Postgres：优化urn的性能加入匹配与INT

Question

在我们的数据库中，我们正在思考到处使用urns而不是ID。

例如urn：ny：mytypeofobject：123

上测试性能

CREATE table addresses AS 
SELECT 
i as id_int,
CONCAT('urn:ny:address:', i::text) as id_urn
FROM 
generate_series(1, 100000000) s(i);  -- 100M


CREATE table addresses_attrib AS 
SELECT 
i as id_int,
CONCAT('urn:ny:address:', i::text) as id_urn,
i%79 as random_filter
FROM 
generate_series(1, 10000000) s(i); -- 10M

-- CREATE index both on id_int and id_urn
CREATE index int_a1 on addresses(id_int);
CREATE index int_aa1 on addresses_attrib(id_int);
CREATE index urn_a1 on addresses(id_urn);
CREATE index urn_aa1 on addresses_attrib(id_urn);

SELECT * FROM addresses_attrib aa
JOIN addresses a on (aa.id_int=a.id_int)
WHERE aa.random_filter=4
LIMIT 100000

使用pg_bench在VS

SELECT * FROM addresses_attrib aa
JOIN addresses a on (aa.id_urn=a.id_urn)
WHERE aa.random_filter=4
LIMIT 100000

，与pg_bench相比，基于int的int基于int的速度约为速度（0.5 avg int）（0.5 avg int）。

有什么办法可以加快这种加入在urns上？