Re: [SR-Users] HTable 'size' parameter

2010-10-21 Thread Elena-Ramona Modroiu
Maybe the term entry should be replaced with slot, looks to be more standard. A good doc about hash tables is on wikipedia: http://en.wikipedia.org/wiki/Hash_table The htable module keeps the number of items in a slot, but not exported via rpc, this could be added in the future. Ramona On

Re: [SR-Users] HTable 'size' parameter

2010-10-17 Thread Iñaki Baz Castillo
2010/10/16 Uriel Rozenbaum : > What would be best to optimize the system is to have some statistic > information, save every now and then the status of the htable and act > accordingly to: > - If the table is pretty much or completely used and you have > collisions, make the size parameter bigger >

Re: [SR-Users] HTable 'size' parameter

2010-10-16 Thread Alex Balashov
On 10/16/2010 09:18 AM, Juha Heinanen wrote: Uriel Rozenbaum writes: What would be best to optimize the system is to have some statistic information, save every now and then the status of the htable and act accordingly to: it would indeed be nice of htable module would exports stats about col

Re: [SR-Users] HTable 'size' parameter

2010-10-16 Thread Juha Heinanen
Uriel Rozenbaum writes: > What would be best to optimize the system is to have some statistic > information, save every now and then the status of the htable and act > accordingly to: it would indeed be nice of htable module would exports stats about collisions. -- juha

Re: [SR-Users] HTable 'size' parameter

2010-10-16 Thread Uriel Rozenbaum
Iñaki, Just as an addition from experience on hash tables, the first "sweet-spot" on the design is to have a fairly good algorithm so that different keys are mapped on different hash values. You cannot pick this (as far as I read) but we'll think it as fair enough. What you can do to "help" the al

Re: [SR-Users] HTable 'size' parameter

2010-10-16 Thread Juha Heinanen
Iñaki Baz Castillo writes: > And those collisions mean that I could look for a key name and > retrieve another key (as both key names produce the same hash), am I > right? no, the values whose keys hash to the same index of hash table, are added into a linked list and then list list will be searc

Re: [SR-Users] HTable 'size' parameter

2010-10-16 Thread Iñaki Baz Castillo
2010/10/16 Juha Heinanen : > size=4 means that the hash table has 16 entries.  it does not mean that > the table could not contain more objects, but collisions will happen. And those collisions mean that I could look for a key name and retrieve another key (as both key names produce the same hash)

[SR-Users] HTable 'size' parameter

2010-10-16 Thread Juha Heinanen
Iñaki Baz Castillo writes: > Hi, in the doc of HTable module all the examples show "a=>size=4;". As > the doc says: > > size - number specifying the size of hash table. The number of entries in >the table is 2^size > > This is, the examples just allow 16 concurrent entries, which i

[SR-Users] HTable 'size' parameter

2010-10-16 Thread Iñaki Baz Castillo
Hi, in the doc of HTable module all the examples show "a=>size=4;". As the doc says: size - number specifying the size of hash table. The number of entries in the table is 2^size This is, the examples just allow 16 concurrent entries, which is not enough in most of the cases. Did I m