Specifications

Escala Tower PL & S, E, T System Hardware
Chapter 3: External Storage and SAN Subsystems 43/116
9es3s1c3.doc
Rev 5.9
02/12/2003
3.15. Performance
Performance takes advantage of the capability of mixing RAID architectures to fit application needs.
The appropriate RAID level must be chosen according to the file size and the type of access. The write
cache drops the write response time for all RAID levels, sweeps out the RAID 5 penalty and boosts the
read after write access.
The raw performance of the DAS1300 reaches in RAID 5 for one 5-disk LUN:
•
350 I/Os per second (100 % of READ, random access of 512 bytes, without cache)
•
120 I/Os per second (100% of WRITE, random access of 512 bytes, with an 8MB cache).
•
a 13ms response time for 20 I/O per second (read without cache)
•
a 7.3 ms response time for 20 I/O per second (write with an 8MB mirrored cache)
•
a 5 ms response time for 20 I/O per second (write with an 8MB single cache).
The raw performance of the DAS2900 is much higher for large configurations (more than 12 drives).
•
1200 I/Os per second (100 % of READ, random access of 4 Kbytes, cache hit = 0, access time =
32 ms, 20-disk 4-LUN RAID5, Dual SP Dual SCSI configuration).
•
4 ms less than DAS1300 in access time for 4 Kbytes I/Os in the same conditions.
•
500 I/Os per second (100 % of WRITE, random access of 4 Kbytes, 30 MB of mirrored write
cache, 20-disk 4-LUN RAID5, Dual SP Dual SCSI configuration).
•
340 I/Os per second (100 % of WRITE, random access of 4 Kbytes, 30 MB of mirrored write
cache, 20-disk 4-LUN RAID5, Single SP Single SCSI configuration).
•
a 7 ms response time for 20 I/O per second (100 % of WRITE, random access of 4 Kbytes, 30 MB
of mirrored write cache, 20-disk 4-LUN RAID5, Single SP Single SCSI configuration).
These values are worst cases, measured within a full size working space, the access time includes a
higher seek time than the one usual applications get.