Anyone poke around with one of these things? Their wording sounds a bit sneaky ("2000 Mbps throughput" when they mean full-duplex Gigabit Ethernet, "inherent" 200 Mbps limitation in other NAS, 1700 Mbps throughput on FC, etc... huh?). Looks like they are still new to this game.
http://www.bluearc.com/products/
I have spoken to several people that have one on demo. The all say it is the fastest machine they have ever tested. I spoke with eWeek about the numbers they were hired to do and they confirmed all tests were done without client side caching. There is no word on what features or functionality in is the software. MO
Brian Tao wrote:
Anyone poke around with one of these things? Their wording soundsa bit sneaky ("2000 Mbps throughput" when they mean full-duplex Gigabit Ethernet, "inherent" 200 Mbps limitation in other NAS, 1700 Mbps throughput on FC, etc... huh?). Looks like they are still new to this game.
http://www.bluearc.com/products/
-- Brian Tao (BT300, taob@risc.org) "Though this be madness, yet there is method in't"
not disrespecting the gentleman from nssolutions, but anyone here have any firsthand experience with testing and results of a bluearc system? It would be very interesting to hear some first hand observations rather than 2nd hand anecdotal statements that "They all say it is the fastest machine they have ever tested."
When you spoke to etesting labs (I take it you meant etesting, not eweek) did you ask them what performance looks like for other kinds of real world scenarios other than 60 machines, each doing 12 simultaneous writes/and reads of 57K files? It seems to me that writing and reading only 57K files won't exercise disk I/O much. It would be interesting to hear how the BlueArc system does with multiple simultaneous reads and writes of 100MB files across those 60 systems. It would be exciting if they can sustain 200MB/s aggregate read/write doing that kind of workload.
Iometer Settings Number of Outstanding I/Os per target = 12 Transfer request size = 59 KB Percent Random/Sequential Distribution = 100% Random Distribution Percent Read/Write Distribution = 50/50 Read/Write Align I/Os on: 59 KB Run time = 5 minutes with a ramp up time of 30 seconds
-----Original Message----- From: owner-dl-toasters@netapp.com [mailto:owner-dl-toasters@netapp.com]On Behalf Of Michael Owen Sent: Tuesday, May 15, 2001 9:56 PM To: Brian Tao Cc: toasters@mathworks.com Subject: Re: BlueArc Si7500 NAS
I have spoken to several people that have one on demo. The all say it is the fastest machine they have ever tested. I spoke with eWeek about the numbers they were hired to do and they confirmed all tests were done without client side caching. There is no word on what features or functionality in is the software. MO
Brian Tao wrote:
Anyone poke around with one of these things? Their wording soundsa bit sneaky ("2000 Mbps throughput" when they mean full-duplex Gigabit Ethernet, "inherent" 200 Mbps limitation in other NAS, 1700 Mbps throughput on FC, etc... huh?). Looks like they are still new to this game.
http://www.bluearc.com/products/
-- Brian Tao (BT300, taob@risc.org) "Though this be madness, yet there is method in't"
_________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
Personally, I'm eager to see the spec fs numbers. =)
-- Jeff
On Thu, May 17, 2001 at 06:46:38PM -0700, Ferd Berfl wrote:
not disrespecting the gentleman from nssolutions, but anyone here have any firsthand experience with testing and results of a bluearc system? It would be very interesting to hear some first hand observations rather than 2nd hand anecdotal statements that "They all say it is the fastest machine they have ever tested."
When you spoke to etesting labs (I take it you meant etesting, not eweek) did you ask them what performance looks like for other kinds of real world scenarios other than 60 machines, each doing 12 simultaneous writes/and reads of 57K files? It seems to me that writing and reading only 57K files won't exercise disk I/O much. It would be interesting to hear how the BlueArc system does with multiple simultaneous reads and writes of 100MB files across those 60 systems. It would be exciting if they can sustain 200MB/s aggregate read/write doing that kind of workload.
Iometer Settings Number of Outstanding I/Os per target = 12 Transfer request size = 59 KB Percent Random/Sequential Distribution = 100% Random Distribution Percent Read/Write Distribution = 50/50 Read/Write Align I/Os on: 59 KB Run time = 5 minutes with a ramp up time of 30 seconds
-----Original Message----- From: owner-dl-toasters@netapp.com [mailto:owner-dl-toasters@netapp.com]On Behalf Of Michael Owen Sent: Tuesday, May 15, 2001 9:56 PM To: Brian Tao Cc: toasters@mathworks.com Subject: Re: BlueArc Si7500 NAS
I have spoken to several people that have one on demo. The all say it is the fastest machine they have ever tested. I spoke with eWeek about the numbers they were hired to do and they confirmed all tests were done without client side caching. There is no word on what features or functionality in is the software. MO
Brian Tao wrote:
Anyone poke around with one of these things? Their wording soundsa bit sneaky ("2000 Mbps throughput" when they mean full-duplex Gigabit Ethernet, "inherent" 200 Mbps limitation in other NAS, 1700 Mbps throughput on FC, etc... huh?). Looks like they are still new to this game.
