Hi all,
Late last year I came into possession (dumpster dive find) of an HP ProLiant rackmount server. It started out with lackluster specs (single Xeon E5205, 2GB RAM, no storage, dead SmartArray/RAID battery), that I eventually upgraded the internals:
Dual Xeon X5270 (did not determine if the particular board supports quad-core/socket CPU's, played it safe with best-available dual-core CPU's instead)
32 GB fully-buffered DDR2 RAM (was my first upgrade via a subsequent dumpster dive)
250GB Mushkin SATA SSD
320GB Seagate 2.5" SATA HDD
Replacement SmartArray battery
With that I could complete an Oreo 8.0 build from clean in 55 minutes - my 3rd-gen i5 PC would do so in 40 - in spite of the Xeons calculating prime numbers faster then my i5, the server has twice the age in terms of I/O.
What I determined to be a significant barrier is the SATA-1 bus (that's right, the 1.5Gbps variety). SATA-1 has a theoretical maximum throughput of 187 MBps, but in real world, my SSD was still falling 50 MBps short in tests (and is not much faster than the HDD in this scenario for that matter):
Code:
[email protected] ~> sudo hdparm -tT /dev/sda
/dev/sda:
Timing cached reads: 15532 MB in 1.99 seconds = 7788.53 MB/sec
Timing buffered disk reads: 412 MB in 3.01 seconds = 137.06 MB/sec
[email protected] ~> sudo hdparm -tT /dev/sdb
/dev/sdb:
Timing cached reads: 16086 MB in 1.99 seconds = 8066.75 MB/sec
Timing buffered disk reads: 322 MB in 3.00 seconds = 107.30 MB/sec
[email protected] ~>
The SmartArray, however has and encourages SAS-1 (3 Gbps/375 MBps theoretical throughput). Since I decided I will be keeping the rack-mount for the long haul, I invested in a few SAS drives (4x 300GB, 10K RPM).
Question time.
While the SAS disks deliver, I wonder what RAID configuration would make the most sense. I was thinking RAID 10 all four (effectively oversaturating the SmartArray E200 at up to 440MBps unless there are two channels to split the load) for 600GB, or to RAID 5 three of the drives for 150-to-220 MBps throughput and have a spare - also making 600GB.
Then it was suggested to me to RAID 5 all four (if a drive fails, just means slower performance while running in that state and will be fine as long as a second does not fail - if I use the server at all in the meantime) for 225-to-330 MBps throughput and 900GB space.
Since the arrays' purpose is to provide a place for source code (lots of reads, including random ones, few writes during system load) as I will still use the SSD for ccache and out (a bit more about that later), which scheme would make the most sense? I am inclined to do the RAID 5 with all four drives now, even though I may need to do so with software (ZFS-on-Linux) since the E200 model can have the RAID 5 option turned off based on the cache size available (I should be fine though, luckily) or want to because the server is a WOL/on-demand machine so the cache battery protection is often drained/recharging for the first 50 minutes of boot anyway (I suppose the way around that for 90% of cases is to change the BIOS setting to power back on if power lost while operational if not already).
225-to-330 MBps throughput may be a reasonable enough improvement or hard to improve upon in the situation anyway, unless there is other important information in play. What do you think?
And if I ever get a hold of a spare SSD, I plan for a PCIe X4 SATA-3 riser to hook it up with. I would have got one by now, but it is doubtful they are bootable with this machine. All this and you may be wondering, why prop up a 10+ year old server? At the moment I have quotes prepared to replace my i5, and the current cost to do it the way I want is quite a bit more (ten-fold at least) than the $120 here, $50 there that I am doing with the server.
Many thanks.
Pinging @rwilco12 and @garwynn
A bit of an update:
Some significant details changed as well, but I have received the four drives and I used the SmartArray to create a RAID5 with all four (no spare), making a logical drive of 838 GiB. Transferred my 320 GB contents over (after making larger swap, /var and /home partitions) offline with a CloneZilla CD, rewrote the related UUID's in fstab and initramfs.
