ZFS Arc or L2Arc defaults for read ahead are quite optimal for all sort of workloads. You can modify for a certain workload but this can decrease performance on other workloads. In general these settings can increase performance for a few percent only.
If you say that your read performance is 30 MB/s then this is bad. If you can tune to increase 30% then you are at 40 MB/s what is bad as well. I do not asume that 30% is possible.
I would first look why this is so low. Maybe one or some disks are bad. Check iostat on load if all disks have same load. You can also build a pool per disk and compare performance. When ZFS writes to a pool it spreads data over the pool ao achieve a constant performance with many files and users. ZFS does not optimize for a single user with a large file. The datasheet disk values are lab values when you fill a disk track ba track. This is not the case with Raid or ZFS. But in general you should achieve at least 80 MB/s per disk.
Mostly sequential or large fileread performance depend on pool layout like n-way mirrors where ZFS can read from all mirror disks concurrently (a mirror can have 2x single disk performance)) or n-vdev Raid-Z and raw single single disk performance. A 1M recsice can help a little over a default 128K with the price of inefficiency on small files.
Arc and L2Arc tunings can help a little on small io but cache is for most/last read ZFS datablocks not files. Mostly performance is bad for small io. L2Arc can help a little as it is persistent or on low RAM situations. If you have enough RAM or not a large number of users with volatile files ex a university mailserver, forget L2Arc. A special vdev mirror is a better option. If you set a small blocksize for example of 64GB with a recsize of 128K+, then all files with size up to 64K (nearly all "office files") are stored on the special vdev, not only metadata. If you set 64K and a recsize 64K, all files of this filesystem are on the special vdev so you can set per filesystem workload.
But first, investigate why average single disk performance is so low.