http://www.bluearc.com/products/
-- Brian Tao (BT300, taob@risc.org) "Though this be madness, yet there is method in't"
Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
eWeek and eTesting labs are one in the same. They would only speak about the tests they published and not any of the other scenarios they tested. Bluearc like everyone else wants to publish the best numbers possible. A 100 MB file is not something that most companies use on a daily basis. The 59 KB size is somewhat small, but not that abnormal. To put their 212 MB/sec claim into perspective this is 4x the fastest performance I have seen from a NetApp 840. I am sure that they would be happy to send you a demo if your company is looking for a new box or currently running a large number of NetApp boxes. They are fighting their way into every large NetApp customer they can find. I do not want to give the impression that I am defending BlueArc. I do find their numbers circumspect in that 212 MB/sec is faster than the theoretical maximum of the pipe they were using. This leads to the conclusion that there was client side caching occurring during the test. Your 100 MB file test would correct for this and give a true maximum throughput for the box. The larger issue that I see with the box is a small feature set. They do not offer Snapshot functionality and they have not listed any of the standards that they support (NDMP, DNS, WINS, etc.). The analogy here is that they have built a top fuel dragster. It is impressive on the track, but not very practical for everyday use. MO
Ferd Berfl wrote:
not disrespecting the gentleman from nssolutions, but anyone here have any firsthand experience with testing and results of a bluearc system? It would be very interesting to hear some first hand observations rather than 2nd hand anecdotal statements that "They all say it is the fastest machine they have ever tested."
When you spoke to etesting labs (I take it you meant etesting, not eweek) did you ask them what performance looks like for other kinds of real world scenarios other than 60 machines, each doing 12 simultaneous writes/and reads of 57K files? It seems to me that writing and reading only 57K files won't exercise disk I/O much. It would be interesting to hear how the BlueArc system does with multiple simultaneous reads and writes of 100MB files across those 60 systems. It would be exciting if they can sustain 200MB/s aggregate read/write doing that kind of workload.
Iometer Settings Number of Outstanding I/Os per target = 12 Transfer request size = 59 KB Percent Random/Sequential Distribution = 100% Random Distribution Percent Read/Write Distribution = 50/50 Read/Write Align I/Os on: 59 KB Run time = 5 minutes with a ramp up time of 30 seconds
-----Original Message----- From: owner-dl-toasters@netapp.com [mailto:owner-dl-toasters@netapp.com]On Behalf Of Michael Owen Sent: Tuesday, May 15, 2001 9:56 PM To: Brian Tao Cc: toasters@mathworks.com Subject: Re: BlueArc Si7500 NAS
I have spoken to several people that have one on demo. The all say it is the fastest machine they have ever tested. I spoke with eWeek about the numbers they were hired to do and they confirmed all tests were done without client side caching. There is no word on what features or functionality in is the software. MO
Brian Tao wrote:
Anyone poke around with one of these things? Their wording soundsa bit sneaky ("2000 Mbps throughput" when they mean full-duplex Gigabit Ethernet, "inherent" 200 Mbps limitation in other NAS, 1700 Mbps throughput on FC, etc... huh?). Looks like they are still new to this game.
http://www.bluearc.com/products/
-- Brian Tao (BT300, taob@risc.org) "Though this be madness, yet there is method in't"
Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
eWeek and eTesting labs are one in the same.
Actually no. eTesting labs are now an independent contract testing and QA laboratory. They used to be owned by Ziff Davis, and were responsible for the lab work behind the assorted product reviewlets that were published in the various ZD periodicals over the years, but ZD sold them off and now they are basically a "lab for hire". They'll do contract QA work, testing, report writing and publishing, whatever. We've subcontracted a little work to them ourselves on the NetCache side. They're good folks. See:
http://www.etestinglabs.com/main/reports/networkapp.pdf
A 100 MB file is not something that most companies use on a daily basis. The 59 KB size is somewhat small, but not that abnormal.
A tertiary scan of my systems would seem to indicate that you and I live in somewhat different worlds! :-)
To put their 212 MB/sec claim into perspective
Their claim is 219.658 MB/sec, at least as far as this test went.
this is 4x the fastest performance I have seen from a NetApp 840.
Have you ever done a memory-to-network test on an F840?
I do find their numbers circumspect in that 212 MB/sec is faster than the theoretical maximum of the pipe they were using.
You're forgetting about the full duplex nature of GbE, which will truck along at 125 MB/sec in each direction, so they're still leaving 38 MB/sec on the table when all the theoretical math is done. Actually, 30.342 MB/sec.
Anyway, we don't normally publish tests like this one that BlueArc commissioned, because we don't have many customers running their filers diskless. However, we do run memory-to-network & network-to-memory tests internally, as those numbers are of interest from the engineering perspective, especially to our engineering units responsible for tuning network stacks, drivers, low level I/O firmware and so forth.
To their credit, BlueArc have done a pretty good job in this area, because as far as the F840 goes, their numbers are only slightly lower than our own. We have the advantage of being able to use multiple GbE interfaces though, so "no fair" really.
Keith
Have you ever done a memory-to-network test on an F840?
What options are available on netapp to check memory-to-network thruput.
diskless. However, we do run memory-to-network & network-to-memory tests internally, as those numbers are of interest from the engineering perspective, especially to our engineering units responsible for tuning network stacks, drivers, low level I/O firmware and so forth.
I tested with following options with HP-UX & F840. dd if=/netapp/file1 of=/dev/null doesn't cross 30MB/s prealloc /netapp/file1 3GB doesn't cross 30MB/s
Yasin
-
Brian Tao -
Ferd Berfl -
Jeffrey Krueger -
Keith Brown -
Michael Owen -
Mohamed Yasin