On Thursday 15 May 2008 06:11, Daniel Cheng wrote:
> On Thu, May 15, 2008 at 7:08 AM, Matthew Toseland
> <toad at amphibian.dyndns.org> wrote:
> > On Sunday 11 May 2008 11:24, j16sdiz at freenetproject.org wrote:
> >> Author: j16sdiz
> >> Date: 2008-05-11 10:24:22 +0000 (Sun, 11 May 2008)
> >> New Revision: 19891
> >>
> >> Modified:
> >>
> >
branches/saltedhashstore/freenet/src/freenet/store/SaltedHashFreenetStore.java
> >> Log:
> >> Quadratic Probing (datastore resize disabled)
> >
> > Cool! Any test results yet?
>
> It is working better then the previous code..
> The store can grow ~70% full with no much resistant.
That's good... after that it significantly slows down? Can we have a rough
idea of how long it takes to get to 50%, 60%, 70%, 80%, 90%?
>
> >> Modified:
> >
branches/saltedhashstore/freenet/src/freenet/store/SaltedHashFreenetStore.java
> >> ===================================================================
> >> ---
> >
branches/saltedhashstore/freenet/src/freenet/store/SaltedHashFreenetStore.java
> > 2008-05-11 09:08:40 UTC (rev 19890)
> >> +++
> >
branches/saltedhashstore/freenet/src/freenet/store/SaltedHashFreenetStore.java
> > 2008-05-11 10:24:22 UTC (rev 19891)
> >> @@ -126,22 +126,29 @@
> >>
> >> private Entry probeEntry0(byte[] routingKey, long probeStoreSize)
throws
> > IOException {
> >> Entry entry;
> >> - long offset = getOffsetFromPlainKey(routingKey,
probeStoreSize);
> >> + long[] offset = getOffsetFromPlainKey(routingKey,
probeStoreSize);
> >>
> >> - if (!lockEntry(offset)) {
> >> - Logger.error(this, "can't lock entry: " + offset);
> >> - return null;
> >> + for (int i = 0; i < offset.length; i++) {
> >> + if (logDEBUG)
> >> + Logger.debug(this, "probing for i=" + i
+ ", offset=" + offset[i]);
> >> +
> >> + if (!lockEntry(offset[i])) {
> >> + Logger.error(this, "can't lock entry: " +
offset[i]);
> >> + continue;
> >> + }
> >> + try {
> >> + entry = readEntry(offset[i], routingKey);
> >> + } catch (EOFException e) {
> >> + // may occur on resize, silent it a bit
> >> + Logger.error(this, "EOFException on
probeEntry", e);
> >> + return null;
> >
> > Shouldn't we go around the loop here? Or are the offsets strictly
increasing?
>
> Fixed in my local repository.
>
> >> + } finally {
> >> + unlockEntry(offset[i]);
> >> + }
> >> + if (entry != null)
> >> + return entry;
> >> }
> >> - try {
> >> - entry = readEntry(offset, routingKey);
> >> - } catch (EOFException e) {
> >> - // may occur on resize, silent it a bit
> >> - Logger.error(this, "EOFException on probeEntry",
e);
> >> - return null;
> >> - } finally {
> >> - unlockEntry(offset);
> >> - }
> >> - return entry;
> >> + return null;
> >> }
> >>
> >> public void put(StorableBlock block, byte[] routingKey, byte[]
fullKey,
> > byte[] data, byte[] header,
> >> @@ -163,16 +170,39 @@
> >> }
> >>
> >> Entry entry = new Entry(routingKey, header, data);
> >> - if (!lockEntry(entry.getOffset())) {
> >> + long[] offset = entry.getOffset();
> >> +
> >> + for (int i = 0; i < offset.length; i++) {
> >> + if (!lockEntry(offset[i])) {
> >> + Logger.error(this, "can't lock entry: " +
offset[i]);
> >> + return;
> >> + }
> >> + try {
> >> + if (isFree(offset[i], entry)) {
> >> + if (logDEBUG)
> >> +
Logger.debug(this, "probing, write to i=" + i + ", offset=" +
> > offset[i]);
> >> + writeEntry(entry, offset[i]);
> >> + incWrites();
> >> + return;
> >> + }
> >> + } finally {
> >> + unlockEntry(offset[i]);
> >> + }
> >> + }
> >> +
> >> + // no free blocks?
> >> + int i = random.nextInt(offset.length);
> >
> > You should try to write to an empty slot, of course... if all are full, is
it
> > better to overwrite an empty slot rather than the first one?
>
> All slot in offset[] are occupied when we reach this code.
Ok.
> Have to pick a random one to overwrite
Is it better to overwrite a random one or the first one?
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL:
<https://emu.freenetproject.org/pipermail/devl/attachments/20080515/7445586d/attachment.pgp>
