Any experience with SATA port selectors?

I've found a couple of products on the market (from Marvell and Exar -
not clear whether they're chips or pluggable products), but not much
other information - e.g., whether these, or any others which may be
available, are something close enough to enterprise-grade to use in a
serious storage fail-over configuration. Even if they are, it would
seem that one would be required for each disk (or at least for each disk
in any mirror pair or parity group) to avoid making it itself a single
point of failure (IIRC the Exar product can be used behind a port
multiplier to handle multiple disks).

The Marvell product talks about a proprietary mode that allows fast
switching to load-balance between two host ports rather than fail over
per se - though I suspect that it might still fall short of what one
would need to treat the disk as concurrently-shared storage.

I've also not found any indication of how power is handled. Does it run
through the port selector (such that it fails over when the data stream
does), or are the disks assumed to have their own power supply (which
would be another single point of failure unless it was duplexed)?

A quick look at the LSI Logic website didn't turn up anything there
(somewhat to my surprise, but I guess they're pushing SAS a lot harder
right now).

In any event, any information would be appreciated.

- bill

In article PdednV_r8Y6SsOHYnZ2dnUVZ_tGsnZ2d@metrocastcablevi sion.com,
Bill Todd wrote:

A quick look at the LSI Logic website didn't turn up anything there
(somewhat to my surprise, but I guess they're pushing SAS a lot harder
right now).

Actually, they have at least one very innovative product that uses
SATA disks behind SAS port multipliers. It's an interesting approach.

--
Thor Lancelot Simon
"The liberties...lose much of their value whenever those who have greater
private means are permitted to use their advantages to control the course
of public debate." -John Rawls

Bill Todd wrote:
I've found a couple of products on the market (from Marvell and Exar -
not clear whether they're chips or pluggable products), but not much
other information - e.g., whether these, or any others which may be
available, are something close enough to enterprise-grade to use in a
serious storage fail-over configuration. Even if they are, it would
seem that one would be required for each disk (or at least for each disk
in any mirror pair or parity group) to avoid making it itself a single
point of failure (IIRC the Exar product can be used behind a port
multiplier to handle multiple disks).

The Marvell product talks about a proprietary mode that allows fast
switching to load-balance between two host ports rather than fail over
per se - though I suspect that it might still fall short of what one
would need to treat the disk as concurrently-shared storage.

I've also not found any indication of how power is handled. Does it run
through the port selector (such that it fails over when the data stream
does), or are the disks assumed to have their own power supply (which
would be another single point of failure unless it was duplexed)?

A quick look at the LSI Logic website didn't turn up anything there
(somewhat to my surprise, but I guess they're pushing SAS a lot harder
right now).

In any event, any information would be appreciated.

Disclaimer: I work for PMC-Sierra.

PMC-Sierra makes a component called the SPS (SATA Port Selector).
I don't know the product # off the top of my head. Using this
component
you can connect a SATA drive to 2 different SAS expanders (which
PMC-Sierra also makes). In turn, a different SAS HBA would be connected
to each SAS expander giving you your redundant connection. SATA drives
are single ported so it and the SPS are a single point of failure.
Until SATA
drives become dual ported I don't see any way around that.

Hope this helps.

Dave


- bill

dbs wrote:
Bill Todd wrote:
I've found a couple of products on the market (from Marvell and Exar -
not clear whether they're chips or pluggable products), but not much
other information - e.g., whether these, or any others which may be
available, are something close enough to enterprise-grade to use in a
serious storage fail-over configuration. Even if they are, it would
seem that one would be required for each disk (or at least for each disk
in any mirror pair or parity group) to avoid making it itself a single
point of failure (IIRC the Exar product can be used behind a port
multiplier to handle multiple disks).

The Marvell product talks about a proprietary mode that allows fast
switching to load-balance between two host ports rather than fail over
per se - though I suspect that it might still fall short of what one
would need to treat the disk as concurrently-shared storage.

I've also not found any indication of how power is handled. Does it run
through the port selector (such that it fails over when the data stream
does), or are the disks assumed to have their own power supply (which
would be another single point of failure unless it was duplexed)?

A quick look at the LSI Logic website didn't turn up anything there
(somewhat to my surprise, but I guess they're pushing SAS a lot harder
right now).

In any event, any information would be appreciated.

Disclaimer: I work for PMC-Sierra.

PMC-Sierra makes a component called the SPS (SATA Port Selector).
I don't know the product # off the top of my head. Using this
component
you can connect a SATA drive to 2 different SAS expanders (which
PMC-Sierra also makes). In turn, a different SAS HBA would be connected
to each SAS expander giving you your redundant connection.

The brochure appears to suggest that you could also connect the SPS to
two different SATA port multiplexers (each connected to a different
host's SATA port) to attain a similar level of redundancy (SAS seeming a
bit like over-kill for at least small fail-over systems).

SATA drives
are single ported so it and the SPS are a single point of failure.
Until SATA
drives become dual ported I don't see any way around that.

The drive would still be a single point of failure then as well. My
objective is to keep any single failure from making *multiple* drives
inaccessible for any significant period - at something approaching
commodity prices (which is why I was also interested in whether the port
selector failed over power as well: otherwise, one needs a redundant
source of power for the fail-overable disks even though one can use
commodity power supplies for their hosts; I haven't yet explored just
how expensive that would be).

Knowing where one could find pluggable components that incorporate your
chips would also be useful.


Hope this helps.

Yes - thanks. Apologies for replying both to the newsgroup and to your
email, but I don't know how often you frequent c.a.storage and thought
that others here might also be interested in the developing options in
this area.

- bill