php - 100% SSD usage Linux

Every 20-30 seconds my HDD usage goes to 100% (iostat).

iotop is showing that [flush-8:0] is using 99% HDD during these times. In between HDD usage is 1-10%.

iostat output:

04/22/2013 08:58:44 AM
Device:         rrqm/s   wrqm/s     r/s     w/s    rMB/s    wMB/s avgrq-sz avgqu-sz   await  svctm  %util
sda               1.55  1188.88    3.43  569.93     0.03     6.88    24.69     0.25    0.43   0.12   7.15

04/22/2013 08:58:46 AM
Device:         rrqm/s   wrqm/s     r/s     w/s    rMB/s    wMB/s avgrq-sz avgqu-sz   await  svctm  %util
sda               0.00   632.50    1.00  753.50     0.01     5.41    14.72     0.77    1.02   0.02   1.35

04/22/2013 08:58:48 AM

Device:         rrqm/s   wrqm/s     r/s     w/s    rMB/s    wMB/s avgrq-sz avgqu-sz   await  svctm  %util
sda               0.00  1001.00    4.50   26.50     0.04     4.01   267.74     0.08    1.63   1.15   3.55

04/22/2013 08:58:50 AM
Device:         rrqm/s   wrqm/s     r/s     w/s    rMB/s    wMB/s avgrq-sz avgqu-sz   await  svctm  %util
sda               0.00     0.00    2.00    0.00     0.03     0.00    26.00     0.00   16.75   1.50   0.30

04/22/2013 08:58:52 AM
Device:         rrqm/s   wrqm/s     r/s     w/s    rMB/s    wMB/s avgrq-sz avgqu-sz   await  svctm  %util
sda               0.00  2332.50    2.00 5370.00     0.03    30.04    11.46   113.70   20.79   0.15  79.30


04/22/2013 08:58:54 AM
Device:         rrqm/s   wrqm/s     r/s     w/s    rMB/s    wMB/s avgrq-sz avgqu-sz   await  svctm  %util
sda               0.00   508.50    3.50 2102.00     0.03    10.21     9.96   143.96   63.78   0.47  99.50

04/22/2013 08:58:56 AM
Device:         rrqm/s   wrqm/s     r/s     w/s    rMB/s    wMB/s avgrq-sz avgqu-sz   await  svctm  %util
sda               0.00    46.50    1.50  423.50     0.01     1.85     8.95   117.26  288.18   2.35 100.05

04/22/2013 08:58:58 AM

Device:         rrqm/s   wrqm/s     r/s     w/s    rMB/s    wMB/s avgrq-sz avgqu-sz   await  svctm  %util
sda               0.00   924.50    3.00   34.00     0.02     3.76   209.30     1.04  203.03   1.73   6.40

04/22/2013 08:59:00 AM
Device:         rrqm/s   wrqm/s     r/s     w/s    rMB/s    wMB/s avgrq-sz avgqu-sz   await  svctm  %util
sda               0.00     0.00    3.50    0.00     0.04     0.00    21.71     0.03    8.43   8.43   2.95

04/22/2013 08:59:02 AM
Device:         rrqm/s   wrqm/s     r/s     w/s    rMB/s    wMB/s avgrq-sz avgqu-sz   await  svctm  %util
sda               9.00  2662.50    9.00 1135.50     0.08    14.86    26.72     1.03    0.90   0.04   4.50

iotop:

[root@a18 ~]# iotop -o -a
unable to set locale, falling back to the default locale

Total DISK READ :      19.47 K/s | Total DISK WRITE :       2.00 M/s
Actual DISK READ:      19.47 K/s | Actual DISK WRITE:       0.00 B/s

  TID  PRIO  USER     DISK READ  DISK WRITE  SWAPIN     IO>    COMMAND                                                                                                            
 2055 be/4 root          0.00 B   1968.00 K  0.00 % 15.75 % [flush-8:0]
20991 be/4 lighttpd      7.90 M      0.00 B  0.00 %  7.05 % lighttpd -f /etc/lighttpd/lighttpd.conf
23832 be/4 root         36.00 K    714.59 M  0.00 %  6.94 % php /var/www/base/bg-worker.php

How can I figure out what is this problem caused by?

Using SSD RAID 1. Filesystem EXT4.

I have PHP server with heavy writing and lots of small files deletion.

CentOS 6 64-bit.