Question

I have created a virtual file system(very simular to fat) based on two files.

1. First file – allocation table.

It store information about files allocation(actually it is not files but donnt cares about it)

Each record has following structure:

1. Name – fixed size(100 bytes). first 4 bytes is string length
2. Length – 4 bytes, total length of file
3. StartCluster – 4 bytes id of start cluster
4. EndCluster – 4 bytes id of end cluster

Each entry has fixed size and i have hashtable in memory which helps me to found position of each entry.

1. Second file – Virtual drive

VD based on clusters. Each cluster has fixed size – 256 bytes. Last 4 bytes are pointer to next cluster in file chain.

The problem is very slow speed when im trying to read all files. How can i improve perfomance? Is there any tips to fast reading from hard drive.

For example: Is it good idea to read file by big blocks? When i read even a little part of file, file is cached by OS right? And in next time i just take data from memory not from HD?

Well i have a few such questions where i can get answers?

Answer 1

Some options;

• You could enlarge your cluster size (256 bytes is small, most OS’es use 4KB+ for a cluster these days)

• If you read all files, you could do a sort based on startCluster so you read files in the order that they are physically close to each other on disk. So whenever the OS fetches a 4K+ block it is more likely you will need other parts of it with the next file.

• You could defrag your virtual disk file

• You seem confident it is a diskread issue. Did you check that what you do with the file after you just read it isn’t the slow part?

• A lot of random access is where SSD storage shines. Move the virtual disk to an SSD