On Mon, Nov 28, 2011 at 7:48 PM, Gregory Farnum <gregory.farnum@xxxxxxxxxxxxx> wrote: > (Added list for future reference.) > On Nov 28, 2011, at 5:47 PM, Xiaofei Du <xiaofei.du008@xxxxxxxxx> wrote: > > Do you remember we talked several days before about the problem I > encountered for testing ceph on ec2? The reason I ask this question is > I am still confused with one of the results. The following is what you > told me. > > You have two nodes that write at 20MB/s and six that write at 45MB/s. > > Journaling cuts each of those in half: two nodes at 10MB/s, six at > > 23MB/s. That's an aggregate 158MB/s of write bandwidth, but each write > > goes to two OSDs so you only get half of that, or 79MB/s of write > > bandwidth. You are getting about the bandwidth I would expect out of a > > slow and unbalanced cluster. > > My question is since journaling cuts the bandwidth of each node into > half. When I did the test for writing a single file. Why could I still > achieve a bandwidth around 45MB/s. If it goes to the 45MB/s node, it > could only achieve 23MB/s. My question is why it could still achieve > 45MB/s for writing a single file? Thanks a lot > > As you reference below, Ceph stripes files over objects (4MB by default). > When you write an object, it can buffer the write, and when it flushes the > buffer out to disk it can send objects out to multiple OSDs simultaneously. Awesome, Ceph's design is really nice in many details > > > On Nov 28, 2011, at 5:27 PM, Xiaofei Du <xiaofei.du008@xxxxxxxxx> wrote: > > Hi Gregory, > I have a question about Ceph's read and write mechanism. If I only write one > file or read one file at a time. Does it doing the read and the write > concurrently for each of the stripe? For example. I have a 160 MB file. > Ceph's stripe is 16MB, which means this file will be striped into 10 > objects. My question is when I do reading or writing for this file, will it > do it concurrently? > > Have you adjusted the stripe size? The default is 4MB. > In any case, yes, it will do concurrent IOs. > -Greg No. I didn't adjust the stripe size. Because I saw 16MB on the Ceph paper. So I thought the default one was 16MB. Thanks a lot again. Greg Best, Xiaofei -- Xiaofei (Gregory) Du Department of Computer Science University of California, Santa Barbara -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
