add v0.6.1 migration guide; mention dataset conversion in import docs

docs/Migration.md

@@ -0,0 +1,97 @@
+# From Cayley 0.6.1 to 0.7.x
+
+First you need to dump all the data from the database via v0.6.1:
+
+```bash
+./cayley-0.6 dump --db <backend> --dbpath <address> --dump_type pquads --dump ./data.pq.gz
+```
+
+or using config file:
+
+```bash
+./cayley-0.6 dump --config <config> --dump_type pquads --dump ./data.pq.gz
+```
+
+And load the data into new database via v0.7.x (`pq` file extension is important):
+
+```bash
+./cayley load --init -d <new-backend> -a <new-address> -i ./data.pq.gz
+```
+
+or using config file:
+
+```bash
+./cayley load --init -c <new-config> -i ./data.pq.gz
+```
+
+## Dump via text format
+
+An above guide uses Cayley-specific binary format to avoid encoding and parsing overhead and to compress output file better.
+
+As an alternative, a standard nquads file format can be used to dump and load data (note `nq` extension):
+
+```bash
+./cayley-0.6 dump --config <config> --dump ./data.nq.gz
+./cayley load --init -c <new-config> -i ./data.nq.gz
+```
+
+## Bolt, LevelDB, MongoDB
+
+Cayley v0.7.0 is still able to read and write databases of these types in old format. It can be accessed by changing
+backend name from `bolt`/`leveldb`/`mongo` to `bolt1`/`leveldb1`/`mongo1`. Thus, you can use guide for moving data between
+different backends for v0.7.x (see below).
+
+**Note:** support for old versions will be dropped starting from v0.7.1.
+
+## SQL
+
+Data format for SQL between v0.6.1 and v0.7.x has not changed significantly. Database can be upgraded
+by executing the following SQL statements:
+
+```sql
+ALTER TABLE quads DROP id;
+ALTER TABLE nodes ADD refs INT DEFAULT 0 NOT NULL;
+ALTER TABLE nodes ALTER COLUMN refs DROP DEFAULT;
+UPDATE nodes SET refs = 
+  (SELECT count(1) FROM quads WHERE subject_hash = nodes.hash) +
+  (SELECT count(1) FROM quads WHERE predicate_hash = nodes.hash) +
+  (SELECT count(1) FROM quads WHERE object_hash = nodes.hash) +
+  (SELECT count(1) FROM quads WHERE label_hash = nodes.hash);
+```
+
+# From different backend (Cayley 0.7+)
+
+First you need to dump all the data from old backend (`pq` extension is important):
+
+```bash
+./cayley dump -d <backend> -a <address> -o ./data.pq.gz
+```
+
+or using config file:
+
+```bash
+./cayley dump -c <config> -o ./data.pq.gz
+```
+
+And load the data into a new backend and/or database:
+
+```bash
+./cayley load --init -d <new-backend> -a <new-address> -i ./data.pq.gz
+```
+
+or using config file:
+
+```bash
+./cayley load --init -c <new-config> -i ./data.pq.gz
+```
+
+## Dump via text format
+
+An above guide uses Cayley-specific binary format to avoid encoding and parsing overhead and to compress output file better.
+
+As an alternative, a standard nquads file format can be used to dump and load data (note `nq` extension):
+
+```bash
+./cayley dump -c <config> -o ./data.nq.gz
+./cayley load --init -c <new-config> -i ./data.nq.gz
+```

docs/Quickstart-As-Application.md

@@ -57,6 +57,15 @@ And wait. It will load. If you'd like to watch it load, you can run
 
 And watch the log output go by.
 
+If you plan to import a large dataset into Cayley and try multiple backends, it makes sense to first convert the dataset
+to Cayley-specific binary format by running:
+
+```bash
+./cayley conv -i dataset.nq.gz -o dataset.pq.gz
+```
+
+This will minimize parsing overhead on future imports and will compress dataset a bit better.
+
 ### Connect a REPL To Your Graph
 
 Now it's loaded. We can use Cayley now to connect to the graph. As you might have guessed, that command is: