Add OceanBase (column store and row store)

README.md

@@ -310,7 +310,6 @@ Please help us add more systems and run the benchmarks on more types of VMs:
 - [ ] LocustDB
 - [ ] Manticore Search
 - [ ] MS SQL Server with Column Store Index (without publishing)
-- [ ] OceanBase
 - [ ] Planetscale (without publishing)
 - [ ] Redshift Spectrum
 - [ ] Seafowl

oceanbase-row/README.md

@@ -0,0 +1,51 @@
+This is the `oceanbase` entry with one change: `hits` is created
+`WITH COLUMN GROUP (all columns)` instead of `(each column)`, so it stays in
+OceanBase's row store — the LSM-tree layout every table had before 4.3
+introduced the columnstore engine, and still what a plain `CREATE TABLE` gives
+you.
+
+Read `../oceanbase/README.md` first: the RPM unpacking, the bootstrap that has
+to happen inside `./install`, the resource unit sized from `GV$OB_SERVERS`, the
+`olap` parameter template, the `VARCHAR` widths and the 1.5 MB row limit, the
+direct load and the `secure_file_priv` Unix-socket rule are all identical here
+and are only described there.
+
+To run the benchmark:
+
+```
+./benchmark.sh
+```
+
+## What differs from the columnstore entry
+
+- `create.sql` says `WITH COLUMN GROUP (all columns)` — one group holding every
+  column, which is what a row is. `SHOW CREATE TABLE` is otherwise identical to
+  the columnstore entry's, down to `COMPRESSION = 'zstd_1.3.8'`,
+  `SKIP_INDEX_LEVEL = 1`, `DELTA_FORMAT = 'encoding'`,
+  `MERGE_ENGINE = DELETE_INSERT` and `COLLATE = utf8mb4_bin`; the table is
+  still sorted by `(CounterID, EventDate, UserID, EventTime, WatchID)`. Only
+  the physical grouping of the columns changes.
+
+  Spelling the group out matters: with the `olap` template's
+  `default_table_store_format = 'column'` in force, simply *omitting* the
+  clause produces a columnstore table, not a rowstore one.
+- `install` sets `default_table_store_format = 'row'` rather than `'column'`,
+  so the parameter and the schema say the same thing.
+
+Everything else — the same 43 queries, the same load path, the same instance
+and tenant sizing — is byte-for-byte the columnstore entry, so the difference
+between the two result sets is the storage format and nothing else.
+
+## Verification
+
+Validated at 1% scale on the same instance as the columnstore entry: the DDL
+produces `WITH COLUMN GROUP(all columns)` with every other table option
+identical, the load gets all rows in, and all 43 queries return a result. The
+comparison against `clickhouse-local` lands in the same place as the
+columnstore entry's — 32 of 43 identical, the rest `LIMIT` ties plus the two
+semantic differences (`AVG(UserID)` overflowing in ClickHouse, and
+`utf8mb4_bin` being a `PAD SPACE` collation) that `../oceanbase/README.md`
+describes.
+
+The 100-million-row load has only been run for the columnstore entry. No
+results yet — those need runs on the benchmark's own EC2 machines.

oceanbase-row/benchmark.sh

@@ -0,0 +1,9 @@
+#!/bin/bash
+export BENCH_DOWNLOAD_SCRIPT="download-hits-tsv"
+# The observer's startup is not instant even when nothing has to be replayed:
+# it re-reads its schema and tablet metadata, and after ./drop_caches all of
+# that comes off the disk. Measured here at 27-45 s warm and a few minutes when
+# the volume is busy, against the driver's 300 s default -- and a ./check that
+# times out aborts the whole run, so give it room.
+export BENCH_CHECK_TIMEOUT=900
+exec ../lib/benchmark-common.sh

