- ...Ioannidis
-
Current affiliation: Computer Science Department,
University of Rochester, Rochester, NY14627
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
- ...Markatos
- Evangelos Markatos
and Julia Sevaslidou are also with the University of Crete,
Department of Computer Science
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
- ...crash.
-
We increase the reliability of remote main memory by
using redundant power supplies (UPS) to survive power failures, and
mirroring of data to survive software failures.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
- ...25].
-
Although optimizations like group transaction commit have been
proposed, these may increase the complexity of the transaction-based system,
and thus have not been incorporated in the EXODUS and RVM systems.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
- ...probability.
- Simple math calculations suggest
that if each workstation crashes once every few months, and stays crashed
for several minutes, two workstations will crash
within the same time interval once every several years, which leads
to higher reliability than current disks provide. If, however, this
level of reliability is not enough, data can be replicated to three
main memories, etc.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
- ...(UPS).
-
Alternatively, both power supplies may be provided by two different UPSs.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
- ...disks.
- Note, however, that the
UPS itself may malfunction, leading to power loss.
Such malfunctions however, happen once every several years, making the
UPS more reliable machine than the magnetic disk. If, however,
a UPS malfunctions more frequently, sensitive data can be replicated
to different workstations connected to different
UPSs that malfunction independent of each other.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
- ...disk.
- UNIX-derivative
systems force all their modified data to the disk every 30-60 seconds.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
- ...EXODUS).
- Our modification were rather small.
Out of about 30,000 lines of RVM
code, we modified (or added) less than 600 lines (2%).
Out of 200,000 lines of EXODUS code, we modified (or added) less than
700 lines (a 0.3% change).
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
- ...imposes.
- Due to technical difficulties
we were not able to run these experiments
on top of Memory Channel.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
- ...second.
-
Since each client/server pair sustains 150 transactions per second,
the total number of transactions per second sustained
in the Ethernet-based system is
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.