Latest revision as of 14:42, 16 October 2017

Overview

QED was an attempt at creating a better performing image format by removing some features compared to qcow2. However, it turned out that the achieved performance improvements were mostly related to an improved implementation rather than the file format per se. These improvements, as well as a few format extensions, have been merged back into qcow2 since. The development of QED has been abandoned.

It is not recommended to use QED for any new images. For existing images, converting to qcow2 should be considered as today qcow2 provides both more features and better performance, and has an actively maintained code base.

QED supports backing files and sparse images.

Status

QED is deprecated and only supported for compatibility with existing images (similar to qcow1)
Base QED is in qemu.git since 2010-12-17 and will form part of QEMU 0.14.
No additional features are planned to get merged

Features

Open specification
Fully asynchronous I/O path
Strong data integrity due to simple design
Backing files
- Backing files may be smaller than the QED image
Sparse files
- Retains sparseness over non-sparse channels (e.g. HTTP)
Zero clusters
Periodic dirty flag flush

Uncompleted work

@@ Line 1: / Line 1: @@
-=Specification=
+= Overview =
-The file format looks like this:
+QED was an attempt at creating a better performing image format by removing some features compared to qcow2. However, it turned out that the achieved performance improvements were mostly related to an improved implementation rather than the file format per se. These improvements, as well as a few format extensions, have been merged back into qcow2 since. The development of QED has been abandoned.
- +----------+----------+----------+-----+
+'''It is not recommended to use QED for any new images.''' For existing images, converting to qcow2 should be considered as today qcow2 provides both more features and better performance, and has an actively maintained code base.
- | cluster0 | cluster1 | cluster2 | ... |
- +----------+----------+----------+-----+
-The first cluster contains a header.  The header contains information about where regular clusters start; this allows the header to be extensible and store extra information about the image file.  A regular cluster may be a data cluster, an L2, or an L1 table.  L1 and L2 tables are composed of one or more contiguous clusters.
+QED supports backing files and sparse images.
-Normally the file size will be a multiple of the cluster size.  Extra information at the end of the file may not be preserved if new clusters are allocated.  Legitimate out-of-band data should use space between the header and the first regular cluster.
+= Status =
-All fields are little-endian.
+* '''QED is deprecated and only supported for compatibility with existing images''' (similar to qcow1)
+* Base QED is in qemu.git since [http://git.qemu.org/qemu.git/commit/?id=75411d236d93d79d8052e0116c3eeebe23e2778b 2010-12-17] and will form part of QEMU 0.14.
+* No additional features are planned to get merged
-==Header==
+= Features =
- Header {
-     uint32_t magic;               /* QED\0 */
-     uint32_t cluster_size;        /* in bytes */
+* [[Features/QED/Specification|Open specification]]
-     uint32_t table_size;          /* in clusters */
+* Fully asynchronous I/O path
-     uint32_t header_size;         /* in clusters */
+* Strong data integrity due to simple design
+* Backing files
-     uint64_t features;            /* format feature bits */
+** Backing files may be smaller than the QED image
-     uint64_t compat_features;     /* compat feature bits */
+* Sparse files
-     uint64_t l1_table_offset;     /* in bytes */
+** Retains sparseness over non-sparse channels (e.g. HTTP)
-     uint64_t image_size;          /* total logical image size, in bytes */
+* Zero clusters
+* Periodic dirty flag flush
-     /* if (features & QED_F_BACKING_FILE) */
-     uint32_t backing_filename_offset; /* in bytes from start of header */
-     uint32_t backing_filename_size;   /* in bytes */
-     /* if (compat_features & QED_CF_BACKING_FORMAT) */
-     uint32_t backing_fmt_offset;  /* in bytes from start of header */
-     uint32_t backing_fmt_size;    /* in bytes */
- }
-==Tables==
+= Uncompleted work =
- #define TABLE_NOFFSETS (table_size * cluster_size / sizeof(uint64_t))
+* [[Features/QED/OutstandingWork|Outstanding work]]
+* [[Features/QED/Streaming|Streaming]]
- Table {
+* [[Features/QED/OnlineDefrag|Online defragmentation]]
-     uint64_t offsets[TABLE_NOFFSETS];
+* [[Features/QED/ParallelSubmission|Parallel submission]]
- }
+* [[Features/QED/ScanAvoidance|Meta-data scan avoidance]]
-The tables are organized as follows:
+[[Category:Obsolete feature pages]]
-                    +----------+
-                    | L1 table |
-                    +----------+
-               ,------'  |  '------.
-          +----------+   |    +----------+
-          | L2 table |  ...   | L2 table |
-          +----------+        +----------+
-      ,------'  |  '------.
- +----------+   |    +----------+
- |   Data   |  ...   |   Data   |
- +----------+        +----------+
-A table is made up of one or more contiguous clusters.  The table_size header field determines table size for an image file.  For example, cluster_size=64 KB and table_size=4 results in 256 KB tables.
-=Operations=
-==Read==
-# If L2 table is not present in L1, read from backing image.
-# If data cluster is not present in L2, read from backing image.
-# Otherwise read data from cluster.
-==Write==
-# If L2 table is not present in L1, allocate new cluster and L2.  Perform L2 and L1 link after writing data.
-# If data cluster is not present in L2, allocate new cluster.  Perform L1 link after writing data.
-# Otherwise overwrite data cluster.
-The L2 link '''should''' be made after the data is in place on storage.  However, when no ordering is enforced the worst case scenario is an L2 link to an unwritten cluster.
-The L1 link '''must''' be made after the L2 cluster is in place on storage.  If the order is reversed then the L1 table may point to a bogus L2 table.  (Is this a problem since clusters are allocated at the end of the file?)
-==Grow==
-# If table_size * TABLE_NOFFSETS < new_image_size, fail -EOVERFLOW.  The L1 table is not big enough.
-# Write new image_size header field.
-=Data integrity=
-==Write==
-Writes that complete before a flush must be stable when the flush completes.
-If storage is interrupted (e.g. power outage) then writes in progress may be lost, stable, or partially completed.  The storage must not be otherwise corrupted or inaccessible after it is restarted.