oceanbase-row/check

@@ -0,0 +1,16 @@
+#!/bin/bash
+set -eu
+
+OB_HOME=${OB_HOME:-/opt/oceanbase}
+source "$OB_HOME/bench.env"
+
+# obclient talks to the server over TCP and needs no privileges of its own;
+# only ./start, ./stop and ./install (which writes under $OB_HOME and has to
+# reach the Unix socket) need root.
+#
+# Check the user tenant, not the sys tenant: after a restart the server accepts
+# root@sys well before the tenant's log stream has replayed and its tablets are
+# readable, and a query issued in that window fails.
+out=$("$OB_HOME/obc" -h127.0.0.1 -P"$OB_MYSQL_PORT" -uroot@"$OB_TENANT" \
+        -A -N -e 'SELECT 1' 2>&1)
+[ "$out" = "1" ]

oceanbase-row/create.sql

@@ -0,0 +1,111 @@
+CREATE TABLE hits
+(
+    WatchID BIGINT NOT NULL,
+    JavaEnable SMALLINT NOT NULL,
+    Title VARCHAR(16384) NOT NULL,
+    GoodEvent SMALLINT NOT NULL,
+    EventTime DATETIME NOT NULL,
+    EventDate DATE NOT NULL,
+    CounterID INTEGER NOT NULL,
+    ClientIP INTEGER NOT NULL,
+    RegionID INTEGER NOT NULL,
+    UserID BIGINT NOT NULL,
+    CounterClass SMALLINT NOT NULL,
+    OS SMALLINT NOT NULL,
+    UserAgent SMALLINT NOT NULL,
+    URL VARCHAR(32768) NOT NULL,
+    Referer VARCHAR(32768) NOT NULL,
+    IsRefresh SMALLINT NOT NULL,
+    RefererCategoryID SMALLINT NOT NULL,
+    RefererRegionID INTEGER NOT NULL,
+    URLCategoryID SMALLINT NOT NULL,
+    URLRegionID INTEGER NOT NULL,
+    ResolutionWidth SMALLINT NOT NULL,
+    ResolutionHeight SMALLINT NOT NULL,
+    ResolutionDepth SMALLINT NOT NULL,
+    FlashMajor SMALLINT NOT NULL,
+    FlashMinor SMALLINT NOT NULL,
+    FlashMinor2 VARCHAR(512) NOT NULL,
+    NetMajor SMALLINT NOT NULL,
+    NetMinor SMALLINT NOT NULL,
+    UserAgentMajor SMALLINT NOT NULL,
+    UserAgentMinor VARCHAR(255) NOT NULL,
+    CookieEnable SMALLINT NOT NULL,
+    JavascriptEnable SMALLINT NOT NULL,
+    IsMobile SMALLINT NOT NULL,
+    MobilePhone SMALLINT NOT NULL,
+    MobilePhoneModel VARCHAR(512) NOT NULL,
+    Params VARCHAR(8192) NOT NULL,
+    IPNetworkID INTEGER NOT NULL,
+    TraficSourceID SMALLINT NOT NULL,
+    SearchEngineID SMALLINT NOT NULL,
+    SearchPhrase VARCHAR(8192) NOT NULL,
+    AdvEngineID SMALLINT NOT NULL,
+    IsArtifical SMALLINT NOT NULL,
+    WindowClientWidth SMALLINT NOT NULL,
+    WindowClientHeight SMALLINT NOT NULL,
+    ClientTimeZone SMALLINT NOT NULL,
+    ClientEventTime DATETIME NOT NULL,
+    SilverlightVersion1 SMALLINT NOT NULL,
+    SilverlightVersion2 SMALLINT NOT NULL,
+    SilverlightVersion3 INTEGER NOT NULL,
+    SilverlightVersion4 SMALLINT NOT NULL,
+    PageCharset VARCHAR(512) NOT NULL,
+    CodeVersion INTEGER NOT NULL,
+    IsLink SMALLINT NOT NULL,
+    IsDownload SMALLINT NOT NULL,
+    IsNotBounce SMALLINT NOT NULL,
+    FUniqID BIGINT NOT NULL,
+    OriginalURL VARCHAR(32768) NOT NULL,
+    HID INTEGER NOT NULL,
+    IsOldCounter SMALLINT NOT NULL,
+    IsEvent SMALLINT NOT NULL,
+    IsParameter SMALLINT NOT NULL,
+    DontCountHits SMALLINT NOT NULL,
+    WithHash SMALLINT NOT NULL,
+    HitColor CHAR NOT NULL,
+    LocalEventTime DATETIME NOT NULL,
+    Age SMALLINT NOT NULL,
+    Sex SMALLINT NOT NULL,
+    Income SMALLINT NOT NULL,
+    Interests SMALLINT NOT NULL,
+    Robotness SMALLINT NOT NULL,
+    RemoteIP INTEGER NOT NULL,
+    WindowName INTEGER NOT NULL,
+    OpenerName INTEGER NOT NULL,
+    HistoryLength SMALLINT NOT NULL,
+    BrowserLanguage VARCHAR(512) NOT NULL,
+    BrowserCountry VARCHAR(512) NOT NULL,
+    SocialNetwork VARCHAR(512) NOT NULL,
+    SocialAction VARCHAR(512) NOT NULL,
+    HTTPError SMALLINT NOT NULL,
+    SendTiming INTEGER NOT NULL,
+    DNSTiming INTEGER NOT NULL,
+    ConnectTiming INTEGER NOT NULL,
+    ResponseStartTiming INTEGER NOT NULL,
+    ResponseEndTiming INTEGER NOT NULL,
+    FetchTiming INTEGER NOT NULL,
+    SocialSourceNetworkID SMALLINT NOT NULL,
+    SocialSourcePage VARCHAR(2048) NOT NULL,
+    ParamPrice BIGINT NOT NULL,
+    ParamOrderID VARCHAR(512) NOT NULL,
+    ParamCurrency VARCHAR(512) NOT NULL,
+    ParamCurrencyID SMALLINT NOT NULL,
+    OpenstatServiceName VARCHAR(2048) NOT NULL,
+    OpenstatCampaignID VARCHAR(2048) NOT NULL,
+    OpenstatAdID VARCHAR(2048) NOT NULL,
+    OpenstatSourceID VARCHAR(2048) NOT NULL,
+    UTMSource VARCHAR(2048) NOT NULL,
+    UTMMedium VARCHAR(2048) NOT NULL,
+    UTMCampaign VARCHAR(2048) NOT NULL,
+    UTMContent VARCHAR(2048) NOT NULL,
+    UTMTerm VARCHAR(2048) NOT NULL,
+    FromTag VARCHAR(2048) NOT NULL,
+    HasGCLID SMALLINT NOT NULL,
+    RefererHash BIGINT NOT NULL,
+    URLHash BIGINT NOT NULL,
+    CLID INTEGER NOT NULL,
+    PRIMARY KEY (CounterID, EventDate, UserID, EventTime, WatchID)
+)
+ORGANIZATION INDEX
+WITH COLUMN GROUP (all columns);

oceanbase-row/data-size

@@ -0,0 +1,14 @@
+#!/bin/bash
+# The observer preallocates both its data file (datafile_size) and its redo log
+# pool (log_disk_size) at startup, so `du` on the store directory reports the
+# reservation rather than the dataset -- tens of gigabytes of untouched zeroes.
+# DATA_DISK_IN_USE and LOG_DISK_IN_USE are the macroblocks and log blocks
+# actually occupied, which is the number this benchmark asks for: user data,
+# indexes, and transaction log.
+set -eu
+
+OB_HOME=${OB_HOME:-/opt/oceanbase}
+source "$OB_HOME/bench.env"
+
+"$OB_HOME/obc" -h127.0.0.1 -P"$OB_MYSQL_PORT" -uroot@sys -A -Doceanbase -N \
+    -e 'SELECT SUM(DATA_DISK_IN_USE) + SUM(LOG_DISK_IN_USE) FROM GV$OB_SERVERS'