The new benchmarks (keep in mind I am currently downloading AOSP as well):
Code:
[email protected] ~> sudo hdparm -tT /dev/sdb
/dev/sdb:
Timing cached reads: 11008 MB in 2.00 seconds = 5515.37 MB/sec
Timing buffered disk reads: 1260 MB in 3.00 seconds = 419.86 MB/sec
[email protected] ~>
Because of the significant changes, there is a benefit to the SSD also:
Code:
[email protected] ~> sudo hdparm -tT /dev/sda
/dev/sda:
Timing cached reads: 11498 MB in 2.00 seconds = 5758.57 MB/sec
Timing buffered disk reads: 784 MB in 3.00 seconds = 261.23 MB/sec
[email protected] ~>
And where it is important, for a 70,296-object Android 8.1 build:
Code:
Package size: 327M
════════════════════════════════════════════════════════════════════════════════
#### build completed successfully (33:32 (mm:ss)) ####
[email protected]:~/carbon$
Because this is still a ProLiant server, I have been playing with the iLO interface - neat things in there.
Wait, did I say "still"?
Yeah, I found an unexpectedly good deal on a DL380 G6 (loaded with G7-era CPU's) that I was compelled to snatch up.
Related
Hey guy's I recently purchased a 8 GB MicroSD card and I got it in the mail today, so before putting it in my phone I put it on my computer and it doesn't read 8 GB it reads 7.59 GB. Why is this?
Capactities will never be exact. That is true of flash drives, hard drives etc. My iPod is theoretically 80GB but is actually only 76. Annoying...but that's just how it is
http://www.technical-assistance.co.uk/kb/hddsize.php
same thing really it's been the case for eons
sd cards
well they are true about how they always vary in capacity for posted size but for hardrives it usually actually is withing a couple percent, where as flash being very unstable memory can come with no memory lol, but also some of the memory is take up by foratting the device to work with operating systems and able to hold files and everything
That's because drive manufacturers measure like this:
1MB = 1000KB
1GB = 1000MB
And operating systems measure like this (the correct way):
1MB = 1024KB
1GB = 1024MB
So the card you bought has 8,000,000KB. The manufacturer claims that's 8GB. Windows sees it as...
8,000,000KB / 1024KB per MB /1024MB per GB = 7.629GB
Then a little bit of space is used for formatting, which leaves us with 7.59GB.
yeap!!
gtg465x said:
That's because drive manufacturers measure like this:
1MB = 1000KB
1GB = 1000MB
And operating systems measure like this (the correct way):
1MB = 1024KB
1GB = 1024MB
So the card you bought has 8,000,000KB. The manufacturer claims that's 8GB. Windows sees it as...
8,000,000KB / 1024KB per MB /1024MB per GB = 7.629GB
Then a little bit of space is used for formatting, which leaves us with 7.59GB.
Click to expand...
Click to collapse
Must of user needs to know how sizes arent exacts! just think about why it arent memoryes for 200MB over 256 or 32 not 30MB
I just got my GNex few days back and I'm new to this exotic way it manage its data.
I installed some apps and some games that are have downloaded their data to the sd-card and some other stuff like nandroid backups and some roms.
The thing is when I get into Settings-Storage the numbers just does't fit. I'm missing close to 3GB of space.
Where is my lost space here?
That's normal.. 16GB really means 13.3GB after all the system stuff, etc is taken.
Just like how on an iPhone/iPod 32GB is actually around 29GB of user-useable storage.
This part OK and I'm fine with this.
Try summing up the numbers to see if you get 13.33GB.
I'm getting 10.37GB when summing everything there including the free space, so there's something very weird here.
It is not so much that it is missing, its the way its measured. Manufactures measure hard drive space one way, OS's read hard drive space in different ways.
Computers do not internally represent HDD or memory capacity in powers of 1024; reporting it in this manner is just a convention.[39] Creating confusion, operating systems report HDD capacity in different ways. Most operating systems, including the Microsoft Windows operating systems use the powers of 1024 convention when reporting HDD capacity, thus an HDD offered by its manufacturer as a 1 TB drive is reported by these OSes as a 931 GB HDD. Apple's current OSes, beginning with Mac OS X 10.6 (“Snow Leopard”), use powers of 1000 when reporting HDD capacity, thereby avoiding any discrepancy between what it reports and what the manufacturer advertises.
Read more here
Hope this helps
All the sizes are reported by the same software and are in the same units: total space, free space and so are most of the other categories.
I still don't get how can it be.
