Don't use XML where it doesn't make sense. XML is not a panacea. You will not get good performance by transferring and parsing a lot of XML files.

Using XML is memory, CPU, and network intensive.

Avoid creating a new parser each time you parse; reuse parser instances. A pool of reusable parser instances might be a good idea if you have multiple threads parsing at the same time.

• Convert the document to US ASCII ("US-ASCII") or Unicode ("UTF-8" or "UTF-16") before parsing. Documents written using ASCII are the fastest to parse because each character is guaranteed to be a single byte and map directly to their equivalent Unicode value. For documents that contain Unicode characters beyond the ASCII range, multiple byte sequences must be read and converted for each character. There is a performance penalty for this conversion. The UTF-16 encoding alleviates some of this penalty because each character is specified using two bytes, assuming no surrogate characters. However, using UTF-16 can roughly double the size of the original document which takes longer to parse.
• Explicitly specify "US-ASCII" encoding if your document is in ASCII format. If no encoding is specified, the XML specification requires the parser to assume UTF-8 which is slower to process.
• Avoid external entities and external DTDs. The extra file opens and transcoding setup is expensive.
• Reduce character count; smaller documents are parsed quicker. Replace elements with attributes where it makes sense. Avoid gratuitous use of whitespace because the parser must scan past it.
• Avoid using too many default attributes. Defaulting attribute values slows down processing.

• Turn validation off if you don't need it. Validation is expensive. Also, avoid using a DOCTYPE line in your XML document. The current version of the parser will always read the DTD if the DOCTYPE line is specified even when not validating.
• For large documents, avoid using DOM which uses a lot of memory. Instead, use SAX if appropriate. The DOM parser requires that the entire document be read into memory before the application processes the document. The SAX parser uses very little memory and notifies the application as parts of the document are parsed.