我有一个具有此模式的表：

CREATE TABLE `data_realtime` (
     `id` mediumint(9) unsigned NOT NULL AUTO_INCREMENT,
     `timestamp` int(10) NOT NULL,
     `ticker_id` smallint(5) unsigned NOT NULL,
     `price` decimal(7,2) unsigned NOT NULL,
     `volume` mediumint(9) unsigned NOT NULL,
     `bid` decimal(7,2) unsigned DEFAULT NULL,
     `bid_sz` smallint(6) unsigned DEFAULT NULL,
     `ask` decimal(7,2) unsigned DEFAULT NULL,
     `ask_sz` smallint(6) unsigned DEFAULT NULL,
     PRIMARY KEY (`id`),
     UNIQUE KEY `ticker_timestamp` (`ticker_id`,`timestamp`) USING BTREE,
     CONSTRAINT `data_realtime_ibfk_2` FOREIGN KEY (`ticker_id`) REFERENCES `tickers` (`id`) ON DELETE CASCADE ON UPDATE CASCADE
    ) ENGINE=InnoDB AUTO_INCREMENT=11330043 DEFAULT CHARSET=latin1

我正在尝试运行一个简单的查询，按时间戳对数据进行排序：

select * from data_realtime ORDER BY timestamp ASC

这需要2.5秒（约4.5米的行，最终将增加到约12米的行）。但如果我只是跑

select * from data_realtime

需要25秒
我在timestamp上有一个复合索引（带有ticker\ id），我认为这有助于解决这个问题。
订购时我可以做些什么来提高性能？
谢谢。
编辑：要添加到原始问题，我有以下查询：

SELECT data_latest.*, data_1m.timestamp timestamp_1m, data_1m.price price_1m, data_1m.volume volume_1m FROM 
    (SELECT B.* FROM 
        (SELECT ticker_id, max(timestamp) max_timestamp FROM `data_rt` GROUP BY ticker_id) 
    A 
    LEFT JOIN 
    data_rt B 
    ON 
    A.ticker_id=B.ticker_id 
    and A.max_timestamp=B.timestamp) 
data_latest 
LEFT JOIN 
data_rt data_1m 
ON 
data_latest.timestamp <= (data_1m.timestamp + (60*1) ) 
AND data_latest.timestamp > (data_1m.timestamp + 60*(1-0.5)) 
AND data_latest.timestamp>data_1m.timestamp 
AND data_latest.ticker_id=data_1m.ticker_id 
ORDER BY data_1m.timestamp ASC

在一组1米的行上，大约需要1.3秒。把最后一个订单加起来会大大增加时间。如果我改为按时间戳订购，只需要0.05秒。
使用临时列进行排序时，我可以做些什么改进？