Re: [tahoe-dev] Automatic rebalancing

2010-12-05 Thread Brian Warner
On 12/5/10 3:52 PM, James A. Donald wrote: > > A centralized coordinator is single point of failure and an additional > configuration issue. If everyone runs the same algorithm, they will > mostly agree without need for a central coordinator - though there > will never be 100% agreement. If the sy

Re: [tahoe-dev] Automatic rebalancing

2010-12-05 Thread James A. Donald
On 2010-12-06 4:43 AM, Ravi Pinjala wrote: I'm pretty sure Ceph runs CRUSH on the servers, too - it's how they handle rebalancing, after all. The biggest issue I see there is that the way they use it depends on setting a new CRUSH map across the entire cluster atomically, which only works on Ceph

Re: [tahoe-dev] Automatic rebalancing

2010-12-05 Thread Ravi Pinjala
I'm pretty sure Ceph runs CRUSH on the servers, too - it's how they handle rebalancing, after all. The biggest issue I see there is that the way they use it depends on setting a new CRUSH map across the entire cluster atomically, which only works on Ceph because they use a Paxos cluster to coordina

Re: [tahoe-dev] Automatic rebalancing

2010-12-05 Thread Greg Troxel
Ravi Pinjala writes: > As far as description languages for data allocation go, Ceph has > already solved this problem - check out the "CRUSH" algorithm. > Basically, it's a description language for data placement that > controls replication and data placement, and I think it also lets > clients

Re: [tahoe-dev] Automatic rebalancing

2010-12-05 Thread Terrell Russell
CRUSH seems to require knowledge of a hierarchy (location) of devices (location, racks, etc). I'm not sure this knowledge is available to tahoe. Or any knowledge we have should be assumed as flat. http://ceph.newdream.net/wiki/Custom_data_placement_with_CRUSH Terrell On 12/5/10 4:42 AM, Ravi

Re: [tahoe-dev] Automatic rebalancing

2010-12-05 Thread Ravi Pinjala
As far as description languages for data allocation go, Ceph has already solved this problem - check out the "CRUSH" algorithm. Basically, it's a description language for data placement that controls replication and data placement, and I think it also lets clients figure out which servers a piece o

[tahoe-dev] Automatic rebalancing

2010-12-05 Thread Shu Lin
Hi, As the answer of this discussion, Tahoe doesn't have automatic rebalancing capability now. http://tahoe-lafs.org/pipermail/tahoe-dev/2010-December/005697.html Also, we have bunch of tickets already tracking this problem. Such as the rebalancing manager: http://tahoe-lafs.org/trac/tahoe-lafs/t