HashMap backup to file with every update in Java

Question

I am wondering of the best way of backing up a hashmap onto disk with each of the maps .put calls. Currently what i have coded is to separate each object being written into a file where its hashcode is the files name and the contents is a hashmap of the entries that match (encase of collisions)

private File val2file(K k) {
    int hash = k.hashCode();
    //bitsIgnored if we ever want to store multiple objects per file
    hash = hash >>> bitsIgnored;
    return new File(rootDir, Integer.toString(hash));
}

with

File toEdit = val2file(k);

    ReentrantLock tempLock = new ReentrantLock();
    ReentrantLock lock = lockMap.putIfAbsent(toEdit, tempLock);
    if(lock == null) {
        lock = tempLock;
    }
    lock.lock();
    try {
        HashMap<K, V> storageMap;
        if(toEdit.exists()) {
            ObjectInputStream ois = new ObjectInputStream(new FileInputStream(toEdit));
            Object obj = ois.readObject();
            ois.close();
            storageMap = (HashMap<K, V>) obj;
        } else {
            storageMap = new HashMap<>();
        }
        storageMap.put(k, v);
        ObjectOutputStream oos = new ObjectOutputStream(new BufferedOutputStream(new FileOutputStream(toEdit)));
        oos.writeObject(storageMap);
        oos.close();
    } catch(FileNotFoundException ex) {
        Logger.getLogger(MapSaver.class.getName()).log(Level.SEVERE, null, ex);
    } catch(ClassNotFoundException ex) {
        Logger.getLogger(MapSaver.class.getName()).log(Level.SEVERE, null, ex);
    } catch(IOException ex) {
        if(lock.isLocked()) lock.unlock();
        throw ex;
    } finally {
        if(lock.isLocked()) lock.unlock();
    }

This does work and i have validated it for 1000 updates per second, with the limiting factor being the serialiser (CPU) and the input/outputs of the drive due to the amount of files its writing to. I have tried some database solutions and other "map to file" solutions suggested here but none seem as fast. I would rather a one file solution compared to my (n file) solution but i cant see anyway to do that for each single update. Does another have any alternative or improvement to the code?