the default, is selected then a write to a block that is 78 cached will go only to the cache and the block will be marked dirty in 79 the metadata. 80 81 If writethrough is selected then a write to a cached block will not 82 complete until it has hit both the origin and cache devices. Clean 83 blocks should remain clean. 84 85 If passthrough is selected, useful when the cache contents are not known 86 to be coherent with the origin device, then all reads are served from 87 the origin device (all reads miss the cache) and all writes are 88 forwarded to the origin device; additionally, write hits cause cache 89 block invalidates. To enable passthrough mode the cache must be clean. 90 Passthrough mode allows a cache device to be activated without having to 91 worry about coherency. Coherency that exists is maintained, although 92 the cache will gradually cool as writes take place. If the coherency of 93 the cache can later be verified, or established through use of the 94 "invalidate_cblocks" message, the cache device can be transitioned to 95 writethrough or writeback mode while still warm. Otherwise, the cache 96 contents can be discarded prior to transitioning to the desired 97 operating mode. 98 99 A simple cleaner policy is provided, which will clean (write back) all 100 dirty blocks in a cache. Useful for decommissioning a cache or when 101 shrinking a cache. Shrinking the cache's fast device requires all cache 102 blocks, in the area of the cache being removed, to be clean. If the 103 area being removed from the cache still contains dirty blocks the resize 104 will fail. Care must be taken to never reduce the volume used for the 105 cache's fast device until the cache is clean. This is of particular 106 importance if writeback mode is used. Writethrough and passthrough 107 modes already maintain a clean cache. Future support to partially clean 108 the cache, above a specified threshold, will allow for keeping the cache 109 warm and in writeback mode during resize. 110 111 Migration throttling 112 -------------------- 113 114 Migrating data between the origin and cache device uses bandwidth. 115 The user can set a throttle to prevent more than a certain amount of 116 migration occurring at any one time. Currently we're not taking any 117 account of normal io traffic going to the devices. More work needs 118 doing here to avoid migrating during those peak io moments. 119 120 For the time being, a message "migration_threshold <#sectors>" 121 can be used to set the maximum number of sectors being migrated, 122 the default being 2048 sectors (1MB). 123 124 Updating on-disk metadata 125 ------------------------- 126 127 On-disk metadata is committed every time a FLUSH or FUA bio is written. 128 If no such requests are made then commits will occur every second. This 129 means the cache behaves like a physical disk that has a volatile write 130 cache. If power is lost you may lose some recent writes. The metadata 131 should always be consistent in spite of any crash. 132 133 The 'dirty' state for a cache block changes far too frequently for us 134 to keep updating it on the fly. So we treat it as a hint. In normal 135 operation it will be written when the dm device is suspended. If the 136 system crashes all cache blocks will be assumed dirty when restarted. 137 138 Per-block policy hints 139 ---------------------- 140 141 Policy plug-ins can store a chunk of data per cache block. It's up to 142 the policy how big this chunk is, but it should be kept small. Like the 143 dirty flags this data is lost if there's a crash so a safe fallback 144 value should always be possible. 145 146 Policy hints affect performance, not correctness. 147 148 Policy messaging 149 ---------------- 150 151 Policies will have different tunables, specific to each one, so we 152 need a generic way of getting and setting these. Device-mapper 153 messages are used. Refer to cache-policies.txt. 154 155 Discard bitset resolution 156 ------------------------- 157 158 We can avoid copying data during migration if we know the block has 159 been discarded. A prime example of this is when mkfs discards the 160 whole block device. We store a bitset tracking the discard state of 161 blocks. However, we allow this bitset to have a different block size 162 from the cache blocks. This is because we need to track the discard 163 state for all of the origin device (compare with the dirty bitset 164 which is just for the smaller cache device). 165 166 Target interface 167 ================ 168 169 Constructor 170 ----------- 171 172 cache <metadata dev> <cache dev> <origin dev> <block size> 173 <#feature args> [<feature arg>]* 174 <policy> <#policy args> [policy args]* 175 176 metadata dev : fast device holding the persistent metadata 177 cache dev : fast device holding cached data blocks 178 origin dev : slow device holding original data blocks 179 block size : cache unit size in sectors 180 181 #feature args : number of feature arguments passed 182 feature args : writethrough or passthrough (The default is writeback.) 183 184 policy : the replacement policy to use 185 #policy args : an even number of arguments corresponding to 186 key/value pairs passed to the policy 187 policy args : key/value pairs passed to the policy 188 E.g. 