Total reported by the phone: 13.33GB
Apps: 2.66GB
Pictures, Videos: 1.41GB
Audio: 20KB
Downloads: 88KB
Misc: 2.63GB
Free: 3.67GB
----------------------------
Total: 10.37GB
AFAIK 13.33GB != 10.37GB
Developers,
I got a free dedicated server the other day that is free for one month. I have little use for the server, so I figured I would share it with all of you for as long as it's free (because I don't intend to pay for it).
If you would like to build on it, please contact me. It can build ParanoidAndroid in just a little over an hour.
The specs are as follows:
Code:
Processor Intel Core i3 21302x2(HT)x3.4+ GHz
Architecture 64 bits
RAM 16 GB DDR3
Hard disk 2x 1 TB - SATA2
RAID SOFT - 0/1
NIC FastEthernet
I have a Zenfone 2 ze551ml that is getting sluggish. Many of my apps are on the external sdcard, which is a 128GB U1/class10 microSD card. I'm using link2sd for these apps.
My question is what experience people may have with using an U3 microSD card? Does U3 improve the overall speed of the phone much when many apps are linked to the external card? I do not have an U3 card to test with, so whatever data or impressions you may have is useful. Of course, the more scientific the results, the better.
I have been using busybox.top to see (via ssh login) what the phone is doing under load.
Mem: 1934648K used, 71860K free, 0K shrd, 28480K buff, 497300K cached
CPU: 40.8% usr 33.0% sys 4.5% nic 6.2% idle 10.9% io 2.0% irq 2.2% sirq
The io number is often small, but I probably does not include IO *wait* time, so I do not know how significant the number is. The "top" numbers are all over the place, from second to second, so I do not think they tell me much, but I have a strange hunch that the sdcard is one of the bottlenecks. Using 497MB of RAM as disk cache might occur whether or not the sdcard is fast. Another example of some "top" numbers, after a reboot.
Mem: 1935396K used, 71112K free, 0K shrd, 14932K buff, 392356K cached
CPU: 32.0% usr 22.9% sys 3.1% nic 15.1% idle 24.5% io 1.2% irq 0.9% sirq
Load average: 13.97 5.47 2.02 4/1903 5604
PID PPID USER STAT VSZ %VSZ CPU %CPU COMMAND
1933 319 app_207 R 1578m 80.4 1 20.8 {facebook.katana} com.facebook.katana
678 319 system S < 1474m 75.1 2 13.9 system_server
As a rule, the bigger the card, the slower it is, compared to the same series cards but smaller size.
I use 32GB card Kingston microSDHC 32GB UHS-I U3 (r/w;90/80MB/s), when i just installed it around 6 months ago it showed around 100MB/s speeds.
p.s. it gets worse over time.
Bump, anyone with specific experience in a general feeling or measurement of phone speeding up when switching from U1 to U3 speed class sdcard? I know a U3 card will have better IO benchmarks, but the real the question is how much does it affect performance of the phone overall.
I can report back that I bought a Netac 128GB micro-SD UHS-I class U3 card, and I have been using it in my phone for a while.
I would say that operations that involve the external U3 card are definitely faster, although I have not run a benchmark yet (having other problems at the moment, may get back to this later). Anyway, I think the U3 card was worth paying a bit extra ($49 U3 versus $40 or so for a U1, I think).
ive been messing around making user builds (roms) for the nexus 6 and the pixel xl, and ive gotten comfortable enough to start releasing my work, but before i do, i would like to get my desktop upgraded and actually set up.
Right now my system and /home is on a low rpm western digital hdd.
Im upgrading the pc all the way around, but my questions are regarding the disk drives. I have 2 of the exact same western digital blues, and a sandisk ssd.
western blue 500gb hdd
western blue 500gb hdd
sandisk 125gb ssd
Im aware disk speed is definitely the biggest bottleneck when it comes to building, so Here are my questions.....
How much space do you recommend when building for two devices while making builds weekly?
Do you think i should dedicate my ssd to /home?
Should i put the system on the ssd, and dedicate a hdd to the /home?
It appeared that I couldn't use the ssd for my home directory because (at a rough estimate) I'll be using around 200gb.
I found the best solution that gives me the best results.
I have split the windows and linux system on the ssd, I then ran the two hdd's in raid 0 and slpit the raid for /home and windows data.