I write here only on the non-coding RNA classes. From my experience:

- in RNAseq from total RNA, rRNA depleted, I could see lncRNAs and snoRNAs. What you see depend on the size of insert selected for sequencing after library preparation. The biggest problem for me was the efficiency of depletion, which will depend from the kit and species you are using (you don't want 80% rRNA reads in your RNAseq).

- in RNAseq on polyA-tail captured RNA, I have seen many lncRNA, no miRNA (although sometimes host nc genes for miRNA)

- for sequencing miRNAs I  used size-fractionation after total RNA mirVana extraction. In the result you should be able to see picks of different length after clean up of adapters. Keep in mind that piwiRNAs are mostly in the germ line cells.

To summarize - for most ncRNA sequencing you want good quality total RNA, depleted from rRNA, checked on Bioanalyser. Then I would split samples into total (long) and short, size selected for miRNA-piwiRNA. All of that depends of the species you use, e.g. Drosophila small rRNA is a pain for miRNA extraction. You can also ask a sequencing platform/company what kits and protocols they usually use.

Adding some mess to Kasia's answer... I would like also to point out that you can detect circRNAs too. It is quite tricky but there are some tools that allow you to pinpoint non-canonical splicing events that will lead to the formation of circRNAs. For circRNAs you do not need any specific method, however there are some papers that came out from Gorospe's lab where they describe protocols for selective isolation of circRNAs.

Hope it helps. Best luck

dear  Luo A non-coding RNA (ncRNA) is a functional RNA molecule that is transcribed from DNA but not translated into proteins. Epigenetic related ncRNAs include miRNA, siRNA, piRNA and lncRNA. In general, ncRNAs function to regulate gene expression at the transcriptional and post-transcriptional level.  Those ncRNAs that appear to be involved in epigenetic processes can be divided into two main groups; the short ncRNAs (200 nts).  The three major classes of short non-coding RNAs are microRNAs (miRNAs), short interfering RNAs (siRNAs), and piwi-interacting RNAs (piRNAs). Both major groups are shown to play a role in heterochromatin formation, histone modification, DNA methylation targeting, and gene silencing.

Short ncRNAs

MicroRNAs (miRNA) generally bind to a specific target messenger RNA with a complementary sequence to induce cleavage, or degradation or block translation. This may be done in the context of a feedback mechanism that involves chromosome methylation. For example, miRNA genes mir-127 and mir-136 were found to be involved in regulating the genetic imprinting of Rtl1, a key gene involved in placenta formation in mice. Methylation of a specific region in the paternal chromosome results in expression of Rtl1. If the chromosome is not methylated, as on the maternal chromosome, mir-127 and mir-136 are produced and bind to the Rtl1 transcript and induce degradation. Lack of Rtl1 protein expression due to improper epigenetic modifications can result in fetal death in mice.12

Short interfering RNAs (siRNA) function in a similar way as miRNAs to mediate post-transcriptional gene silencing (PTGS) as a result of mRNA degradation. In addition to this function, siRNAs have also been shown to induce heterochromatin formation via an RNA-induced transcriptional silencing (RITS) complex which when bound to siRNA promotes H3K9 methylation and chromatin condensation.3

Piwi-interacting RNAs (piRNA) are so named due to their interaction with the piwi family of proteins. The primary function of these RNA molecules involves chromatin regulation and suppression of transposon activity in germline and somatic cells. PiRNAs that are antisense to expressed transposons target and cleave the transposon in complexes with PIWI-proteins.  This cleavage generates additional piRNAs which target and cleave additional transposons. This cycle continues to produce an abundance of piRNAs and augment transposon silencing.45

Long ncRNAs

Many lncRNAs can complex with chromatin-modifying proteins and recruit their catalytic activity to specific sites in the genome, thereby modifying chromatin states and influencing gene expression.6 The majority of non-coding RNA transcripts belong to the group lncRNAs. Long ncRNAs function in chromatin remodeling, transcriptional regulation, post-transcriptional regulation, and as precursors for siRNAs.7 One particular subgroup of lncRNAs, the large intergenic non-coding RNAs (lincRNAs), has been associated with chromatin modifying complexes which can target specific genomic loci to promote specific epigenetic states.8 One widely known example of this is the role of X-inactive specific transcript gene (Xist), in X-chromosome inactivation (XCI). This process involves two lncRNAs; Xist and its antisense transcript Tsix, a negative regulator of Xist. Prior to differentiation, Xist and Tsix are actively transcribed due to H3K4 dimethylation of the Xist gene. In this state XCI is a random event. Upon differentiation, Xist expression is elevated resulting in Xist RNA coating the future inactive X chromosome which triggers extensive histone methylation and chromosome inactivation

Thank you all for the detailed suggestions. Guess I could get rRNA depleted samples first, then start total RNA sequencing for mRNA&lncRNA and small RNA sequencing for other small RNA classes. 

I also find a paper "Deciphering the RNA landscape by RNAome sequencing" analyzed all RNA classes in one experiment. But this is the only paper I found using this method and I am not sure whether it's necessary to study that many RNA classes for my project.