'sequential_threshold 1024' 189 See cache-policies.txt for details. 190 191 Optional feature arguments are: 192 writethrough : write through caching that prohibits cache block 193 content from being different from origin block content. 194 Without this argument, the default behaviour is to write 195 back cache block contents later for performance reasons, 196 so they may differ from the corresponding origin blocks. 197 198 passthrough : a degraded mode useful for various cache coherency 199 situations (e.g., rolling back snapshots of 200 underlying storage). Reads and writes always go to 201 the origin. If a write goes to a cached origin 202 block, then the cache block is invalidated. 203 To enable passthrough mode the cache must be clean. 204 205 metadata2 : use version 2 of the metadata. This stores the dirty bits 206 in a separate btree, which improves speed of shutting 207 down the cache. 208 209 A policy called 'default' is always registered. This is an alias for 210 the policy we currently think is giving best all round performance. 211 212 As the default policy could vary between kernels, if you are relying on 213 the characteristics of a specific policy, always request it by name. 214 215 Status 216 ------ 217 218 <metadata block size> <#used metadata blocks>/<#total metadata blocks> 219 <cache block size> <#used cache blocks>/<#total cache blocks> 220 <#read hits> <#read misses> <#write hits> <#write misses> 221 <#demotions> <#promotions> <#dirty> <#features> <features>* 222 <#core args> <core args>* <policy name> <#policy args> <policy args>* 223 <cache metadata mode> 224 225 metadata block size : Fixed block size for each metadata block in 226 sectors 227 #used metadata blocks : Number of metadata blocks used 228 #total metadata blocks : Total number of metadata blocks 229 cache block size : Configurable block size for the cache device 230 in sectors 231 #used cache blocks : Number of blocks resident in the cache 232 #total cache blocks : Total number of cache blocks 233 #read hits : Number of times a READ bio has been mapped 234 to the cache 235 #read misses : Number of times a READ bio has been mapped 236 to the origin 237 #write hits : Number of times a WRITE bio has been mapped 238 to the cache 239 #write misses : Number of times a WRITE bio has been 240 mapped to the origin 241 #demotions : Number of times a block has been removed 242 from the cache 243 #promotions : Number of times a block has been moved to 244 the cache 245 #dirty : Number of blocks in the cache that differ 246 from the origin 247 #feature args : Number of feature args to follow 248 feature args : 'writethrough' (optional) 249 #core args : Number of core arguments (must be even) 250 core args : Key/value pairs for tuning the core 251 e.g. migration_threshold 252 policy name : Name of the policy 253 #policy args : Number of policy arguments to follow (must be even) 254 policy args : Key/value pairs e.g. sequential_threshold 255 cache metadata mode : ro if read-only, rw if read-write 256 In serious cases where even a read-only mode is deemed unsafe 257 no further I/O will be permitted and the status will just 258 contain the string 'Fail'. The userspace recovery tools 259 should then be used. 260 needs_check : 'needs_check' if set, '-' if not set 261 A metadata operation has failed, resulting in the needs_check 262 flag being set in the metadata's superblock. The metadata 263 device must be deactivated and checked/repaired before the 264 cache can be made fully operational again. '-' indicates 265 needs_check is not set. 266 267 Messages 268 -------- 269 270 Policies will have different tunables, specific to each one, so we 271 need a generic way of getting and setting these. Device-mapper 272 messages are used. (A sysfs interface would also be possible.) 273 274 The message format is: 275 276 <key> <value> 277 278 E.g. 279 dmsetup message my_cache 0 sequential_threshold 1024 280 281 282 Invalidation is removing an entry from the cache without writing it 283 back. Cache blocks can be invalidated via the invalidate_cblocks 284 message, which takes an arbitrary number of cblock ranges. Each cblock 285 range's end value is "one past the end", meaning 5-10 expresses a range 286 of values from 5 to 9. Each cblock must be expressed as a decimal 287 value, in the future a variant message that takes cblock ranges 288 expressed in hexadecimal may be needed to better support efficient 289 invalidation of larger caches. The cache must be in passthrough mode 290 when invalidate_cblocks is used. 291 292 invalidate_cblocks [<cblock>|<cblock begin>-<cblock end>]* 293 294 E.g. 295 dmsetup message my_cache 0 invalidate_cblocks 2345 3456-4567 5678-6789 296 297 Examples 298 ======== 299 300 The test suite can be found here: 301 302 https://github.com/jthornber/device-mapper-test-suite 303 304 dmsetup create my_cache --table '0 41943040 cache /dev/mapper/metadata \ 305 /dev/mapper/ssd /dev/mapper/origin 512 1 writeback default 0' 306 dmsetup create my_cache --table '0 41943040 cache /dev/mapper/metadata \ 307 /dev/mapper/ssd /dev/mapper/origin 1024 1 writeback \ 308 mq 4 sequential_threshold 1024 random_threshold 8'